Apply Data Science Introduction Applications And

Projects Thomas Barton download

https://ebookbell.com/product/apply-data-science-introduction-
applications-and-projects-thomas-barton-47492184

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Econometrics And Data Science Apply Data Science Techniques To Model

Complex Problems And Implement Solutions 1st Edition Tshepo Chris
Nokeri

https://ebookbell.com/product/econometrics-and-data-science-apply-
data-science-techniques-to-model-complex-problems-and-implement-
solutions-1st-edition-tshepo-chris-nokeri-35700072

Econometrics And Data Science Apply Data Science Techniques To Model

Complex Problems And Implement Solutions For Economic Problems 1st
Edition Tshepo Chris Nokeri

https://ebookbell.com/product/econometrics-and-data-science-apply-
data-science-techniques-to-model-complex-problems-and-implement-
solutions-for-economic-problems-1st-edition-tshepo-chris-
nokeri-36067372

15 Math Concepts Every Data Scientist Should Know Understand And Learn
How To Apply The Math Behind Data Science Algorithms 1st Edition David
Hoyle

https://ebookbell.com/product/15-math-concepts-every-data-scientist-
should-know-understand-and-learn-how-to-apply-the-math-behind-data-
science-algorithms-1st-edition-david-hoyle-58994148

Velocitybased Training How To Apply Science Technology And Data To

Maximize Performance Nunzio Signore

https://ebookbell.com/product/velocitybased-training-how-to-apply-
science-technology-and-data-to-maximize-performance-nunzio-
signore-33981280
Advances In Data Science And Computing Technology Methodology And
Applications 1st Edition Suman Ghosal Editor

https://ebookbell.com/product/advances-in-data-science-and-computing-
technology-methodology-and-applications-1st-edition-suman-ghosal-
editor-46502118

Advances In Data Science And Computing Technology 1st Suman Ghosal

https://ebookbell.com/product/advances-in-data-science-and-computing-
technology-1st-suman-ghosal-46508620

Proceedings Of 6th International Conference Of Young Scientists On

Solutions Of Applied Problems In Control And Communications Data
Processing And Data Analysis March 19 2015 Koethen Germany Perm Nat
Research Polytechnic Univ Perm
https://ebookbell.com/product/proceedings-of-6th-international-
conference-of-young-scientists-on-solutions-of-applied-problems-in-
control-and-communications-data-processing-and-data-analysis-
march-19-2015-koethen-germany-perm-nat-research-polytechnic-univ-
perm-43840028

Python Data Analysis Learn How To Apply Powerful Data Analysis

Techniques With Popular Open Source Python Modules Ivan Idris

https://ebookbell.com/product/python-data-analysis-learn-how-to-apply-
powerful-data-analysis-techniques-with-popular-open-source-python-
modules-ivan-idris-5472844

Data Engineering With Alteryx Helping Data Engineers Apply Dataops

Practices With Alteryx Houghton

https://ebookbell.com/product/data-engineering-with-alteryx-helping-
data-engineers-apply-dataops-practices-with-alteryx-houghton-56140498
Thomas Barton
Christian Müller
Editors

Apply Data
Science
Introduction, Applications and Projects
Apply Data Science
Thomas Barton · Christian Müller
Editors

Apply Data Science

Introduction, Applications and Projects
Editors
Thomas Barton Christian Müller
FB Informatik, Hochschule Worms FB Wirtschaft, Informatik, Recht (WIR)
Worms, Germany Technische Hochschule Wildau
Wildau, Germany

ISBN 978-3-658-38797-6 ISBN 978-3-658-38798-3 (eBook)

https://doi.org/10.1007/978-3-658-38798-3

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Fachmedien Wiesbaden
GmbH, part of Springer Nature 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does
not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors
or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

Responsible Editor: Petra Steinmueller

This Springer Vieweg imprint is published by the registered company Springer Fachmedien Wiesbaden GmbH,
part of Springer Nature.
The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany
Contents

Part I Introduction
1 Data Science: From Concept to Application. . . . . . . . . . . . . . . . . . . . . . . . . . 3
Thomas Barton and Christian Müller
1.1 What is Data Science?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 What is and What Does a Data Scientist?. . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Introduction to Data Science. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Systems, Tools and Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Part II Introduction to Data Science

2 Visualization and Deep Learning in Data Science. . . . . . . . . . . . . . . . . . . . . 13
Jens Kaufmann and Daniel Retkowitz
2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Methods for the Visual Preparation of Data. . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1 Representing Simple Data and Text. . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.2 Simplifying and Representing Complex Data. . . . . . . . . . . . . . . . 17
2.2.2.1 Matrixplot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2.2 Principal Component Analysis and
Multidimensional Scaling. . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2.3 t-SNE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Extract Image Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 Recognizing Visual Structures with Deep Learning . . . . . . . . . . . 23
2.3.2 Architectures for Practice. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Bringing Together Image and Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4.1 Generation of Enriching Detail Information . . . . . . . . . . . . . . . . . 24
2.4.2 Transformation of Visual Representations. . . . . . . . . . . . . . . . . . . 25
2.4.3 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

v
vi Contents

2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3 Digital Ethics in Data-Driven Organizations and AI Ethics as Application
Example. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Claudia Lemke, Dagmar Monett and Manuel Mikoleit
3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Data-driven Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.1 The concept of the Data-driven Organization . . . . . . . . . . . . . . . . 33
3.2.2 Technology Use of Data-driven Organizations . . . . . . . . . . . . . . . 34
3.2.3 Data-driven Corporate Culture. . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Digital Ethics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.1 Terminology and Moral Theories. . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.2 Overview of Digital Ethical Principles . . . . . . . . . . . . . . . . . . . . . 38
3.4 Digital Ethics and Data-driven Organizations. . . . . . . . . . . . . . . . . . . . . 39
3.4.1 Digital Ethical Principles and Data Value Creation. . . . . . . . . . . . 39
3.4.2 Consequences for the Design of Data-driven
Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.5 Case Study Deutsche Telekom AG: Operationalization of
AI Ethics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.1 The Company’s Motivation for Developing
Digital Ethics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.2 AI Ethics at the DTAG. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6 Summary and Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Multiple Perspectives for the Implementation of Innovative
Technological Solutions in the Context of Data-Driven
Decision-Making. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Anna-Maria Nitsche, Christian-Andreas Schumann, Christoph Laroque
and Olga Matthias
4.1 Why the Implementation of Innovative Technologies Requires a
Comprehensive Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2 Models from the Literature and their Weaknesses. . . . . . . . . . . . . . . . . . 52
4.3 The Technological and Organizational Coherence Implementation
Model (TOCI Model). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Benefits and Features of the TOCI Model. . . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Possible Useful Extensions to the TOCI Model . . . . . . . . . . . . . . . . . . . 59
4.6 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Contents vii

5 Don’t Be Afraid of Failure—Insights from a Survey on the Failure

of Data Science Projects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Jule Aßmann, Joachim Sauer and Michael Schulz
5.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2 Characteristics of and Hypotheses about Data Science Projects. . . . . . . 67
5.3 Design and Conduct of the Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.4 Evaluation of the Survey. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.5 Conclusion and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Part III Systems, Tools and Methods

6 Recommendation Systems and the Use of Machine Learning Methods . . . 79
Thomas Barton and Andreas Peuker
6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.2 Collaborative Recommendation Systems . . . . . . . . . . . . . . . . . . . . . . . . 82
6.2.1 Approaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.2.2 Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.3 Content-based Recommendation Systems. . . . . . . . . . . . . . . . . . . . . . . . 85
6.3.1 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.3.2 Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.4 Other Concepts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.4.1 Demographic Recommendation Systems . . . . . . . . . . . . . . . . . . . 87
6.4.2 Knowledge-based Recommendation Systems. . . . . . . . . . . . . . . . 88
6.4.3 Hybrid Recommendation Systems. . . . . . . . . . . . . . . . . . . . . . . . . 88
6.5 Current Developments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
7 Comparison of Machine Learning Functionalities of Business
Intelligence and Analytics Tools. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Gabriele Roth-Dietrich, Michael Gröschel and Benedikt Reiner
7.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2 Evaluation Framework for Business Intelligence Tools. . . . . . . . . . . . . . 97
7.2.1 Selection of BI Tools. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.2.2 Personas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.2.2.1 Persona 1: Expert. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.2.2.2 Persona 2: Layperson . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.2.3 Comparison Criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
7.2.4 Test Datasets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.3 Comparison of ML Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.3.1 SAP Analytics Cloud. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.3.2 Tableau Online/Tableau Desktop. . . . . . . . . . . . . . . . . . . . . . . . . . 104
viii Contents

7.3.3 Qlik Sense Business/Qlik Sense Desktop . . . . . . . . . . . . . . . . . . . 105

7.3.4 TIBCO Cloud Spotfire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
7.3.5 RapidMiner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.4 Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8 Using the Data Science Process Model Version 1.1. (DASC-PM v1.1)
for Executing Data Science Projects : Procedures, Competencies,
and Roles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Stephan Kuehnel, Uwe Neuhaus, Jens Kaufmann, Michael Schulz
and Emal M. Alekozai
8.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8.2 The Project Flow when Using DASC-PM v1.1. . . . . . . . . . . . . . . . . . . . 121
8.2.1 DASC-PM v1.1 at a Glance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
8.2.2 Project Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
8.2.3 Data Provision. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.2.4 Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.2.5 Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.2.6 Application. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
8.3 Overarching Key Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.4 Competence-driven Team Management Using Roles. . . . . . . . . . . . . . . 130
8.5 Concluding Remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Part IV Applications
9 Integration of Renewable Energies—AI-Based Prediction Methods for
Electricity Generation from Photovoltaic Systems . . . . . . . . . . . . . . . . . . . . 137
Boris Brandherm, Matthieu Deru, Alassane Ndiaye, Gian-Luca Kiefer,
Jörg Baus and Ralf Gampfer
9.1 Introduction and Motivation: Integration of Renewable Energies. . . . . . 138
9.2 Data Preparation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9.2.1 Data Collection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.2.2 Data Exploration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
9.2.3 Data Cleansing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9.2.4 Data Transformation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9.3 AI-based Prediction Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
9.3.1 Approaches Based on Artificial Neural Networks. . . . . . . . . . . . . 148
9.3.2 Approaches Based on Ensemble Machine Learning. . . . . . . . . . . 149
9.4 Fusion of Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
9.5 Application Examples and Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Contents ix

10 Machine Learning for Energy Management Optimization . . . . . . . . . . . . . 159

Gabriele Roth-Dietrich and Rainer Gerten
10.1 Digital Twin for an Air Conditioning System with Passive
and Active Heat Recovery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
10.2 Conception and Architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
10.3 Analysis and Evaluation of the Data Processing Steps. . . . . . . . . . . . . . 165
10.3.1 Step 1: Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
10.3.2 Step 2: Data Cleansing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
10.3.3 Step 3: Classify Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.3.4 Step 4: Filter Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
10.3.5 Step 5: Calculate Prediction. . . . . . . . . . . . . . . . . . . . . . . . . . . 172
10.4 Proof-of-Concept. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
10.4.1 Methods and Technologies Stack. . . . . . . . . . . . . . . . . . . . . . . 173
10.4.2 Visualization of results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
10.5 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
10.6 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
10.6.1 Further Analysis Approaches. . . . . . . . . . . . . . . . . . . . . . . . . . 176
10.6.2 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
11 Text Mining in Scientific Literature Evaluation: Extraction of
Keywords for Describing Content. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Thomas Barton and Arthur Kokoev
11.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
11.2 Explainable Artificial Intelligence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
11.3 Extraction of Keywords. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
11.4 Extraction of keywords for a literature review on
“Explainable AI”. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11.5 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
12 Identification of Relevant Relationships in Data Using
Machine Learning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Joshua Hammesfahr and Martin Spott
12.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
12.2 Expertise Problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
12.3 Approaches to Reducing the Number of Rules. . . . . . . . . . . . . . . . . . . . 193
12.3.1 Association Rule Discovery. . . . . . . . . . . . . . . . . . . . . . . . . . . 193
12.3.2 Subgroup Discovery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
12.4 Determining the Quality of Reduced Rule Sets. . . . . . . . . . . . . . . . . . . . 197
12.5 Combination Schema. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
12.6 Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
12.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
x Contents

13 Framework for the Management and Analysis of Vehicle Data for

Model-Based Driver Assistance System Development in Teaching
and Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Tobias Peuschke-Bischof and Stefan Kubica
13.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
13.2 Wildauer Maschinen Werke at TH Wildau . . . . . . . . . . . . . . . . . . . . . . . 209
13.3 Presentation of the Vehicle Fleet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
13.3.1 Trikes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
13.3.2 Trucks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
13.4 Introduction to Infrastructure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
13.4.1 ROS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
13.4.2 Node-RED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
13.4.3 MQTT-Bridge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
13.4.4 ROS Car2X. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
13.4.5 Traffic Light Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
13.4.6 VDI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
13.5 Development Framework. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
13.5.1 Implementation of Vehicle Communication. . . . . . . . . . . . . . . 215
13.5.2 Model-based Development and Code Generation
for Vehicles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
13.5.3 Agile Project Management, Knowledge Management
and Source Code Management. . . . . . . . . . . . . . . . . . . . . . . . . 220
13.6 Scenario-based Teaching and Research. . . . . . . . . . . . . . . . . . . . . . . . . . 222
13.6.1 ROS Car2X as Data Aggregation and Function
Behavior Across Vehicles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
13.6.2 NodeRED for Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 223
13.6.3 Interdisciplinary Scenario using the Example of
Material Management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
13.7 Summary and Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Editors and Contributors

About the Editors

Prof. Dr. Thomas Barton studied and obtained his doctorate at TU Kaiserlautern. He
then worked for SAP SE for approximately 10 years with a focus on application devel-
opment, consulting, training and project management. Since 2006 he has been working
at the University of Applied Sciences Worms as a professor of computer science with a
focus on business informatics. His areas of activity include the development of business
applications, e-business and data science. He is the author and editor of numerous pub-
lications. In addition, he is active in various committees and working groups. He is also
the spokesman for the GI Board of Universities of Applied Sciences.
Prof. Dr. Christian Müller studied mathematics at the Freie Universität Berlin and
received his doctorate in 1989 on network flows with side conditions. From 1990 to 1992
he worked for Schering AG and from 1992 to 1994 for the Berliner Verkehrsbetriebe
(BVG) in the area of travel
​​ and duty plan optimization. In 1994 he received a call to
the Technische Hochschule Wildau, Germany. He is a professor in the field of business
informatics and dean of the faculty of business, computer science and law. His research
focuses are the conception of information systems, mathematical optimization and the
simulation of business processes.

Contributors

Dr. Emal M. Alekozai Robert Bosch GmbH, Finance Shared Service Digital, Stuttgart, Germany
Jule Aßmann Hamburg, Germany
Prof. Dr. Thomas Barton Homburg, Germany
Dr. Jörg Baus Kognitive Assistenzsysteme, DFKI GmbH, Saarbrücken, Germany
Dr. Boris Brandherm Kognitive Assistenzsysteme, DFKI GmbH, Saarbrücken, Germany
Dr. Matthieu Deru Kognitive Assistenzsysteme, DFKI GmbH, Saarbrücken, Germany

xi
xii Editors and Contributors

Prof. Dr. Ralf Gampfer Hochschule Worms, Worms, Germany

Prof. Dr. Rainer Gerten Hochschule Mannheim, Mannheim, Germany
Prof. Dr. Michael Gröschel Hochschule Mannheim, Mannheim, Germany
Joshua Hammesfahr Walldorf, Germany
Prof. Dr. Jens Kaufmann Niederrhein University of Applied Sciences, FB Economics, Möncheng­
ladbach, Germany
Gian-Luca Kiefer Kognitive Assistenzsysteme, DFKI GmbH, Saarbrücken, Germany
Arthur Kokoev Worms, Germany
Prof. Dr. Stefan Kubica Zernsdorf, Germany
Dr. Stephan Kühnel Martin-Luther-Universität Halle-Wittenberg, Lehrstuhl für Wirtschaftsinformatik,
insb. Betriebliches Informationsmanagement, Halle (Saale), Germany
Prof. Dr. Christoph Laroque University of Applied Sciences Zwickau, Faculty of Economics,
Zwickau, Germany
Prof. Dr. Claudia Lemke Berlin School of Economics and Law (HWR Berlin), Department 2,
Berlin, Germany
Prof. Dr. Olga Matthias Leeds Beckett University, Leeds Business School, City Campus, Leeds,
UK
Manuel Mikoleit Gückingen, Germany
Prof. Dr. Dagmar Monett Hochschule für Wirtschaft und Recht Berlin (HWR Berlin),
Fachbereich 2, Berlin, Germany
Prof. Dr. Christian Müller Berlin, Germany
Dr. Alassane Ndiaye Kognitive Assistenzsysteme, DFKI GmbH, Saarbrücken, Germany
Dipl.-Inform. Uwe Neuhaus NORDAKADEMIE Hochschule der Wirtschaft, Elmshorn,
Germany
Anna-Maria Nitsche University of Applied Sciences Zwickau, Faculty of Economics, Zwickau,
Germany
Andreas Peuker Harthausen, Germany
Tobias Peuschke-Bischof Wildau, Germany
Benedikt Reiner Hochschule Karlsruhe, Karlsruhe, Germany
Prof. Dr. Daniel Retkowitz Hochschule Niederrhein, FB Wirtschaftswissenschaften,
Mönchengladbach, Germany
Prof. Dr. Gabriele Roth-Dietrich Hochschule Mannheim, Mannheim, Germany
Prof. Dr. Joachim Sauer NORDAKADEMIE Elmshorn, Fachbereich Informatik, Elmshorn,
Germany
Prof. Dr. Michael Schulz NORDAKADEMIE Hochschule der Wirtschaft, Elmshorn, Germany
Editors and Contributors xiii

Prof. Dr.-Ing. habil. Christian-Andreas Schumann Westsächsische Hochschule Zwickau,

Faculty of Economics, Zwickau, Germany
Prof. Dr. Martin Spott HTW Berlin, Berlin, Germany
 Part I
Introduction
Data Science: From Concept
to Application 1
Thomas Barton and Christian Müller

Contents

1.1 What is Data Science?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2 What is and What Does a Data Scientist?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Introduction to Data Science. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Systems, Tools and Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Abstract

This article provides an introduction to the topic of applying data science, which
appears in the series “Applied Business Informatics”. After a discussion of the terms
data science and data scientist, the article presents the contributions that appear in the
thematic areas. The starting point is an introduction to the field of data science. This is
followed by various articles on systems, tools and methods. One focus of this book is
the presentation of concrete applications.

T. Barton (*)
Homburg, Germany
e-mail: [email protected]
C. Müller
Berlin, Germanye-mail: [email protected]

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, 3

part of Springer Nature 2023
T. Barton and C. Müller (eds.), Apply Data Science,
https://doi.org/10.1007/978-3-658-38798-3_1
4 T. Barton and C. Müller

Keywords

Data Science · Data Scientist · Methods · Tools · Systems · Applications

1.1 What is Data Science?

A first visual representation describing the term data science based on the competencies
of the people who practice this discipline is shown in Fig. 1.1. According to Drew Con-
way, data science is shown as the intersection of three sets in a Venn diagram, with each
set representing a competence of the people who practice data science [1]. A competence
consists of hacking skills, which include abilities like editing text files at the command
line level and an understanding of vector operations, as well as algorithmic thinking.
Here, a hacker is seen as a person who enjoys having a deeper understanding of the inner
workings of a system, a computer, or especially a computer network [2]. Other compe-
tencies are mathematical and statistical knowledge. Substantial competencies complete
the profile. They should embody the willingness to discover and create knowledge for
which motivated questions about the world are needed, from which hypotheses can be
derived that can be verified using data.
From a statistician’s point of view, data science can be seen as a combination of sta-
tistics and computer science [3]. Statistics is one of the most important disciplines that
provides tools and methods to gain deeper insights from data; it is the most important
discipline to investigate and quantify uncertainty [4].
For the German Academy of Technical Sciences acatech, data science can be divided
into four core areas [5]: data engineering, data analytics, data prediction and machine
learning. Data engineering includes processes and methods for storing, accessing and
traceability of data, while data analytics focuses on data analysis. The prediction of top-
ics and situations based on experience is the subject of data prediction. Machine learning
is seen as a cross-sectional area to these three areas.

Fig. 1.1 Venn diagram for

data science based on [1]
Hacking Knowledge of
Abilities mathematics
and statistics

Data science

Substantive
competences
1 Data Science: From Concept to Application 5

A newer definition for data science, based on an interdisciplinary approach, dates

from 2017 [6]:

 Data Science = (Statistics + Applied Informatics + Computing + Communication + Soci-

ology + Management | (Data + Environment + Thinking)).

Data science is based on statistics, applied informatics, computing, communication,

management and sociology, with the term sociology standing for social aspects. The sub-
ject is the investigation of data and their environment, which includes in particular the
associated domains in order to transform data into knowledge and decisions. Here, think-
ing and methods should be used that make it possible to generate knowledge on the basis
of data. The vertical bar | means “under the condition that” and is known from condi-
tional probability.
The assessment of social aspects is based on morality. Morality can be understood
as the “total of observable behavior, behavioral norms and attitudes and values related
to behavior” [7]. The subject of ethics is morality. As a science, ethics is a discipline of
philosophy. Digital ethics is more or less seen as a synonym for information ethics, and
“information ethics has the morality of those as its subject who offer and use information
and communication technologies (ICT) and new media” [8].
Data science is also the subject of degree programs and continuing education offer-
ings at universities. Proposals for the content design of degree programs and continuing
education offerings for data science are available from the GI working group “Data Sci-
ence / Data Literacy” [9].

1.2 What is and What Does a Data Scientist?

Data Scientist is advertised as the most attractive job of the 21st century [10]. In 2015,
a Chief Data Scientist was appointed during the tenure of President Barack Obama [11].
But what does a Data Scientist do, and what qualifications are required? The duties of
a Data Scientist can be very diverse. In order to describe the extensive knowledge that
Data Scientists need to perform their jobs, a list of qualifications was drawn up and
assigned to different subject areas. A representation of the qualifications in the assigned
subject areas is shown in Table 1.1 [12].
Eighteen qualifications are assigned to five topic areas. Under the topic area Machine
Learning/Big Data, qualifications for structured, unstructured and distributed data are
also summarized. The topic area Mathematics/Operation Research also includes quali-
fications for algorithms, Bayesian statistics and Monte Carlo methods, graphical models
and simulations. The topic area Programming/System Administration also includes sys-
tem administration and programming of back-end and front-end. The topic area Statistics
includes statistics as well as qualifications for surveys and marketing as well as for visu-
alization.
6 T. Barton and C. Müller

Table 1.1  Topics for Data Scientists’ qualifications

Business/Product Machine Learning/ Mathemat- Programming / Statistics and
Development Big Data ics/Operation System- admin- Visualization
Research istration
- Business -B  ig Data and - Algorithms - Back-end - Statistics
- Product distributed data -B ayesian statis- p­ rogramming - S  urveys and
Development - Machine Learning tics and Monte -F  ront-end marketing
- Structured data Carlo methods programming - Visualization
- Unstructured data - Graphical models - System
- Mathematics administration
- Optimization
- Simulation

Four specializations can be derived for data scientists. Depending on the extent to
which the qualifications can be assigned to the five topic areas, the following specializa-
tions result [12]:

• Data-Business-Person: A person with a focus on qualifications for business

• Data Creative: A person for whom the qualifications are evenly distributed across the
five topic areas
• Data Developer: A person with a focus on qualifications for programming
• Data Researcher: A person with advanced qualifications for statistics

Good data scientists should [13]

• have technical expertise that can be demonstrated, for example, by completing a sci-
entific degree,
• be curious with a desire to discover and go into depth to break down a problem into
hypotheses that can be tested,
• practice storytelling by using data to tell a story and communicate it effectively,
• be clever to approach problems creatively and in different ways.

In addition, a team that carries out data science is characterized by intensive collabora-
tive work [14].

1.3 Introduction to Data Science

The introduction to the topic of data science begins in Chap. 2 with a contribution by the
authors Jens Kaufmann and Daniel Retkowitz. Their contribution entitled “Visualization
and Deep Learning in Data Science” introduces the visual presentation of data and the
extraction of information from images.
1 Data Science: From Concept to Application 7

Ethical considerations are playing an increasingly important role in the digital trans-
formation of companies. “Digital Ethics in Data-Driven Organizations and Their Appli-
cation to the Example of AI Ethics” is the title of the contribution by Claudia Lemke,
Dagmar Monett and Manuel Mikoleit. Digital ethics and its application in intelligent sys-
tems of a company is the subject of their contribution in Chap. 3.
Digital transformation leads to the implementation of technological solutions to sup-
port decision making. Anna-Maria Nitsche, Christian-Andreas Schumann, Christoph
Laroque and Olga Matthias present a process framework for the implementation and
evaluation of these technologies. Their contribution is entitled “Multiple Perspectives on
the Implementation of Innovative Technological Solutions in the Context of Data-Driven
Decision Making” and includes Chap. 4.
Investigations into the failure of data science projects are the subject of the contribu-
tion by Jule Aßmann, Joachim Sauer and Michael Schulz. Their contribution, which is
assigned to Chap. 5, is entitled “Don’t be afraid of failures—findings from a survey on
the failure of data science projects”.

1.4 Systems, Tools and Methods

The first contribution in this section comes from Thomas Barton and Andreas Peuker. In
their contribution in Chap. 6 entitled “Recommendation systems and the use of machine
learning methods”, the authors present the basics and the use of recommendation sys-
tems taking into account recent developments.
Gabriele Roth-Dietrich, Michael Gröschel and Benedikt Reiner compare BI systems
with regard to their ability to provide machine learning functionality for applications.
Chap. 7 consists of their contribution “Comparison of the machine learning functionality
of business intelligence and analytics tools”.
The contribution of the author team Stephan Kühnel, Uwe Neuhaus, Jens Kaufmann,
Michael Schulz and Emal M. Alekozai in Chap. 8 is about the use of a data science pro-
cess model. The title of the article is “Using the Data Science Process Model Version 1.1.
(DASC-PM v1.1) for executing Data Science projects: procedures, competencies and roles”.

1.5 Applications

The presented applications deal with the topics of renewable energy and optimization of
energy management, scientific literature evaluation, customer satisfaction in the automo-
tive industry and driver assistance system development.
The energy transition with the exit from nuclear energy and its challenges for the
electricity system in Germany are the starting point for the contribution by Boris Brand-
herm, Matthieu Deru, Alassane Ndiaye, Gian-Luca Kiefer, Jörg Baus and Ralf Gampfer.
The integration of renewable energies using the example of photovoltaic systems and the
8 T. Barton and C. Müller

use of artificial intelligence is the subject of their contribution “Integration of renewable

energies—AI-based prediction methods for electricity generation by photovoltaic sys-
tems” in Chap. 9.
Sustainability and a more efficient use of heat energy are the starting point for the
contribution by Gabriele Roth-Dietrich and Rainer Gerten. Their contribution “Machine
learning for energy management optimization” describes the optimization of an air con-
ditioning system using Data Science based on a concrete project. Chap. 10 consists of
this contribution.
Chapter 11 shows how the analysis of publications can be supported by text mining
as part of a scientific literature review. To do this, authors Thomas Barton and Arthur
Kokoev present a concrete application example. Their contribution is entitled “Text Min-
ing in a Scientific Literature Review: Extraction of Keywords to Describe Content”.
The use of data to determine relevant relationships is the subject of the contribution
by authors Joshua Hammesfahr and Martin Spott. As part of their contribution “Identifi-
cation of Relevant Relationships in Data Using Machine Learning”, they establish a con-
nection between, on the one hand, configurations of products or an infrastructure and, on
the other hand, errors in order to reduce errors. Chap. 12 presents this contribution.
The management and analysis of vehicle data in a digital learning factory is the sub-
ject of the contribution by Tobias Peuschke-Bischof and Stefan Kubica. Their contribu-
tion in Chap. 13 entitled “Framework for the Management and Analysis of Vehicle Data
for Model-based Driver Assistance System Development in Teaching and Research”
illustrates mechanisms in the development and project management.

References

1. Conway D (2010) The data science venn diagram. http://drewconway.com/zia/2013/3/26/

the-data-science-venn-diagram. Accessed: 21. Nov. 2022
2. Malkin G, Parker TL (1993) Request for comments 1392. Internet users’ glossary. https://
tools.ietf.org/html/rfc1392. Accessed: 21. November 2022
3. Kauermann G (2019) Data Science – Einige Gedanken aus Sicht eines Statistikers. Informatik
Spektrum 42(6):387–393
4. Weihs C, Ickstadt K (2018) Data science: the impact of statistics. Int J Data Sci Analytics
6:189–194
5. Gausemeier J, Guggemos M, Kreimeyer A (Hrsg) (2017) Auswahl, Beschreibung, Bewertung
und Messung der Schlüsselkompetenzen für das Technologiefeld Data Science. acatec. https://
www.acatech.de/wp-content/uploads/2019/02/acatech_NKM_Data_Science_WEB-2.pdf.
Accessed: 21. November 2022
6. Cao L (2017) Data science: a comprehensive overview. ACM Comput Surv 50(3): Article 43
7. Werner MH (2021) Einführung in die Ethik. Metzler. https://doi.org/10.1007/978-3-476-
05293-3
8. Bendel O (2019) 400 Keywords Informationsethik, 2nd edn. Springer Gabler, Wiesbaden
1 Data Science: From Concept to Application 9

9. GI (2019) Data Science: Lern- und Ausbildungsinhalte (Arbeitspapier). https://gi.de/fileadmin/

GI/Allgemein/PDF/GI_Arbeitspapier_Data-Science_2019-12_01.pdf. Accessed: 21. Novem-
ber 2022
10. Davenport TH, Patil DJ (2012) Data scientist: the sexiest job of the 21st century. Harv Bus
Rev 90(5):70–76
11. The White House (2015) The white house names Dr. DJ Patil as the first U.S. chief data sci-
entist. https://obamawhitehouse.archives.gov/blog/2015/02/18/white-house-names-dr-dj-patil-
first-us-chief-data-scientist. Accessed: 21. November 2022
12. Harris HD Murphy SP, Vaisman M (2013) Analyzing the analyzers. O’Reilly Media, Sevas-
topol
13. Patil DJ (2011) Building data science teams. O’Reilly Media, Sebastopol
14. Zhang AX, Muller M, Wang D (2020) How do data science workers collaborate? Roles, work-
flows, and tools. Proc ACM on Hum-Comput Interact 4(CSCW1):1–23

Thomas Barton studied and obtained his doctorate at the TU Kaiserlautern. He then worked for
SAP SE for approximately 10 years with a focus on application development, also consulting,
training and project management. Since 2006 he has been working at the Hochschule Worms as a
professor for computer science with a focus on business informatics. His areas of activity include
the development of business applications, e-business and data science. He is the author and editor
of numerous publications. In addition, he is active in various committees and working groups. He
is also engaged as a speaker of the GI Board for Universities of Applied Sciences.

Christian Müller studied mathematics at the Freie Universität Berlin and obtained his doctorate
in 1989 on network flows with side conditions. From 1990 to 1992 he worked at Schering AG and
from 1992 to 1994 at the Berliner Verkehrsbetriebe (BVG) in the area of timetable
​​ and service
planning optimization. In 1994 he was appointed to the Technische Hochschule Wildau, Germany.
He is a professor in the field of business informatics and dean of the Faculty of Economics, Com-
puter Science and Law. His research focus is on the conception of information systems, mathemat-
ical optimization and the simulation of business processes.
 Part II
Introduction to Data Science
Visualization and Deep Learning in Data
Science 2
Jens Kaufmann and Daniel Retkowitz

Contents

2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Methods for the Visual Preparation of Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1 Representing Simple Data and Text. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.2 Simplifying and Representing Complex Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Extract Image Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.1 Recognizing Visual Structures with Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 Architectures for Practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Bringing Together Image and Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4.1 Generation of Enriching Detail Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4.2 Transformation of Visual Representations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.3 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Abstract

People quickly and efficiently take in visually processed information. The process-
ing of data of any kind is therefore an important and heavily researched area in data
science and all its surrounding fields. The aim is to simplify complex information as

J. Kaufmann (*) · D. Retkowitz

Hochschule Niederrhein, FB Wirtschaftswissenschaften, Mönchengladbach, Germany
e-mail: [email protected]
D. Retkowitz e-mail: [email protected]

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, 13

part of Springer Nature 2023
T. Barton and C. Müller (eds.), Apply Data Science,
https://doi.org/10.1007/978-3-658-38798-3_2
14 J. Kaufmann and D. Retkowitz

much as possible so that the core information can be transported as simply and clearly
as possible without significant loss of meaning. In turn, there is an increasing need
to automatically process images and image information, whether for facial recogni-
tion as a biometric feature, for personal assistants, or for the evaluation of camera
images in driverless cars. The article shows what possibilities exist in each case and
how algorithms, e.g. in deep learning, can train and improve themselves or each other.

Keywords

Visualization · Deep Learning · Data Science · Generative Adversarial Networks ·

Dimension Reduction · Image Data

2.1 Introduction

The human brain has a core competence in pattern recognition. Its performance is so
high that modern methods of image and object recognition are based on a replica of this
organ, namely artificial neural networks (ANN). Deeply nested, with various adapta-
tions, additions and specializations, applications in the field of deep learning now deliver
results in a very short time that compete with those of the human brain and, for example,
exceed them in the evaluation of medically used photographs. At the same time, data sci-
ence is concerned in the opposite direction with the question of how complex data struc-
tures, multidimensional data sets and extensive analysis results can be designed to be as
appealing, understandable and interpretable as possible for the users of these systems.
This article shows in Sect. 2.2 which methods exist to visually prepare complex data
and how dimension-reducing methods use, for example, similarities in data and statisti-
cal distributions to avoid information loss. It describes which algorithm parameters the
users can adjust for which purpose and how the results can be interpreted in practice.
In contrast to data representation, deep learning can read structures and recognize
objects from visually available information. Here, too, users have extensive configuration
options to obtain answers to operational questions with justifiable (computational) effort.
The corresponding considerations and hints for practical applications can be found in
Sect. 2.3.
In the combination of image generation and image recognition, there are also so-
called Generative Adversarial Networks (GAN). They are an example of how ANN can
be used to create the most perfect copies of existing graphics or to reproduce familiar
people with “Deep Fakes”. Sect. 2.4 shows what effects this can have on good data visu-
alizations, but also on security aspects in companies.
2 Visualization and Deep Learning in Data Science 15

2.2 Methods for the Visual Preparation of Data

Data can take on various forms—also without visual preparation. Number series, text
data, financial tables, image data, etc. are just a few examples. This section is mainly
concerned with the graphical representation of data that originally exist in numerical
form. In addition, the visualization of text is dealt with. Only a black and white represen-
tation is shown here. While modern visualizations also convey information through the
use of colors and are especially aesthetically appealing, all forms of representation are
initially based on basic principles and structures that are usually color-independent and
can also be realized with gray tones. The advantage here is that a subsequent coloration
is always possible, but not required. For example, it can be ensured that the representa-
tions remain well readable even in black and white prints and regardless of the format. In
addition, the problem is avoided that it is not always known whether the graphic recipi-
ents have a color vision deficiency, for example a red-green weakness. In particular, the
focus of this article is therefore on the methodology and its understanding, not on exam-
ples of the most colorful graphics.

2.2.1 Representing Simple Data and Text

Simple structured data is usually available in large quantities in (business) everyday life.
Examples are sales figures, product properties, production times, gasoline consumption
or infection rates in the population of a country. Various standard diagrams are available
for the representation of these data, which are used in descriptive statistics. Since data
science is to a large extent the application of statistics, every data scientist should also be
familiar with the corresponding diagram types. Figure 2.1 shows six different diagrams
of a data set (without further description) as a selection.

Fig. 2.1 Six different standard plots of a dataset

16 J. Kaufmann and D. Retkowitz

The first one is a vertical bar chart (a). For several elements (shown on the x-axis)
the frequency of occurrence is given here on the y-axis. If the individual columns do
not represent individual elements (for example, individual people of a certain height),
but classes (155‒165 cm, 166‒175 cm …), it is a histogram. The dot diagram (b) is a
variant in which the individual entries are shown more clearly. Finally, the bar chart (c)
presents the rotated version by 90 degrees, which can be used as the standard variant for
most single value representations. Individual elements (here: countries) are listed from
top to bottom, their assigned values (here: gross domestic product; GDP) are depicted
by the length of the bars to the right. This representation allows quick and easy com-
parisons between values and follows the reading direction from left to right common in
many countries of the world.
The lower row shows a Boxplot (d) on the left, a compressed representation of a single
data set, in which location parameters such as the median and the quartiles can also be
seen (described by the “box”, the rectangle) as well as “outliers”, i.e. values classified as
unusually large (or small) (described by the individual points). Graphic e shows a scatter
plot. Here, individual data points are represented, which oppose two different variables to
each other. On the horizontal and vertical axes, the birth and death rates of the countries
of the world are plotted in this example, each point representing a country. A third infor-
mation dimension is represented by the color saturation of the point. The darker the point,
the higher the GDP of the respective country. Here, a fourth dimension, e.g. the popula-
tion, could be included by representing the points by differently sized circles. In general,
it should be noted that only information should be included in the visualizations that is
relevant for the respective statement to be made, since the reader is forced to interpret
each (unnecessary) complex representation, which contradicts the idea of simplifying data
interpretation. Finally, graphic f shows a Heatmap of the same data. Here, the available
space is divided into blocks. The more elements (“points”) there are in a block, the darker
it is colored. This makes it possible to quickly record and display higher-level patterns.
In addition to the variants presented here, there are various other representations for
specific purposes such as network diagrams. In addition, there are extensive considera-
tions of color variants, axis labeling, 3D representations, etc. Here, introductory works
on (descriptive) statistics (e.g. [1]) or compilations with a focus on data visualization
(e.g. [2]) should be consulted.
While numerical data is usually easy to visualize in different forms, there is a need to
quantify text in some form, that is, to provide it with numbers again, so that its proper-
ties can be expressed. This can be done, for example, by counting words, identifying key
terms, word lengths, etc.—depending on whether only individual texts or several texts
are to be displayed in comparison. The Word cloud has established itself as a widely used
standard tool for representing text content. It shows words in relation to their importance
(usually determined by the frequency of their occurrence) in a text in different sizes and
color intensity. Fig. 2.2 shows an example of a word cloud (created with the generator
2 Visualization and Deep Learning in Data Science 17

Fig. 2.2 Wordcloud of an article on data visualization and artificial neural networks (for creation
see [3])

from [3]). It is based on keywords that could occur in an article on data visualization
and artificial neural networks. It should be noted that a comprehensive pre-processing of
texts is usually necessary in order to remove different spellings, punctuation and mean-
ingless words (“a”, “the”, “and” …). The corresponding considerations are the subject of
a separate field of research in data science—text mining. For more information, see, for
example, [4].

2.2.2 Simplifying and Representing Complex Data

In data science, there is often a need to simplify complex data for presentation purposes.
Data that is already available in tabular form can become complex to view when three or
more variables (or: properties, dimensions, characteristics) are required to express the
relationship between multiple observations (or: objects, data elements). For illustration,
a dataset of German cities is used below, which clearly differ from each other in terms of
their geographic position (latitude, longitude), as well as certain properties (area, popula-
tion) or their combination (population per area). The problem can be transferred to any
number of variables. To illustrate the principles of the methods, this article is limited to
a dataset of approximately 50 to 200 cities, and the described maximum of five variables
18 J. Kaufmann and D. Retkowitz

(„lat”—latitude; “lon”—longitude; “area”—City area in km2; “population”—City popu-

lation; “pop per area”—Population divided by area) (for the origin of the data, see [5]).

2.2.2.1 Matrixplot
If a dataset with multiple variables is available, it can be helpful to first create all pos-
sible combinations of two variables and to represent them using the standard diagrams
described above, in particular scatter plots. If these diagrams are all combined, a matrix
plot results, as shown in Fig. 2.3 for the data set presented. The individual data points
are shown here semi-transparently in order to be able to detect overlaps and clustering of
data points.
There are 10 possible unique combinations of five variables, or 20 if the order of the
variables is relevant. Each row/column combination is then the two-dimensional repre-
sentation based on the two variables that are listed in the corresponding column and row.
The plot marked with “B” in this example therefore shows the combination of “popula-
tion” and “area”, with all plots in the “population” row showing the population data on
the vertical axis (scaling on the right-hand side) and the area on the horizontal axis in
plot B (scaling at the top). This representation makes it possible to find (suspected) rela-
tionships quickly. In the example, it can be seen that, as the area increases, the popula-
tion tends to increase as well. This is to be expected from a professional standpoint, but
the variety of such patterns can only be seen if many combinations can be quickly evalu-
ated. It should be noted that this representation cannot make any statements about more
complex relationships, e.g. if two variables influence a third variable in a certain com-
bination. Also, the scaling, or compression or stretching of the graphics may need to be
adjusted for interpretation. Plot A (top left) shows the data points (cities) as a combina-
tion of longitude and latitude. A more detailed view of the same plot including the names
of the cities can be found to the right in the graphic. Here it can be seen that, when the
ratio of the representation of the horizontal and vertical axes is correct, the individual
points quite accurately reproduce a map of Germany.

2.2.2.2 Principal Component Analysis and Multidimensional Scaling

In order to take into account combinations of variables and thus also reduce the num-
ber of graphics to be evaluated, the underlying data must be changed or reordered them-
selves. The principal component analysis determines linear combinations of the variables
in the data in order to be able to map the data as correctly as possible with a small num-
ber of principal components. Simply put, principal components can be considered “arti-
ficial” dimensions that are formed by cleverly combining the existing variables so that
as much variance, that is, dispersion or “diversity”, of the data is already given by using
a few of these artificial dimensions. In this way, data with many variables can be repre-
sented by a two-dimensional image, which is not “completely correct”, but “helpfully
sorted”. For the example, only three of the five available dimensions are chosen: “lat”,
“lon”, “pop per area”. A “map” like the one in the previous image is no longer possi-
ble, since the number of inhabitants per area should also be taken into account when
 2

A A
lat

lon

B
population
Visualization and Deep Learning in Data Science

area

pop per area

Fig. 2.3 Matrix plot of city data. (Data basis: [5])

19
20 J. Kaufmann and D. Retkowitz

d­ etermining the similarity of cities, which has to be represented graphically. Figure 2.4
shows the result of a corresponding PCA as a biplot, a representation of the actual data
based on the newly calculated principal components plus a representation of the con-
struction of these principal components.
The horizontal axis represents the first main component, the vertical axis the second.
Each data point (city name) is now mapped taking into account its similarity to other
cities. The positions on the axes are no longer directly interpretable. Rather, the arrows
drawn can help to find out which original variables have an influence on the construction
of the graphic in which form. Starting from the center, the extent to which the arrows of
the three variables extend to the left/right or up/down is observed. The arrow “pop per

lat

pop per area

lon

Fig. 2.4 Principal component analysis of city data. (Data basis: [5])
2 Visualization and Deep Learning in Data Science 21

area” points almost directly to the right. This means that cities with a high population
density tend to be located more to the right on the horizontal axis. At the same time, the
arrow from “lat” goes up and to the left, that is, in the negative direction with respect to
the horizontal axis. This means that cities with a high latitude (ie “northern” cities) will
be located more to the left in the graph. The observation for the vertical axis is analo-
gous. Here, the population density has hardly any influence. Northern cities (high value
for “lat”) tend to be located further up (positive influence), eastern cities (high value for
“lon”) tend to be located further down (negative influence).
The correct interpretation of the PCA requires some practice. It should also be noted
that the two artificial dimensions contain many, but not all, information about the data.
Principal component analyses serve, like the (partially related) multidimensional scaling,
to achieve a compressed representation of complex data. They help to identify superor-
dinate patterns and combinations, but do not provide a complete picture of the data. A
further introduction to the topic can be found, for example, in [6].

2.2.2.3 t-SNE
A form of dimension-reducing representation that has gained a lot of popularity is “t-dis-
tributed stochastic neighbor embedding” (t-SNE). Here, too, data points are represented
closer or further apart from each other in two or three dimensional space according to
their similarity. Unlike principal component analysis, however, similarity is based on
probability distributions [7]. For the pure application of the methodology, it is particu-
larly relevant to understand again that the representation does not show any directly
interpretable axis values. In addition, users have the option of parametrizing the proce-
dure using the so-called “perplexity”. The “perplexity” describes the influence of “local”
similarities (at small values of the “perplexity”) compared to “global” similarities (at
large values of the “perplexity”) of the data points in the preparation in low-dimensional
space. The parameter usually takes a value between 5 and 50, but its effect is strongly
dependent on the data basis and the size of the data set used. Therefore, smaller values
of the “perplexity” can also lead to results that are significantly different from other val-
ues for small data sets. Figure 2.5 shows the well-known data set, reduced to geographic
coordinates of cities with more than 200,000 inhabitants, in a t-SNE-based representa-
tion. While a “perplexity” of 3 (left) leads to clearly recognizable groups of local simi-
larities (e.g. Nuremberg, Augsburg, Munich), a value of 20 already results in a map of
Germany that is quite appropriate in terms of overall similarity (here tilted 45 degrees to
the left).
When interpreting, it should be noted that all the information in the representation
shown is available (unlike the two main components of PCA from many), but only a
distorted and “perplexity” -influenced approximation of the “actual” similarities can be
achieved. [8] provides a clear overview of the possible hurdles and assistance for correct
application.
 22

Fig. 2.5 t-SNE representation of city data. (Data basis: [5])

J. Kaufmann and D. Retkowitz
2 Visualization and Deep Learning in Data Science 23

2.3 Extract Image Information

In addition to data visualization, a key area of data science and machine learning is infor-
mation extraction from images. In the reverse direction, data is recovered from the visual
representation and structures are read out. So it is about analyzing and understanding
images automatically. This area is also called computer vision and did not arise only
in the course of the rapid development in the field of data science and machine learn-
ing (ML), but is a well-known discipline of computer science. Nevertheless, the recent
advances in artificial intelligence (AI) and ML have also had a significant impact in the
field of computer vision.

2.3.1 Recognizing Visual Structures with Deep Learning

The German Bitkom industry association has published a “periodic table of artificial
intelligence” (see [9]), which systematizes the various application areas of artificial intel-
ligence. 28 “AI elements” are distinguished, which are divided into the three groups
“Assess”, “Infer” and “Respond”. The three groups stand for typical processing steps of
an AI-based solution. A key part of the “Assess” group deals with the analysis of image
and audio data in which certain patterns are to be recognized and identified. In particular,
the field of image data analysis is of high importance in many application areas. Artifi-
cial neural networks (ANNs) have led to considerable progress in their various variants
and have gained a strong influence in the field of computer vision.
The task of classification can take on many different forms. In addition to evaluat-
ing situations, e.g. whether financial transactions are fraudulent activities, or predicting
future events, computer vision is about automatically capturing and interpreting visual
information. With the introduction of ML, new developments have been made possible in
this area of application that would not have been possible otherwise or only with tremen-
dous difficulties. Thanks to advances in hardware development and the use of special-
ized graphics processors for the calculation of ANNs, it has become possible to develop
more complex models with an ever increasing number of layers. In this context, one also
speaks of deep neural networks and deep learning.
In the meantime, ANNs have reached a level that allows them to perform comparably
to or better than humans in many scenarios. An example of this is the field of imaging
methods in medicine to support the diagnosis of various diseases. For example, doc-
tors can be supported in the analysis and evaluation of mammography images for cancer
diagnosis (cf. [10]) or in the diagnosis of Alzheimer’s disease (cf. [11]), to name just two
applications in this context.
In addition to classifying whole images, other important questions in the context of
computer vision are the detection of objects within an image or the segmentation of an
image into different areas as well as the analysis and interpretation of these.
24 J. Kaufmann and D. Retkowitz

2.3.2 Architectures for Practice

ANN are based on a mathematical model which simulates the functioning of biological
neurons in the brain of living beings in a greatly simplified form. In many approaches,
the McCulloch-Pitts neuron model, which was developed in 1943, is still in use today. In
this neuron model, the input signals of a neuron are added up and compared with a given
threshold value. If the threshold value is exceeded, the neuron emits a signal, otherwise
no signal is emitted. Various developments of this neuron model now make a variety of
AI algorithms possible, which are implemented in the form of different network architec-
tures.
As a standard procedure in the field of computer vision applications, so-called Con-
volutional Neural Networks (CNNs) have proven themselves. This type of ANN is char-
acterized in that the image data to be analyzed are first processed in several convolution
layers, which are trained to identify different structures in the image. This results in a
reduction and optimization of the data complexity in comparison to a pixel-wise view
and thus allows for more efficient processing.
The basic principle of the CNN is used in different network architectures that have
emerged over time for deep learning. The aim is to carry out ever more detailed and
accurate image analysis while simultaneously achieving efficient calculation. Known
architectures in the field of image classification, object recognition and segmentation,
which have also been taken up in many other works, are VGG, Region Based Convo-
lutional Neural Networks (R-CNNs), U-Net, Inception, GoogLeNet, Residual Networks
(ResNet), and You Only Look Once (YOLO).

2.4 Bringing Together Image and Data

The extraction of data from visual structures and, conversely, the representation of data
in visual form enable further applications in combination with each other. For these
approaches, different architectures of ANNs are used, which allow a reduction of details,
an enrichment with artificially generated details as well as a transformation from one
representation into another.

2.4.1 Generation of Enriching Detail Information

An example is the enrichment of images, e.g. by generating detail information in order

to achieve a higher image resolution. This is also referred to as “super resolution” and
deals with enriching images with missing detail information if, for example, they are
only available in low resolution or relevant objects have been recorded too small or are
blurry [12]. There is a whole range of different ANN architectures that can be used in
this area. Super resolution can also be used in medicine, but also to improve the visual
2 Visualization and Deep Learning in Data Science 25

quality of photos in general. Today’s photo editing software already includes functions
based on ML that, for example, can upgrade images through super resolution or com-
pensate for the deficiencies of the ever smaller camera hardware in mobile devices. How-
ever, it should be noted that with this technique the missing image information can only
be approximately reconstructed on the basis of an ANN and the image data used in the
training. Of course, it is not possible to exactly reconstruct the missing image informa-
tion afterwards.
Many more applications are based on generating additional or missing information.
This makes it possible to complete pictures with hidden areas [13] and to generate plau-
sible and very realistic looking image fragments that allow a suitable completion of an
incomplete image. For this purpose, in addition to CNNs, so-called recurrent neural net-
works (RNNs) are used. The approach can even be further developed in the sense that no
incomplete image is required as input, but instead keywords can be used to generate a
matching and completely artificially generated image [14].
However, it is not only possible to generate data artificially with image data, but also
with audio data [15]. For example, it is possible to implement text-to-speech applica-
tions with naturally sounding, artificially generated voices. With the same technique, it
is also possible to generate music in addition to speech generation, since the ANN works
directly on the unprocessed audio data and thus no prior knowledge of the type of audio
data determines the model.

2.4.2 Transformation of Visual Representations

The transformation of one image into another image is another task in which the essence
of a visual representation is extracted in order to subsequently generate a new visual rep-
resentation from it. This is the direct coupling of both processing directions. In this way,
for example, maps can be generated from aerial photographs or detailed images can be
derived from abstract line drawings [16].
For such purposes, often Generative Adversarial Networks (GANs) are used [17]. The
architecture of such a GAN is schematically shown in Fig. 2.6. It consists of two ANNs
that compete against each other in a so-called zero-sum game in the training process. The
first ANN (generator) realizes a generative model and has the task of artificially creat-
ing data sets that reproduce the real training data as accurately as possible. The second
ANN (discriminator) realizes a discriminative model, which has the task of reliably dis-
tinguishing the “fakes” of the generator from the real training data. The training pro-
cess now ensures that the generator always creates better “fakes” while the discriminator
becomes better and better at recognizing the “fakes”.
A well-known example of such a GAN is StyleGAN, which was developed by
researchers at NVIDIA. It is trained to generate as realistic artificial portraits of peo-
ple as possible [18]. In many cases, these images are hardly distinguishable from those
of real people. The approach can be transferred to many other areas. For example, it is
26 J. Kaufmann and D. Retkowitz

Original
Training data

Original or
fake?

Fake
Random data

Discriminator
(discriminative model)

Generator
(generative model)

Fig. 2.6 Schematic structure of a Generative Adversarial Network (GAN)

p­ ossible to generate artificial paintings that can convincingly imitate the different styles
of real artists [19].
However, the GAN architecture is not the only one used to generate visual represen-
tations. Another architecture used for this purpose are the so-called autoencoders [20].
These are primarily used to reduce the number of dimensions in data sets. In this way, a
high-dimensional feature space can be reduced to a manageable number of dimensions.
This can be used as a replacement for a principal component analysis (see Sect. 2.2.2.2)
for nonlinear relationships. Autoencoders also offer the possibility of generating artificial
data sets due to their architecture. With certain variants, very detailed images can also be
generated in this way [21].

2.4.3 Applications

The possibilities that arise from the different ANN architectures can be used for differ-
ent purposes. The enrichment of image information with details or the complete artificial
generation of image information is one of the core tasks in the entertainment industry,
e.g. in the production of films and video games. The requirements have been constantly
increasing and, in combination with applications of augmented reality and virtual reality,
are still further increasing.
An example application called “GauGAN”, based on the SPADE approach developed
by NVIDIA, shows how image synthesis can be simplified [22]. Tools like “GauGAN”
offer a variety of uses, for example in the development of video games, to create
2 Visualization and Deep Learning in Data Science 27

p­ hotorealistic scenes with the desired properties with much less effort than the manual
approach. Similar approaches enable the artificial creation of game worlds and can be
used for level design in the development of video games.
On the other hand, this also opens up many possibilities for misuse. Since the gener-
ated images are often hardly distinguishable from real images, they can be used very eas-
ily to deceive people. Such counterfeits, which are generated by means of deep learning,
are often also referred to as deep fakes. For example, the confidence in images, such as
photos in media reports, is lost, because any photo could also be a very realistic fake.
Another aspect in this context is the generation of manipulative image information
that is not used to deceive human observers, but which deliberately leads to incorrect
results in the classification by an ANN. It has been known for some time that it is effi-
cient to generate manipulated inputs for an ANN that lead to incorrect results with a
high degree of reliability [23]. The mapping between the input and output of an ANN is
highly unstable, which means that even small, barely visible changes to an input image
can result in an incorrect classification. From a security perspective, this property is very
problematic, and it therefore shows that in many security-relevant applications, special
precautions must be taken when using ANNs to prevent incorrect behavior with possibly
drastic consequences. In addition to all the impressive application possibilities, a great
challenge therefore lies in making ANNs safe from deliberate manipulation or uncon-
scious errors.

2.5 Summary

In order to understand complex data, people typically make use of various forms of
visual representation. This article has shown which common methods of data visual-
ization can be used for this purpose. By means of dimension-reducing methods, even
high-dimensional data can be transferred into a representation that makes the essential
structures and relationships in the data visible.
The opposite direction, i.e. extracting data from visual representations, has gained
particular importance through the development in the field of artificial neural networks
and deep learning. Different architectures of networks make it possible to learn which
essential structures in the visual representation can be used, for example, for the classifi-
cation of objects. The recognition of objects in images is a widely used application area
in which today’s results partially already supersede those achieved by human experts.
By linking both directions, further applications are possible that go beyond this.
Autoencoders and generative adversarial networks consist of several components that
can compress visual information and thus extract the essential structures from image data
on the one hand, and generate new visual representations from data on the other hand.
This structure is particularly suitable for generating realistic artificial image data that
hardly differ from real ones, or for transforming one visual representation into another.
28 J. Kaufmann and D. Retkowitz

In addition to useful applications such as in the entertainment industry, there is also the
danger of misuse for malicious purposes, to deceive and manipulate.
What applications are possible, desirable and ethically justifiable and how the abusive
use of the technical possibilities can be prevented remains an extremely relevant research
field in the future.

References

1. Brell C, Brell J, Kirsch S (2016) Statistik von Null auf Hundert, 2. Aufl. Springer, Berlin
2. Kahl T, Zimmer F (Hrsg) (2020) Interaktive Datenvisualisierung in Wissenschaft und
Unternehmenspraxis. Springer, Wiesbaden
3. Davies J (2020) Word cloud generator. https://www.jasondavies.com/wordcloud/. Accessed: 8.
Dec 2020
4. Provost F, Fawcett T (2017) Data Science for Business: What you need to know about data
mining and data-analytic thinking. ‎ O'Reilly and Associates
5. OpenGeoDB Downloads (2020) OpenGeoDb, Die freie Geoinformatik-Wissensdatenbank.
http://opengeodb.giswii.org/index.php?title=OpenGeoDB_Downloads&oldid=13822.
Accessed: 8. Dec 2020
6. Backhaus K, Erichson B, Plinke W, Weiber R (2018) Multivariate analysemethoden, 15th edn.
Springer, Berlin
7. Maaten L, Hinton G (2008) Visualizing data using t-SNE. J Mach Learn Res 9(2008):2579–
2605
8. Wattenberg et al (2016) How to use t-SNE effectively. Distill. https://doi.org/10.23915/dis-
till.00002
9. Bitkom e. V. (2018) Digitalisierung gestalten mit dem Periodensystem der Künstlichen Intelli-
genz. Ein Navigationssystem für Entscheider. Hg. v. Bitkom e. V. Berlin. https://www.bitkom.
org/sites/default/files/2018-12/181204_LF_Periodensystem_online_0.pdf
10. Abdelhafiz D, Yang C, Ammar R, Nabavi S (2019) Deep convolutional neural networks for
mammography: advances, challenges and applications. BMC Bioinformatics 20 (Suppl 11):
281. https://doi.org/10.1186/s12859-019-2823-4
11. Ding Y, Sohn JH, Kawczynski MG et al (2019) A deep learning model to predict a diagnosis
of alzheimer disease by using 18F-FDG PET of the brain. Radiology 290(2):456–464. https://
doi.org/10.1148/radiol.2018180958
12. Wang Z, Chen J, Hoi SCH (2019) Deep learning for image super-resolution: a survey. https://
arxiv.org/pdf/1902.06068
13. van den Oord A, Kalchbrenner N, Kavukcuoglu K (2016) Pixel recurrent neural networks.
https://arxiv.org/pdf/1601.06759
14. van den Oord A, Kalchbrenner N, Vinyals O, Espeholt L, Graves A, Kavukcuoglu K (2016)
Conditional image generation with pixelCNN decoders. https://arxiv.org/pdf/1606.05328
15. van den Oord A, Dieleman S, Zen H et al (2016) WaveNet: a generative model for raw audio.
https://arxiv.org/pdf/1609.03499
16. Isola, P, Zhu, J-Y, Zhou T, Efros AA (2016) Image-to-image translation with conditional
adversarial networks. https://arxiv.org/pdf/1611.07004
17. Goodfellow I, Pouget-Abadie J, Mirza M et al (2014) Generative adversarial nets. In:
Advances in neural information processing systems 27: annual conference on neural informa-
tion processing systems, pp 2672–2680. https://papers.nips.cc/paper/5423-generative-adver-
sarial-nets
2 Visualization and Deep Learning in Data Science 29

18. Karras T, Laine S, Aila T (2018) A style-based generator architecture for generative adver-
sarial networks. https://arxiv.org/pdf/1812.04948
19. Gatys LA, Ecker AS, Bethge M (2015) A neural algorithm of artistic style. http://arxiv.org/
pdf/1508.06576v2
20. Hinton GE, Salakhutdinov RR (2006) Reducing the dimensionality of data with neural net-
works. Science 313(5786):504–507. https://doi.org/10.1126/science.1127647
21. Razavi A, van den Oord A, Vinyals O (2019) Generating diverse high-fidelity images with
VQ-VAE-2. https://arxiv.org/pdf/1906.00446
22. Park T, Liu M-Y, Wang T-C, Zhu J-Y (2019) Semantic image synthesis with spatially-adaptive
normalization. https://arxiv.org/pdf/1903.07291
23. Goodfellow I, Shlens J, Szegedy C (2014) Explaining and harnessing adversarial examples.
https://arxiv.org/pdf/1412.6572

Prof. Dr. Jens Kaufmann holds the professorship for Business Information Systems—Data Sci-
ence at Hochschule Niederrhein University of Applied Sciences. Previously, he worked for several
years in consulting at Horváth & Partners and in the area of Global CIO at ERGO Group AG in
Düsseldorf. He lectured as a visiting professor at the University of North Carolina in Charlotte,
NC, USA, and focuses his teaching and research on the application of data science and its transfer
into business practice.

Prof. Dr. Daniel Retkowitz is Professor of Business Informatics, especially Software Engineer-
ing, at Hochschule Niederrhein University of Applied Sciences. He studied computer science at
RWTH Aachen University and Chalmers University of Technology in Gothenburg, Sweden. In his
dissertation, he worked on software development for smart homes. He then worked for an IT ser-
vice provider in the insurance industry as a project manager in system development and as an IT
architect and senior software developer, and lectured at the FH Aachen University of Applied Sci-
ences. In research and teaching, he is particularly active in the areas of software engineering and
machine learning.
Digital Ethics in Data-Driven
Organizations and AI Ethics 3
as Application Example

Claudia Lemke, Dagmar Monett and Manuel Mikoleit

Contents

3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Data-driven Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.1 The concept of Data-driven Organization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.2 Technology Use of Data-driven Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2.3 Data-driven Corporate Culture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Digital Ethics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.1 Terminology and Moral Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.2 Overview of Digital Ethical Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Digital Ethics and Data-driven Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4.1 Digital Ethical Principles and Data Value Creation. . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4.2 Consequences for the Design of Data-driven Organizations. . . . . . . . . . . . . . . . . . 40
3.5 Case Study Deutsche Telekom AG: Operationalization of AI Ethics . . . . . . . . . . . . . . . . . 42
3.5.1 The Company’s Motivation for Developing Digital Ethics. . . . . . . . . . . . . . . . . . . 42
3.5.2 AI Ethics at the DTAG. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6 Summary and Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

C. Lemke (*) · D. Monett

Department of Cooperative Studies Business and Technology, Berlin School of Economics and
Law (HWR Berlin), Berlin, Germany
e-mail: [email protected]
D. Monett
e-mail: [email protected]
M. Mikoleit
Deutsche Telekom AG, DTAG, Gückingen, Germanye-mail: [email protected]

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, 31

part of Springer Nature 2023
T. Barton and C. Müller (eds.), Apply Data Science,
https://doi.org/10.1007/978-3-658-38798-3_3
32 C. Lemke et al.

Abstract

Digital ethics focuses on the moral values and norms when dealing with modern tech-
nologies in the digital age, reflects on human actions in the design of those technolo-
gies and provides principles for their responsible use. Data-driven organizations live
from the value-creating data analysis for decision making. The ethically correct use of
such technologies becomes a key factor thereby. This chapter introduces data-driven
organizations, presents essential concepts of digital ethics and explains the implica-
tions for data-driven value creation on the basis of ethical principles. The example of
Deutsche Telekom AG demonstrates how digital-ethical principles are developed and
used concretely in an increasing data-driven use of intelligent systems in data-driven
organizations. It shows how AI ethics can successfully be applied in everyday busi-
ness life for creating a better way of dealing with these technologies.

Keywords

Data-driven organization · Data-driven culture · Digital ethics · Ethical principles · AI

ethics

3.1 Introduction

The volume of data generated worldwide is expected to grow to an estimated 175 zetta-
bytes per year by 20251. Out of it, 30% will be real-time data and more than half will be
generated by the Internet of Things. In addition, 60% of this data volume will be corpo-
rate data [1].
This immense amount of data potentially available to every company opens up new
opportunities for data analysis and evaluation as a prerequisite for their entire decision-
making processes, but also for generating new business models. The value of data for the
future success of a company is now acknowledged by the majority of companies, even
if only a few have so far been able to actually use its potential [2]. The reasons for this
are multidimensional and may require setting up a data-driven organization [3]. Usually,
these companies use the possibilities of data analysis and evaluation in all their areas in
order to gain deeper and more comprehensive insights about their structures and pro-
cesses. Correlations and relationships can be better identified. Decisions and the actions
associated with them are increasingly and often exclusively made on the basis of the
evaluation results from the processed data. Data and its analysis, thus, become the key
factor for success [3]. As a result, products and services that are tailored to the needs of
customers are created, which in turn become both a data source and a place of data pro-
cessing.

11 ZB (zettabyte) is approximately 1012 GB (gigabytes).

3 Digital Ethics in Data-Driven Organizations … 33

This chapter is dedicated to the topic of data-driven organizations. It provides an

overview of the effects of digital-ethical considerations, especially when using intelli-
gent systems for data-driven structures and processes and the resulting decisions for the
products and services of a company. Digital ethics deals with the moral reflection that
arises from the interactions of humans with modern technologies in the digital age [4].
In data-driven organizations, data processing is increasingly based on intelligent systems.
This increases the risks of harms or violations not complaint with digital-ethical princi-
ples, whether in the process of data collection and selection or in the application of the
underlying algorithms, and thus significantly affects the decisions for the value-creating
services of a company. The definition and compliance with universally valid digital-ethi-
cal principles is increasingly becoming a holistic corporate task for data-driven organiza-
tions in order to use such systems responsibly. The example of Deutsche Telekom AG
shows such a holistic approach that already starts with a digital-ethical consideration of
the company’s projects and can thus sustainably shape a data-driven corporate culture.
The contribution begins with a general understanding of the concept of a data-driven
organization and shows in particular the facets of a data-driven corporate culture as a
guarantee for success. It follows with an introduction to the field of digital ethics and the
presentation of basic moral theories as a basis for ethical understanding and digital-eth-
ical thinking. Key digital-ethical principles are briefly explained and related to a generic
decision-making process in data-driven organizations. The case study of Deutsche Tel-
ekom AG shows this process in a real, corporate-like implementation as good practice
for digital ethics in data-driven organizations.

3.2 Data-driven Organizations

3.2.1 The concept of Data-driven Organization

Data-driven organizations use data as one of their most valuable assets to make deci-
sions based on analyses and their evaluations, thereby creating new value for the com-
pany in the long term and thus having a tremendous impact on the business model as
well as on the products and services of the company. It is a central strategic goal and
alligns the values and attitudes within the company to this goal. The processes and struc-
tures created for data analysis and decision making change the company. Data-driven
organizations show high efficiency based on the entrepreneurial actions that arise from
the analysis and evaluation of data.
Organizations that refer to themselves as “data-driven” show higher productivity and
profitability than their competitors when examined more closely [5]. Data-driven organi-
zations, for example, can win new customers easier because they can analyze and evalu-
ate users and purchase behavior better. This makes it more likely to bind these customers
to the company in the long term.
34 C. Lemke et al.

Data-based organizations, on the other hand, use data as a product of the company
and, thus, regard it as part of their business model. These data-based organizations have
data-based business models, such as Netflix, Uber, Spotify or Instagram have as part of
their digital platforms [6]. Industrial companies such as General Electric or Siemens are
increasingly changing certain areas of their business towards data-based business models
in which, among other things, the manifold sensor data from development and produc-
tion, as well as the products, are part of the value creation [7].
For the necessary discourse on the digital-ethical implications for today’s companies
when dealing with modern technologies, these distinctions are rather secondary. Compa-
nies that increasingly base their decisions in a data-driven analysis use the possibilities
of further development towards data-based organizations as a consequence, resulting in
completely new, data-based business models. The following definition serves as the basis
for the explanations that follow:

 Definition A data-driven organization is a company that uses data to create value

• for the analysis and evaluation of structures, processes and performance of the
company for more efficient decision-making and/or
• for the further development and redesign of the company’s services in the form of
data-based products and service combinations

by the strategically-defined use of modern technologies of the digital age.

3.2.2 Technology Use of Data-driven Organizations

Data-driven organizations use data analysis and evaluation in all forms of decision mak-
ing to gain a more comprehensive and deeper understanding of their business, which
may increase both the efficiency of the business and the value of the business that is
generated. A comprehensive data analysis is also required to use the insights gained from
it for redesigning the business model in the sense of a data-based organization. There is
a general evolutionary development in the field of data analytics that is essentially char-
acterized by four stages, as Davenport has described with the terms “Analytics 1.0” to
“Analytics 4.0” [8].
The first form of data analytics (data analytics 1.0) as the classical form of reporting
in companies is the descriptive or descriptive data analytics, which shows limited rela-
tionships and connections inside the company and is therefore not the preferred form of
data analytics for data-driven organizations. It uses past data to find out what happened.
The forms of business intelligence currently in use in most companies are essentially
based on data analytics.
3 Digital Ethics in Data-Driven Organizations … 35

Data analytics 2.0 combines more comprehensive analytic methods that include
external data sources and use modern data management technologies for additional, new
insights that can already influence the performance of the company and thus the business
model. Data analytics 3.0 builds on these approaches by using an increasingly mixed
set of available internal and external data in any given structure as part of the company’s
strategy for decision making and support as well as performance transformation of the
entire company.
The first two stages use various forms of a predictive and prescriptive analytics, which
data analytics 3.0 applies compulsively for making use of all their advantages. They
shape the core technologies for data-driven organizations [9]. For predictive, forward-
looking analytics, both model building and simulation are necessary to find out what is
likely to happen. Various scenarios can be simulated by specifying parameters and deci-
sions in order to predict and determine the probability of certain events occurring. Pre-
scriptive analytics, on the other hand, not only collects data and evaluates existing data;
it also tries to answer the question, whether there is something that should be done opti-
mally, in order to achieve certain goals (corporate goals) in the best possible way.
For more advanced data analytics, data analytics 4.0 uses intelligent systems. Such
intelligent systems like logic-, knowledge- or connectionism-based systems [10] open up
a higher automation of decision-making processes, thereby reducing the direct involve-
ment of humans. More complex forms of machine intelligence could arise in these kind
of systems were they able to “adapt to their environment while working with insuffi-
cient knowledge and resources” [11]. This action requires a certain degree of autonomy,
in that the systems themselves select their resources and decide how they can achieve a
given goal. At the same time, the number of requirements increases, e.g. those dealing
with the correct selection and diversity of data, its quality and type of access, as well as
the quality of processing it by the algorithms. Those systems require advanced and spe-
cialized skills and competencies from individuals and even companies in the selection
and use of the right intelligent algorithms; they entail high-performance technologies for
processing huge amounts of data, and therefore might require a special embedding in the
corporate culture or when setting up a data-driven organization [8, 12].
These intelligent systems are no longer found only in data-driven organizations in
areas where the automation of decisions makes internal structures and processes more
efficient, but they are also increasingly present as examples of concrete instances or spe-
cific applications that are part of services in the form of product and service bundles that
contain facets of intelligence in machines. With regard to the case study presented in
this chapter, examples of such applications of artificial intelligence (AI) are chatbots and
software programs for the optimization of network planning, which include certain forms
of “intelligent” systems and use data for processing.
In Fig. 3.1 the systematics of data analytics in data-driven organizations is summa-
rized building upon Berndtsson et al. (2018, p. 2) [9]. It starts from specific business
requirements that determine the data selection and cleansing before they are analyzed
36 C. Lemke et al.

Analysis & Evaluation Corporate

Organization contexts
Business
requirements Data Technologies People Relationships

Processes Performance

Fig. 3.1 Data value creation of a data-driven organization

and evaluated. The results provide better and deeper insights into the company’s relation-
ships (also “business insights”). The data-driven insights so gained make it possible to
derive actions or measures that create value for the company. These insights can be used
to further increase the efficiency of internal structures and processes or for the provision
of services in the form of products and services.

3.2.3 Data-driven Corporate Culture

Leadership, governance, technology, and specific skills, among other factors, are seen as
barriers to a company’s transformation into a data-driven organization [13]. At the same
time, setting up a data-driven corporate culture is a key prerequisite for the success of a
data-driven organization [9, 14, 15]. The most important factors include, according to
Berndtsson et al. (2018, p. 3) [9, 16]:

• The implementation of a data governance that considers regulations for data collec-
tion, access to or making it available, as well as for data quality in order to set the
vision of what should be achieved with data-driven decision making in the company.
The existence of a data governance enables management to favor open and transpar-
ent communication about the goals of data analytics in the company. In addition, rules
must be defined for how, in interaction with external sources, data can be collected
and processed according to law, and to what extent data networks could be built and
used for comprehensive value creation.
• The implementation of a data-driven leadership means, on the one hand, the direct
responsibility of the top management for the institutionalized implementation of a
data-driven organization. It is the inner attitude of the corporate management towards
a data-driven organizational design. On the other hand, it requires the implementa-
tion of suitable management principles at all management levels in order to be able to
apply data-driven decision-making processes well.
3 Digital Ethics in Data-Driven Organizations … 37

• The “art” of decision-making requires a certain entrepreneurial and managerial atti-

tude towards the way data-based decisions are made and transferred into actions or
activities that also determine the design of the business model and the provision of
services. Testing plays a special role during the decision-making process from which
validated learning should take place. The decision-making process itself should be
traceable and applied correctly. In addition, and in combination with a data govern-
ance, acceptance should be achieved among all employees of the company regarding
data-driven decisions and how to bind them when implementing operational actions
according to which rules.
• The development of skills and competencies of data-driven decision making requires
training and further education for and of all management and application areas in
relation to modern data processing.
• The structures and processes must be set up and embedded in an organizational
way. The tasks, powers and competences as well as the processes should be docu-
mented and made accessible to all employees.
• The selection and use of suitable tools and technologies constitute an essential tech-
nical prerequisite for any form of data-driven decision making. The corresponding
tools must be accessible and usable, the benefits communicated and the correct use
trained upon. Technical questions about the design, operation and integration into the
company’s IT landscape are also relevant here.

3.3 Digital Ethics

3.3.1 Terminology and Moral Theories

Digital ethics defines and examines moral problems in relation to data and information,
algorithms and infrastructure in order to achieve good moral solutions [17]. Digital eth-
ics should give answers to which legitimate options for action arise from the develop-
ment, use and application of digital technologies (Grimm et al. 2019, p. 14) [18]. Ethics
itself addresses in the moral-philosophical sense the totality of values, norms and atti-
tudes under the term of morality. These values and attitudes are subject to a cultural
imprint and affiliation [19]. The cultural and social backgrounds determine the concrete
morality of a corresponding group of people.
Digital ethics reflects on our moral values in interaction with technology and, in this
particular context, with modern technologies of the digital age. Thus, the way humans
interact with modern technologies also depends on the cultural context. The question of
which moral attitudes and values people develop in interaction with technology can be
answered by looking at three of the most important theories of morality. This considera-
tion helps to understand the concrete design of certain digital technologies better.
38 C. Lemke et al.

These theories of morality include, on the one hand, duty ethics or deontological eth-
ics, which is preferred, for example, in Germany. It considers the actions themselves and
evaluates them in terms of compliance with or violation of basic moral duties. According
to this ethics, absolute values should be worked out for a digital ethics [18], i.e. those
after which the development and deployment of technology as well as its application or
use take place. In contrast, consequentialism and also utilitarianism as a special case of
it, consider the results or consequences of any action in terms of how they might affect
people [18]. Thus, the use or purpose of technology determine the ethical principles of a
digital ethics. A third theory of morality, virtue ethics, is centered on basic and universal
virtues. In this case, striving for a good, happy life determines the value of these tech-
nologies [20]. Whilst the first two theories of morality define basic principles according
to which either the action itself or the consequences of the action are evaluated, virtues
show how we may want to live together as human beings, in general, and how we can
use technology for good in the context of digital ethics, in particular. For a practical
application when dealing with technology, these theories show the reasons why certain
technological perceptions are considered right or wrong [21].

3.3.2 Overview of Digital Ethical Principles

Any concrete example of digital-ethical frameworks represents its inherent cultural con-
ceptions and sets of moral-theoretical priorities. As a result, it becomes clear that despite
their general or universal acceptance, digital technologies also carry the values and
norms of their developers as well as the moral conceptions of their respective cultural
circles within themselves. Digital technologies are not ethically neutral [22].
For data-driven organizations that make decisions based on data collection, data
management and data analytics, it is essential to know from which cultural circle the
data originates and which moral-theoretical principles are represented thereby, in order
to make decisions that are digitally and ethically right. The types of algorithms that are
used should also be subject to digital-ethical regulations, as they determine how the data
is processed, analyzed and evaluated. It is then necessary to define digital-ethical princi-
ples in order to obtain a practicable standard for the digital-ethical assessment of findings
and/or actions generated by data-driven decisions. Trust in and acceptance of automated
decisions is what creates responsible use (and design, development, deployment and
application) of modern technologies for the benefit of all.
An analysis of the global landscape of digital-ethical frameworks, mostly with direct
reference to the use of intelligent systems or, generally speaking, of systems that use AI,
shows different priorities when determining which ethical principles are considered [23].
For example, some ethical principles show a certain global consensus when weighted
according to their relevance and frequency of use in the analyzed frameworks. This way
they can generate trust and acceptance when dealing with intelligent systems across dif-
ferent cultures (including corporate culture) and moral values. These include, among oth-
ers (cf. Jobin et al. 2011, p. 7 ff. [23]):
3 Digital Ethics in Data-Driven Organizations … 39

• Transparency in the sense of explainability, comprehensibility and interpretability of

those systems, when communicating about them as well as the disclosure and demon-
stration of their functioning.
• Justice and fairness in the sense of consistency, inclusion, diversity, plurality, equal-
ity and equal treatment of all users of those systems, as well as the exclusion of (cog-
nitive) biases or any forms of discrimination. Furthermore, fairness and inclusion in
terms of availability, distribution and access to those systems.
• Harm prevention (non-maleficence principle/“non-maleficence”) in the sense of
safety and protection as well as prevention and integrity of the systems to avoid any
damage or harm to the users.
• Responsibility in the sense of liability of and obligation by the producers or develop-
ers of those systems towards the users.
• Privacy in the sense of the protection of personal and private information acquired,
used, processed, displayed in or by those systems.

These five ethical principles constitute the underlying framework for the design, devel-
opment, deployment and use of intelligent systems, in particular, and digital technolo-
gies, in general. In data-driven organizations, they are particularly suitable for deriving
how data can be used in a digital-ethical way, which digital-ethical principles the algo-
rithms that process it are subject to, and how data-driven corporate culture can take into
account moral values and norms. Based on this, specific requirements for the design and
application of technologies for data-driven organizations can be derived.

3.4 Digital Ethics and Data-driven Organizations

3.4.1 Digital Ethical Principles and Data Value Creation

Digital-ethical principles act in data-driven organizations not only as yet another neces-
sary success factor for a data-driven corporate culture. They also serve as basis when
dealing with digital technologies for data analytics, e.g. when addressing the design and
use of data and intelligent algorithms for the analysis and evaluation and, thus, for deci-
sion support in the products and services of a company.
With an increasing degree of autonomy in intelligent systems (from the level of data
analytics 3.0 onwards), the need for transparency or traceability of the decisions made
by the system, as well as for data value creation as a whole, increases. Data-driven
organizations should be able to explain how their decisions are made. They must also be
able to show which data is used for the analysis and evaluation and how they take place,
as well as through which steps and why the individual tasks are performed.
In addition, some automated decision-making systems tend to generate questionable
decisions and reinforce them. In data-driven organizations, justice and fairness with
regard to data access, data selection and data processing play a significant role when in
40 C. Lemke et al.

compliance with digital-ethical principles. Factors such as the discrimination of certain

groups of people or the transfer of cognitive biases already present in the data into deci-
sion making are currently the best-known forms of digital-ethical violations.
Moreover, data processing in or by intelligent systems should be designed and
implemented in a way that avoids physical, mental and economic harm to the users
or recipients of the decisions that are made. Harm prevention means for data-driven
organizations to align the systems for data analysis and evaluation such that they can
detect and prove that no harm (e.g. abuse, careless handling or insufficient risk manage-
ment) can be done when using these systems. Security and protection of intelligent sys-
tems are also subject to this premise.
Responsibility and accountability are not new to intelligent sysrtems, but especially
for data-driven organizations it becomes very clear that they have to explicitly define
the degree of responsibility through the entire data value creation. Topics like integrity,
clarification about the assignment of legal liability, as well as the usually contractually
regulated assumption of responsibility of all parties involved in the process of data value
creation (for example, software developers, data analysts, managers) are key to data-
driven organizations.
Privacy protection is a basic, inalienable human right; in dealing with data, it becomes
one of the central digital-ethical principles that data-driven organizations must adhere to
without exception. In addition to existing regulations, such as the European General Data
Protection Regulation (GDPR), this principle means to data-driven organizations that the
“right to data production and use” must be clarified in order to protect personal data. It also
includes proof of further use in both internal and external data value creation processes.

3.4.2 Consequences for the Design of Data-driven Organizations

Compliance with digital-ethical principles requires the extension of the data-driven cor-
porate culture by the factor of digital ethics (cf. Fig. 3.2).
Furthermore, general ethical principles must be transferred into applicability. The use
of valid regulations, such as the ethical guidelines for trustworthy AI of the European
Union [24], is suitable here in order to continuously check their compliance and effective-
ness in the form of governance. More than 200 organizations worldwide, including the
OECD (Organization for Economic Co-operation and Development), have now published
principles for the trustworthy, responsible and ethical use of AI or intelligent systems. The
time has come to operationalize these principles, because the increase of disruptive tech-
nologies and the mistrust of consumers towards algorithms force companies to answer
how a responsible use of technology should look like. The case study of Deutsche Tel-
ekom AG, introduced in short, shows such an operationalization of AI ethics.
For a data-driven organization, further standards and norms as well as general regu-
lations can be taken into account, especially in relation to data and algorithmic ethi-
cal questions which, through data governance or even further through a separate AI
 3

Data Governance Data-driven leadership Digital ethics

The 'art' of decision making
Transparency

Justice/Fairness

Skills and Structures

Digital Ethics in Data-Driven Organizations …

competences and processes Loss prevention

Responsibility

Technologies Tools Privacy

Fig. 3.2 Extended data-driven corporate culture by the factor “digital ethics”
41
Random documents with unrelated
content Scribd suggests to you:
 T h e P ro j e c t G u t e n b e r g e B o o k o f K i r s t e e n :
T h e S t o r y o f a S c o t c h F a m i l y S e v e n t y Ye a r s
Ago
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.

Title: Kirsteen: The Story of a Scotch Family Seventy Years Ago

Author: Mrs. Oliphant

Release date: January 16, 2019 [eBook #58707]

Most recently updated: January 24, 2021

Language: English

Credits: Produced by Chuck Greif and the Online Distributed

Proofreading Team at http://www.pgdp.net (This file was
produced from images available at The Internet Archive)

*** START OF THE PROJECT GUTENBERG EBOOK KIRSTEEN: THE

STORY OF A SCOTCH FAMILY SEVENTY YEARS AGO ***
Volume I.: Volume II.:
PART I. I. II. III. IV. I. II. III. IV. V. VI. VII. VIII.
PART II. V. VI. VII. VIII. IX. IX. X. XI. XII. XIII. XIV. XV.
X. XI. XII. XIII. XIV. XV. XVI. XVII. XVIII. XIX. XX.
XVI. XVII. XVIII. XIX. XX. XXI. XXII. XXIII.
XXI. XXII. XXIII.
COLLECTION
OF

BRITISH AUTHORS
TAUCHNITZ EDITION.

VOL. 2719.

KIRSTEEN BY MRS. OLIPHANT.

IN TWO VOLUMES.

VOL. I.
TAUCHNITZ EDITION.
By the same Author,
The Chronicles of Carlingford:
THE RECTOR AND THE DOCTOR’S FAMILY 1 vol.
SALEM CHAPEL 2 vols.
THE PERPETUAL CURATE 2 vols.
MISS MARJORIBANKS 2 vols.
PHŒBE, JUNIOR 2 vols.

THE LAST OF THE MORTIMERS 2 v.

MARGARET MAITLAND 1 v.
AGNES 2 v.
MADONNA MARY 2 v.
THE MINISTER’S WIFE 2 v.
OMBRA 2 v.
 MEMOIR OF COUNT DE MONTALEMBERT 2 v.
MAY 2 v.
INNOCENT 2 v.
FOR LOVE AND LIFE 2 v.
A ROSE IN JUNE 1 v.
THE STORY OF VALENTINE AND HIS BROTHER 2 v.
WHITELADIES 2 v.
THE CURATE IN CHARGE 1 v.
MRS. ARTHUR 2 v.
CARITÀ 2 v.
YOUNG MUSGRAVE 2 v.
THE PRIMROSE PATH 2 v.
WITHIN THE PRECINCTS 3 v.
THE GREATEST HEIRESS IN ENGLAND 2 v.
HE THAT WILL NOT WHEN HE MAY 2 v.
HARRY JOSCELYN 2 v.
IN TRUST 2 v.
IT WAS A LOVER AND HIS LASS 3 v.
THE LADIES LINDORES 3 v.
HESTER 3 v.
THE WIZARD’S SON 3 v.
A COUNTRY GENTLEMAN AND HIS FAMILY 2 v.
NEIGHBOURS ON THE GREEN 1 v.
THE DUKE’S DAUGHTER 1 v.
THE FUGITIVES 1 v.

KIRSTEEN
THE STORY
OF
A SCOTCH FAMILY SEVENTY YEARS AGO

BY
MRS. OLIPHANT,
AUTHOR OF “THE CHRONICLES OF CARLINGFORD,”
ETC. ETC.

COPYRIGHT EDITION.
IN TWO VOLUMES.—VOL. I.

LEIPZIG
BERNHARD TAUCHNITZ
1891.

INSCRIBED

WITH LOVE AND RESPECT

TO

CHRISTINA ROGERSON.

KIRSTEEN.
 PART I .

CHAPTER I.
“Where is Kirsteen?”
“ ’Deed, mem, I canna tell you; and if you would be guided by me you
wouldna wail and cry for Kirsteen, night and day. You’re getting into real ill
habits with her to do everything for you. And the poor lassie has not a
meenit to hersel’. She’s on the run from morning to night. Bring me this,
and get me that. I ken you’re very weakly and life’s a great trouble, but I
would fain have ye take a little thought for her too.”
Mrs. Douglas looked as if she might cry under Marg’ret’s reproof. She
was a pale pink woman seated in a large high easy-chair, so-called,
something like a porter’s chair. It was not particularly easy, but it was filled
with pillows, and was the best that the locality and the time could supply.
Her voice had a sound of tears in it as she replied:
“If you were as weak as I am, Marg’ret, and pains from head to foot, you
would know better—and not grudge me the only comfort I have.”
“Me grudge ye ainything! no for the world; except just that bairn’s time
and a’ her life that might be at its brightest; but poor thing, poor thing!” said
Marg’ret, shaking her head.
The scene was the parlour at Drumcarro, in the wilds of Argyllshire, the
speakers, the mistress of the house de jure, and she who was at the head of
affairs de facto, Marg’ret the housekeeper, cook, lady’s maid, and general
manager of everything. Mrs. Douglas had brought Marg’ret with her as her
maid when she came to Drumcarro as a bride some thirty years before; but
as she went on having child after child for nearly twenty years, without
much stamina of either mind or body to support that continual strain,
Marg’ret had gradually become more and more the deputy and
representative, the real substitute of the feminine head of the house. Not
much was demanded of that functionary so far as the management of its
wider affairs went. Her husband was an arbitrary and high-tempered man,
whose will was absolute in the family, who took counsel with no one, and
who after the few complaisances of a grim honeymoon let his wife drop
into the harmless position of a nonentity, which indeed was that which was
best fitted for her. All her active duties one by one had fallen into the hands
of Marg’ret, whose first tender impulse to save the mistress whom she
loved from toils unfitted for her, had gradually developed into the self-
confidence and universal assumption of an able and energetic housekeeper
born to organize and administer. Marg’ret did not know what these fine
words meant, but she knew “her work,” as she would have said, and by
degrees had taken everything in the house and many things outside it into
her hands. It was to her that the family went for everything, who was the
giver of all indulgences, the only person who dared speak to “the maister,”
when clothes were wanted or any new thing. She was an excellent cook, a
good manager, combining all the qualities that make a house comfortable,
and she was the only one in the house who was not afraid of “the maister,”
of whom on the contrary he stood in a little awe. A wife cannot throw up
her situation with the certainty of finding another at a moment’s notice as a
good housekeeper can do—even if she has spirit enough to entertain such
an idea. And poor Mrs. Douglas had no spirit, no health, little brains to
begin with and none left now, after thirty years of domestic tyranny and “a
bairntime” of fourteen children. What could such a poor soul do but fall into
invalidism with so many excellent reasons constantly recurring for adopting
the habits of that state and its pathos and helplessness? especially with
Marg’ret to fall back upon, who, though she would sometimes speak her
mind to her mistress, nursed and tended, watched over and guarded her with
the most unfailing care. Drumcarro himself (as he liked to be called)
scarcely dared to be very uncivil to his wife in Marg’ret’s presence. He
knew better than to quarrel with the woman who kept so much comfort with
so little expense in his spare yet crowded house.
“Who is your ‘poor thing, poor thing’?” said a cheerful voice, with a
mimicry of Marg’ret’s manner and her accent (for Marg’ret said poor as if it
were written with a French u, that sound so difficult to English lips) “would
it be the colley dogue or the canary bird or maybe the mistress of the
house?”
Marg’ret turned round upon the only antagonist in the house who could
hold head against her, or whom she could not crush at a blow—Kirsteen,
the second daughter, who came in at this moment, quite softly but with a
sudden burst open of the door, a sort of compromise between the noise it
would have been natural to her to make, and the quietness essential to the
invalid’s comfort. She was a girl of nearly twenty, a daughter of the hills,
strongly built, not slim but trim, with red hair and brown eyes and a
wonderful complexion, the pure whiteness like milk which so often goes
with those ruddy locks, and the colour of health and fine air on her cheeks. I
would have darkened and smoothed my Kirsteen’s abundant hair if I could,
for in those days nobody admired it. The type of beauty to which the palm
was given was the pale and elegant type, with hair like night and starry eyes
either blue or dark; and accordingly Kirsteen was not considered a pretty
girl, though there were many who liked her looks in spite of her red hair,
which was how people expressed their opinion then. It was so abundant and
so vigorous and full of curl that it cost her all the trouble in the world to
keep it moderately tidy, whereas “smooth as satin” was the required
perfection of ladies’ locks. Her eyes were brown, not nearly dark enough for
the requirements of the time, a kind of hazel indeed, sometimes so full of
light that they dazzled the spectator and looked like gold—also quite out of
accordance with the canons of the day. She was slightly freckled: she was,
as I have said, strongly built; and in the dress of the time, a very short
bodice and a very straight and scanty skirt, her proportions were scarcely
elegant, but her waist was round if not very small, and her arms, in their
short sleeves, shapely and well formed, and whiter than might have been
expected from their constant exposure to air and sun, for Kirsteen only put
on her gloves on serious occasions. The air of health and brightness and
vigour about her altogether, made her appearance like that of a burst of
sunshine into this very shady place.
“ ’Deed,” said Marg’ret, putting her hands on each side of her own
substantial waist in a way which has always been supposed to imply a
certain defiance, “it was just you yoursel’.”
“Me!” the girl cried with a sort of suppressed shout. She cast a laughing
glance round with an apparent attempt to discover some cause for the pity.
“What have I done wrong now?” Then her eyes came back to the troubled
almost whimpering pathos of her mother’s looks, and a cloud came over her
bright countenance. “What has she been saying, mother, about me?”
“She says I’m crying on you for something day and night, and that you
never have a minute to yourself; and oh, Kirsteen, my dear, I fear it’s true.”
Kirsteen put her arms akimbo too, and confronted Marg’ret with
laughing defiance. They were not unlike each other, both of them types of
powerful and capable womanhood, the elder purely and strongly practical,
the other touched with fancy and poetry and perhaps some of the instincts
of gentle blood, though neither in father nor mother were there many graces
to inherit. “You are just a leein’ woman,” said the girl with a flash of her
bright eyes. “Why, it’s my life! What would I do without my Minnie?—as
the song says.” And she began to sing in a fresh, sweet, but uncultivated
voice:

“He turned him right and round about,

Said, Scorn not at my mither,
True loves I may get mony an ane
But Minnie ne’er anither.”

Before Kirsteen’s song came to an end, however, her eyes suddenly

filled with tears. “What were you wanting, mother,” she said hastily as she
dropped the tune which was a very simple one, “to make her speak?”
“Oh, I was wanting nothing, nothing in particular. I was wanting my
pillows shifted a little, and the big plaiden shawl for my knees, and one of
my wires that fell out of my reach, and my other clew for I’m nearly at the
end of this one. Ay! that’s better; there is nobody that knows how to make
me comfortable but you.”
For Kirsteen in the meantime had begun to do, with swift and noiseless
care, all that was wanted, finding the clew, or ball of worsted for the
stocking her mother was knitting, as she swept softly past to get the big
shawl, on her way to the side of the chair where she arranged the pillows
with deft accustomed skill. It did not take a minute to supply all these
simple requirements. Marg’ret looked on, without moving while all was
done, and caught the look half-soothed, half-peevish, which the invalid cast
round to see if there was not something else that she wanted. “You may put
down that book off the mantelpiece that Robbie left there,” Mrs. Douglas
said, finding nothing else to suggest; “it will curl up at the corners, and your
father will be ill-pleased—”
“Weel,” said Marg’ret, “now ye’ve got your slave, I’m thinking ye’ve
nae mair need of me, and there’s the grand supper to think of, that the
maister’s aye sae keen about. When will ye have markit a’ thae things, Miss
Kirsteen? For I maun see to the laddie’s packing before it’s ower late.”
“There’s the last half dozen of handkerchiefs to do; but I’ll not take long,
and they’re small things that can go into any corner. I’ll do them now,” said
Kirsteen with a little sigh.
“There’s nae hurry;” Marg’ret paused a little, then caught the girl by the
sleeve, “just take another turn in the bonnie afternoon before the sun’s
down,” she said in a low tone, “there’s plenty of time. Run away, my bonnie
lamb. I’ll see the mistress wants naething.”
“And you that have the supper and the packing and all on your hands!
No, no. I’ll do them now. You may go to your work,” said Kirsteen with a
look half tender, half peremptory. She carried her work to the window and
sat down there with the white handkerchiefs in her hand.
“And what colour will you mark them in, Kirsteen? You have neither
cotton nor silk to do it.”
Kirsteen raised her head and pulled out a long thread of her red hair. “I
am going to do it in this colour,” she said with a slight blush and smile. It
was not an unusual little piece of sentiment in those days and the mother
accepted it calmly.
“My colour of hair,” she said, smoothing with a little complaisance her
scanty dark locks under her cap, “was more fit for that than yours, Kirsteen,
but Robbie will like to have it all the same.”
Kirsteen laughed a little consciously while she proceeded with her work.
She was quite willing to allow that a thread of her mother’s dark hair would
be better. “I will do one with yours for Robbie,” she said, “and the rest with
mine.”
“But they’re all for Robbie,” said the mother.
“Yes, yes,” Kirsteen replied with again that conscious look, the colour
mantling to her cheeks, a soft moisture filling her eyes. The handkerchief
was marked in fine delicate little cross stitches upon the white cambric, and
though Mrs. Douglas’s dark hair was like a spider’s web, the red of
Kirsteen’s shone upon the fine fabric like a thread of gold.
The handkerchiefs were not yet finished when two young men came into
the room, one so like Kirsteen that there was no difficulty in identifying him
as her brother, the other a swarthy youth a little older, tall and strong and
well knit. Robbie was on the eve of his start in life, leaving home, and
Ronald Drummond, who was the son of a gentleman in the neighbourhood,
was going with him. They were both bound with commissions in the
Company’s service for India, where half of the long-legged youths, sons of
little Highland lairds and Lowland gentlemen, with good blood and plenty
of pride and no money, the Quentin Durwards of the early nineteenth
century, found an appropriate career. The period was that of the brief peace
which lasted only so long as Napoleon was at Elba, long enough, however,
to satisfy the young men that there was to be no chance of renewed fighting
nearer home and to make them content with their destination. They had
been bred for this destination from their cradles, and Robbie Douglas at
least was not sorry to escape from the dullness of Drumcarro to a larger life.
Several of his brothers were already in India, and the younger ones looked
to no other fate but that of following. As for the girls they did not count for
much. He was sorry to say good-bye to Kirsteen, but that did not weigh
down his heart. He was in high excitement, eager about his new outfit, his
uniform, all the novel possessions which were doubly enchanting to a boy
who had never before possessed anything of his own. He was eighteen, and
to become all at once a man, an officer, an independent individuality, was
enough to turn the head of any youth.
Ronald Drummond was different. He was going from a much more
genial home: he had already tasted the sweets of independence, having
served in the last campaign in the Peninsula and been wounded, which was
a thing that raised him still higher in the scale of life than the three years’
advantage in respect of age which he had over his young comrade. He was
neither so cheerful nor so much excited as Robbie. He came and stood over
Kirsteen as she drew closer and closer to the window to end her work
before the light had gone.
“You are working it with your hair!” he said, suddenly, perceiving the
nature of the long curling thread with which she threaded her needle.
“Yes,” she said, demurely, holding up her work to the light. “What did
you think it was?”
“I thought it was gold thread,” he said. And then he took up one of the
handkerchiefs already completed from the table. “R. D.,” he said. “That’s
my name too.”
“So it is,” said Kirsteen, as if she had now discovered the fact for the
first time.
“Nobody will do anything like that for me,” he added, pathetically.
“Oh, Ronald! if not the hairs of their heads but the heads themselves
would do ye good ye should have them—and that ye know.”
 “It is very true,” said Ronald, “and thank you, Kirsteen, for reminding
me how good they are; but,” he added, after a moment, in a low voice,
“they are not you.”
She gave vent to a very feeble laugh which was full of emotion. “No,
they could not be that,” she said.
“And R. D. is my name too,” said the young man. “Kirsteen!” She
looked up at him for a moment in the light that was fading slowly out of the
skies. He had taken one of the handkerchiefs from the pile, and touching her
sleeve with one hand to call her attention, put the little glistening letters to
his lips and then placed the handkerchief carefully in the breast pocket of
his coat. Standing as he did, shutting out, as she complained, all the light
from Mrs. Douglas, this little action was quite unseen, except by the one
person who was intended to see it. Kirsteen could make no reply nor
objection, for her heart was too full for speech. Her trembling hand, arrested
in its work, dropped into his for a moment. He whispered something else,
she scarcely knew what—and then Marg’ret marched into the room with the
two candles which were all the lights ever used in Drumcarro parlour, and
all was over and done.

CHAPTER II.
There was “a grand supper,” as Marg’ret had announced, at Drumcarro this
evening, for which, though it was almost entirely a family party, solemn
preparations were being made. The house was full of an unusual odour of
good cheer, unusual goings and comings through the house betrayed the
excitement and sense of a great event approaching which was diffused
through the family. On ordinary occasions the family dinner took place
between two and three o’clock in the afternoon, followed by tea at seven
with much wealth of scones and jam, new-laid eggs and other home
produce—and the day ended for the elders by the production of “the tray”
with its case of spirit-bottles and accompanying hot water. Now and then by
times, however, this great ceremonial of a supper took place, always on the
eve of the departure of one of the boys to make their fortune in the world.
These occasions were consequently not surrounded by the brightest
recollections to the grown-up portion of the family, or to their mother. The
supper indeed to her was a feast of tears, probably as great, though a more
usual indulgence than the other characteristics of the festival. It was rarely
that Mrs. Douglas ventured to weep in presence of her lord, but on that
night he said nothing, made no comment upon her red eyes, and suffered
the whimper in her voice without any harsh, “Hold your tongue, woman!”
such as usually subdued her. And it was recognized in the house that it was
the mother’s rôle and privilege on these occasions to cry. The children were
not disturbed by it as they might have been by tears which they were less
accustomed to see shed.
The dining-room was the best room in Drumcarro, as in many Scotch
houses of the kind, being recognized as the real centre of life, the special
room of “the maister” and the scene of all the greater events in the family.
There were two windows in it which at a time when the existence of the
window-tax curtailed the light, was of itself a fine feature, and it was well-
sized and not badly furnished, with a multitude of substantial mahogany
chairs, sideboard, cellaret, and a long dining table of very dark mahogany,
shining like a black mirror, which was capable of being drawn out to almost
any length, and which had attained the very highest polish of which wood
was capable. Covered with a dazzling white cloth, lighted with four
candles, a most unusual splendour—set in the silver candlesticks, which
were the pride of the family—and surrounded by all the Douglases who still
remained at home, it was an imposing sight. Flowers had not yet been
thought of as decorations of a table; such frivolities were far in the depths
of time. A large square dish set in a high stand of plated silver with
straggling branches extending from it on every side, each of which
contained a smaller dish full of confectionery, pieces of coloured “rock”
from Edinburgh, and sweeties procured from “the merchant’s” for the
occasion, occupied the centre of the table. It was called the épergne and was
considered very splendid. The central dish was piled high with ruddy
apples, which gave an agreeable piece of colour, if any one had thought of
such fantastic folly. The four candlesticks, each with a pair of snuffers in its
tray placed between them, completed the decorative portion of the table.
The candles were not the delicate articles which advancing civilization has
learned how to produce, but smoky “moulds” which tinged the atmosphere
with a perceptible emanation, especially when they stood in need of
snuffing. They threw a ruddy light upon the faces closely assembled round
the board, bringing out most fully those of the more youthful members of
the family, and fading dismally towards the ends of the long table at which
the principal personages were placed. There were but two visitors of the
party, one the minister, invited in right of having more or less superintended
Robbie’s studies, such as they were, and seated on Mrs. Douglas’s right
hand; the other an old Miss Douglas known as Aunt Eelen, from whom
there were certain expectations and who occupied a similar place of honour
by the side of Drumcarro. The hero of the evening was at his father’s left
hand. The rest of the party were Mary the eldest daughter, Jeanie the
youngest, Kirsteen, and two boys aged fourteen and twelve respectively, the
remaining sons of the house. The fare was excellent, and in another region
might have been thought luxurious; but it was impossible to conceal that the
large dish of delicious trout which stood smoking before Mrs. Douglas, and
the corresponding hecatomb of grouse to which her husband helped the
company after the trout had been disposed of, came from the loch and the
moor on Drumcarro estate, and therefore were as much home produce as
the eggs and the cream. This fact elicited a somewhat sharp criticism from
Miss Eelen at the foot of the table.
“The grouse is no doubt very good,” she said, “and being to the manner
born as ye may say, I never tire of it; but for a genteel supper like what you
have always given to the lads—”
“Faith,” said the laird, “they’ll find it most genteel where they’re going.
The Englishmen will think it the finest table in the world when they hear we
have grouse every day; and Robbie’s no bound to condescend upon the
number of other dishes. I know what I am doing.”
“No doubt, no doubt: I was only making a remark. Now I think a bit of
cod from the sea or a made dish of fine collops, or just a something tossed
up with a bit of veal, they’re more genteel—and I know that’s what you’re
always thinking of, Neil—of course, for the boys’ sakes——”
“There’s a made dish coming, mem,” said Merran, who was waiting.
“Oh, there’s a made dish coming! I thought Marg’ret would mind what
was for the credit of the house. Robbie, my man, ye ought to feel yourself a
great personage with all the phrase that’s made for you. When Sandy went
away, who was the first, there was nothing but a haggis—but we’ve learned
many things since then.”
“A haggis is a very good thing, it’s fit for a king’s table.”
“But not what you would call refined, nor genteel. Give me the leg and a
piece of the back—there’s more taste in it. I hope you will always be
grateful to your father for giving ye such a grand set out.”
 “I think,” said the minister at the other end, “that you and Drumcarro,
mem, give yourselves more and more trouble every son that leaves ye. This
is the fifth I have seen.”
“Oh, don’t say me, Mr. Pyper,” said the mother. “I know just nothing
about it—when your son’s going away, and ye think ye may never set eyes
on him again, who’s to think of eating and drinking? He may do it, but not
me.”
“That’s very true,” said Mr. Pyper. “Still, to give the lad a something
pleasurable to look back upon, a last feast, so to speak, has many points in
its favour. A lad’s mind is full of materialism, as you may call it, and he will
mind all the faces round the friendly board.”
“It’s not very friendly to me,” said the mother, with a sob, “my four
bonny boys all away, and now Robbie. It just breaks my heart.”
“But what would you do with them, mem, if they were here?” said the
sensible minister; “four big men, for they’re all men by this time, about the
house? No, no, my dear leddy, you must not complain. Such fine openings
for them all! and every one getting on.”
“But what does that matter to me, Mr. Pyper, if I am never to see one of
them again?”
“Oh, yes, mem, it matters—oh, ay, it matters much. The young of no
species, much less the human, can bide at home. Fathers and mothers in the
lower creation just throw them off, and there’s an end. But you do more
than that. You put them in the best way of doing for themselves, and the
King himself cannot do better. Alas!” said the minister, “no half so well,
decent man—for look at all these young princes, one wilder than the other.
And every one of yours doing so well.”
“Oh, yes, they’re doing well enough—but such a long way away. And
me so delicate. And Robbie never quite strong since he had the measles. It’s
borne in upon me that I will never see him again.”
“You need not say it, mother,” said Kirsteen, “for that’s what nobody can
know; and it’s just as likely he may be sent home with despatches, or some
great grandee take a fancy to him and bring him back. And when we’re
sitting some day working our stockings he’ll come linking in by the parlour
door.”
“Oh, you’re just as light as air,” said the mother; “there’s nothing serious
in ye. You think going to India is just like going to the fair.”
 Kirsteen darted a quick glance at her mother, but said no more. Her eyes
kept filling much against her will. She was in great terror lest a big drop
might brim over and run down her cheek, to be spied at once by Jeanie or
the boys. For nothing would be hid from these little things: they could note
at the same moment the last bit of a bird which they had all counted on,
being transferred to Aunt Eelen’s plate, and keep an eye upon the favourite
apple each had chosen, and spy that suspicious brightness in Kirsteen’s
eyes. Nothing could be hid from their sharp, little, all-inspecting looks.
There was a breathless moment when the cloth was drawn, and the black
gleam of the mahogany underneath changed in a moment the lights of the
picture, and gave the children a delightful opportunity of surveying
themselves in that shining surface. It was a moment full of solemnity.
Everybody knew what was coming. The port and sherry, with their little
labels, in the silver holders intended to prevent the bottles from scratching
the table, were placed before Mr. Douglas. Then there was also placed
before him a trayful of tall glasses. He rose up: the eyes of all followed his
movements: Jock and Jamie projecting their red heads forward in the smoky
glow of the candles, then much in want of snuffing: Jeanie’s paler locks
turned the same way. Mary, who had her mother’s brown smooth hair,
rested her clasped hands upon the edge of the table with calm expectation.
Kirsteen leant her elbows on the same shining edge, and put down her face
in her hands. Miss Eelen shook her head, and kept on shaking it like a china
mandarin. The laird of Drumcarro went to an old-fashioned wine-cooler,
which stood under the sideboard. He took from it one bottle of champagne,
which occupied it in solitary dignity. Marg’ret stood ready with a knife in
her hand to cut the wire, and a napkin over her arm to wipe up anything that
might be spilt. Not a word was said at table while these preliminaries were
gone through. Aunt Eelen, as the catastrophe lingered, went so far as to
make a suppressed Tchish! Tchish! of her tongue against her palate. The rest
were full of serious excitement too important for speech. The bottle was
opened finally without spilling a drop: it was perhaps not so much “up” as it
might have been. Drumcarro filled all the glasses, one for each person at
table, and another for Marg’ret. There was perhaps more foam than wine in
a number of the glasses. He held up his own in his hand. “It’s Robbie’s last
night at Drumcarro,” he said, “for the present. Have you all your glasses?
Before the fizz is out of the wine drink to Robbie’s good health, and good
luck to him, and to all our lads that have gone before.” He touched the foam
in his glass, now fast dying away, with his lips. “May they all come back
with stars on their breasts,” he said, “and do credit to their name—and not a
laggard, nor a coward, nor one unworthy to be a Douglas among them all!”
The other male members of the party were standing up also, “Here’s to
you, Robbie! Here’s to you, Robbie!” cried the two boys. The foam in their
glasses merely moistened their throats; the minister, however, whose glass
had been full, gravely swallowed its contents in little sips, with pauses
between. “A very good health to them all, and the Lord bless them,” he said
with imposing authority. Mrs. Douglas, taking advantage of the privilege
awarded to her, began to cry, and Marg’ret lifted up a strong voice, from the
foot of the table where she stood with her hand upon the shoulder of the
hero.
“Be a good lad, Robbie—and mind upon your Minnie and a’ the family
—and be a credit to us a’: here’s to you, and to the rest o’ the young
gentlemen, them that’s gone, and them that are to go!”
“Ye’ll have to get a new bottle for the little one,” said Aunt Eelen, “Neil,
my man, for your half-dozen will be out with Jock.” She gave a harsh laugh
at her own joke. “And then there’s the lasses’ marriages to be thought
upon,” she added, setting down her glass.
Drumcarro resumed his seat, the ceremonial being over. “Let the lasses’
marriages alone,” he said impatiently. “I’ve enough to think upon with my
lads. Now, Rob, are you sure you’re all ready? Your things packed and all
your odds and ends put up? The less of them you take the better. Long
before you’ve got the length of Calcutta ye’ll be wishing you had left the
half of your portmanteaux at home.”
“I’ve just two, father.”
“Well, ye’ll be wishing ye had but one. Bring ben the hot water,
Marg’ret; for wine’s but a feeble drink, and cold on the stomach. My wife
never moves at the right time—will I give her a hint that you’re waiting,
Eelen?”
“Not on my account, Drumcarro. Your champagne’s no doubt a grand
drink; but a glass out of your tumbler, if you’re going to make one, is more
wholesome and will set all right.”
“I thought ye would say that,” said the laird. She had said it already on
every such occasion—so that perhaps his divination was not wonderful. He
proceeded with care to the manufacture of “the tumbler,” at which the
minister looked from the other end of the table with patient interest, abiding
his time.
“Snuff the candles,” said the laird, “will nobody pay a little attention?
You three little ones, you can run away with your apples, it’s near your bed-
time; but don’t make more noise than you can help. Marg’ret, take the hot
water to the minister. Champagne, as ye were saying, Eelen, is a grand
drink; I think it right my sons should drink it at their father’s table before
they plunge into the extravagance of a mess. It teaches a lad what he’s
likely to meet with, and I would not have one of mine surprised with any
dainty, as if he had come out of a poor house. But a wholesome glass like
what I’m helping you to is worth twenty of it.” He was filling a wine-glass
with his small silver toddy-ladle as he spoke, and the fumes of the pungent
liquid rose in curls of steam pleasant to the accustomed nostrils. Robbie
kept an eye upon the hot water which Mr. Pyper detained, knowing that one
of the privileges of his position to-night was “to make a tumbler” for
himself, with the privilege of offering it then to his sisters, as each of his
brothers had done.
“Can I assist you to a glass, mem? just a drop. It will do ye good,” the
minister said.
“Nothing will do me good,” said Mrs. Douglas. “I’m far past that; but
I’ll take a little for civility, not to refuse a friend; whether it’s toddy or
whether it’s wine it’s all sounding brass and a tinkling cymbal to me. A
woman when her bairns go from her is little comforted by the like of that.”
“And yet the creature comforts have their place, a homely one but still a
true one,” said the minister. “There’s a time to feast as well as a time to
refrain from feasting. Miss Mary, may I have the pleasure of assisting you?”
“I’ll take a little from Robbie,” said the elder daughter, wisely instructed
that it was well thus to diminish the unwonted tumbler allowed to the
novice. Kirsteen rose quickly to her feet as these interchanges went round.
“Mother, I think if ye’ll let me, I’ll just give an eye to what the little ones
are doing,” she said, “and see that Robbie’s things are all ready. One of the
boxes is open still and there are these handkerchiefs.”
Kirsteen’s eyes were brimming over, and as she spoke a large drop fell
upon her hand: she looked at it with alarm, saying, “I did not mean to be so
silly,” and hastened away.
 “Where is Kirsteen away to? Can she not take her share of what is going
like the rest?” said her father. “You breed these lasses to your own
whimsies, Mistress Douglas. The bairns are well out of the road; but them
that are grown up should bide where they are, and not disturb the family. I
have no patience with them.”
“I’m here, father,” said Mary in her mild voice.
“Oh, ay, you’re there,” said the inconsistent head of the house, “for
you’re just nobody, and never had two ideas in your head,” he continued in
a lower tone. “Now, Robbie, my man, take your glass, there is no saying
when you will get another. It’s just second nature to a Scotsman, but it’s as
well for you to be out of the way of it; for though it’s the most wholesome
drink, it’s very seductive and you’re much better without it at your age. It’s
like the strange woman that you’re warned against in Scripture.”
“Drumcarro!” said Aunt Eelen. “Oh fie! before ladies.”
“Ladies or no ladies I cannot let the occasion pass without a word of
warning,” said the father. “Ye will have every temptation put before ye, my
lad; not drink perhaps, for the climate will not stand it, but other things, that
are worse.”
“I’m thinking, Christina,” said the old lady, “that now your goodman has
begun his moralities it may be as well for us to go, for you know where that
begins and you never can tell where it may end; a man has cognizance of
many things that cannot enter into the experience of you and me. Mind you
what your father says, Robbie, but it’s not intended for your mother and
me.”

CHAPTER III.
Kirsteen hurried out of the room, out of the fumes of the toddy and the
atmosphere of the half-festive, half-doleful occasion which made a not
altogether unpleasant excitement in the monotony of the home life. She
gazed in at the open door of the parlour, and saw the three younger children
gathered in the firelight upon the hearthrug munching their apples, and the
sweets with which they had been allowed to fill their pockets. The firelight
made still more ruddy the red heads and freckled faces of the boys, and lit
up Jeanie, who sat on a footstool a little higher than her brothers, in her
more delicate tints. Kirsteen was much attached to her younger sister, who
promised to be the beauty of the family, and thought her like an angel,
especially as seen through the dew of her wet eyes. “Dinna make a noise,”
she said; “be awfu’ quiet or you’ll be sent to your beds;” and then closed the
door softly and stole through the dark passage towards the principal
entrance. There was no light save a ruddy gleam from the kitchen in the
depths of that dark passage which traversed the whole breadth of the house,
and that which shone through the crevices of the dining-room door. She had
to find her way groping, but she was very well used to this exercise, and
knew exactly where the hall-table and the heavy wooden chairs on each
side stood. The outer door stood half open according to the habit of the
country where there were no burglars to fear, and little to tempt them, and a
perfect capacity of self-defence inside. There was a full moon that night, but
it had not yet risen, though the sky was full of a misty light which preceded
that event. A faint shadow of the group of trees outside was thrown upon the
doorway; they were birches slender and graceful, with their leaves half
blown away by the October gales; those that remained were yellow with the
first touches of the frost, and in themselves gave forth a certain light.
Kirsteen stole out to a bench that stood against the wall, and sat down in a
corner. She was not afraid of cold with her uncovered head and bare arms.
All the moods of the elements were familiar to the Highland girl. She
thought it mild, almost warm: there was no wind, the yellow birches
perceptible in their faint colour stood up like a group of long-limbed youths
dangling their long locks in the dim light: the further landscape was but
faintly visible, the shoulder of the hill against the sky, and a single gleam of
the burn deep down among the trees.
She sat pressing herself into the corner of the seat, and the long pent-up
tears poured forth. They had been getting too much for her, like a stream
shut in by artificial barriers, and now came with a flood, like the same
stream in spate and carrying every obstruction away. It was almost a
pleasure to see (if there had been any one to do so) the good heart with
which Kirsteen wept: she made no noise, but the tears poured forth in a
great shower, relieving her head and her heart. They were very heavy, but
they were not bitter. They meant a great deal of emotion and stirring up of
her whole being, but though her feelings were very poignant they were not
without pleasure. She had never felt so elevated above herself, above every
dull circumstance that surrounded her. She had been very sorry and had
shed tears plentifully when the other boys went away. But this was not the
same. She perhaps did not confess to herself, yet she knew very well that it
was not altogether for Robbie. Robbie had his share, but there was another
now. For years Kirsteen and Ronald Drummond had been good friends.
When he went away before she had felt a secret pang, and had been very
eager to hear the news of the battles and that he was safe: but something
had changed this friendship during the last summer while he had been at
home. Not a word had been said: there was no love-making; they were both
too shy to enter upon any revelation of feeling, nor was there any
opportunity for explanations, since they were always surrounded by
companions, always in the midst of a wandering, easy-minded party which
had no respect for any one’s privacy. But Kirsteen when she marked her
brother’s handkerchiefs with her hair had fully intended that Ronald should
see it, and be struck with the similarity of the initials and ask for or take one
of them at least. Her heart beat high when this happened according to her
prevision; and when he stooped and whispered, “Will ye wait for me,
Kirsteen, till I come back?” the answering whisper, “That I will!” had come
from the bottom of her heart. She had scarcely been aware of what was said
in the hurry of the moment. But it had come back to her, every syllable and
every tone as soon as it was all over. Their spirits had floated together in
that one moment, which was only a moment, yet enough to decide the
course of two lives. They were too much bound by the laws of their
youthful existence to think of breaking any observance in order to expand
these utterances, or make assurance sure. That Ronald should spend his last
evening at home with his mother and sister, that Kirsteen should be present
at Robbie’s parting supper, was as the laws of the Medes and the Persians to
these two. No emergency could be imagined of sufficient weight to interfere
with such necessities of life. And there was something in their simple
absolutism of youthful feeling which was better expressed in the
momentary conjunction, in the sudden words so brief and pregnant, than in
hours of lovers’ talk, of which both boy and girl would have thought shame.
“Will ye wait for me till I come back?” What more could have been said in
volumes? and “That I will!” out of the fervour of a simple heart? Kirsteen
thought it all over again and again. He seemed to stand by her side bending
a little over her with a look half smile, half tears in his eyes; and she was
aware again of the flash of the sweet discovery, the gold thread of the little
letters put to his lips, and then the question, “Will ye wait?” Wait! for a
hundred years, for all the unfathomed depths of life, through long absence
and silence, each invisible to the other. “That I will!” She said it over and
over again to herself.
In those days there was no thought of the constant communications we
have now, no weekly mails, no rapid courses overland, no telegraph for an
emergency. When a young man went away he went for good—away; every
trace of him obliterated as if he had not been. It was a four months’ voyage
to India round by the Cape. Within the course of the year his mother might
hope to hear that he had arrived. And if an Indian letter had come even at
that long interval for a girl in another family, what a host of questions
would she not have had to go through! “A letter for Kirsteen! Who’s writing
to Kirsteen? What is he writing to her about? What is the meaning of it all?
I must know what that means!” such would have been the inquiries that
would have surged up in a moment, making poor Kirsteen the object of
everybody’s curious gaze and of every kind of investigation. She never
dreamed of any such possibility. Robbie, when he wrote home, which he
would no doubt do in time, might mention the companion of his voyage;
Agnes Drummond might say “There’s a letter from our Ronald.” These
were the only communications that Kirsteen could hope for. She was very
well aware of the fact, and raised no thought of rebellion against it. When
she gave that promise she meant waiting for interminable years—waiting
without a glimpse or a word. Nor did this depress her spirits: rather it gave a
more elevating ideal form to the visionary bond. All romance was in it, all
the poetry of life. He would be as if he were dead to her for years and years.
Silence would fall between them like the grave. And yet all the time she
would be waiting for him and he would be coming to her.
And though Kirsteen cried, it was not altogether for trouble. It was for
extreme and highly-wrought feeling, sorrow and happiness combined.
Through all her twenty years of life there had been nothing to equal that
moment, the intensity of it, the expectation, the swift and sudden realisation
of all vague anticipations and wishes. It was only a minute of time, a mere
speck upon the great monotonous level of existence, and yet there would be
food enough in it for the thoughts of all future years. When the thunder-
shower of tears was exhausted, she sat quite still in a kind of exalted
contentment, going over it and over it, never tired. The hot room and the
smoky glare of the candles, and the fumes of the whisky and the sound of
all the voices, had been intolerable to her; but in the fresh coldness of the
night air, in that great quiet of Nature, with the rustle of the leaves going
through it like breath, and the soft distant tinkle of the burns, what room
and scope there was for remembering; which was what Kirsteen called
thinking—remembering every tone and look, the way in which he
approached the table where her work was lying, her wonder if he would
notice, the flush of perception on his face as he said, “It’s my name too,”
and then that tender theft, the act that left Robbie for ever without one of his
pocket-handkerchiefs,—she thought with a gleam of fun how he would
count them and count them, and wonder how he had lost it—the little
visionary letters put to his lips. Oh that her heart had been sewn in with the
hair to give to him! But so it was, so it was! He had that pledge of hers, but
she had nothing of his, nor did she want anything to remind her, to bind her
faith to him, though it should be years before she saw him again. The tears
started into her eyes again with that thought, which gave her a pang, yet one
which was full of sweetness: for what did it matter how long he was away,
or how dark and still the time and space that separated them now. “Will ye
wait for me till I come back?” that would be the gold thread that should run
through all the years.
The sound of a little movement in the dining-room from which all this
time she had heard the murmur of the voices, the tinkle of the glasses, made
her pause and start. It was the ladies withdrawing to the parlour. She
thought with a little gasp that they would find the children scorching their
cheeks on the hearthrug, instead of being sent off to bed as should have
been done, and held her breath expecting every moment the call of
“Kirsteen!” which was her mother’s appeal against fate. But either the
general license of the great family event, or the sedative effect of her
mouthful of champagne and glass of toddy, or the effect of Aunt Eelen’s
conversation which put her always on her defence whatever was the
subject, had subdued Mrs. Douglas: there came no call, and Kirsteen,
though with a slightly divided attention, and one ear anxiously intent upon
what was going on indoors, pursued her thoughts. It gave them a more vivid
sweetness that they were so entirely her own, a secret which she might
carry safely without any one suspecting its existence under cover of
everything that was habitual and visible. It would be her life, whatever was
going on outside. When she was dull—and life was often dull at Drumcarro
—when her mother was more exacting than usual, her father more rough,
Mary and the children more exasperating, she would retire into herself and
hear the whisper in her heart, “Will ye wait till I come back?”—it would be
like a spell she said to herself—just like a spell; the clouds would disperse
and the sun break out, and her heart would float forth upon that golden
stream.
The sound of a heavy yet soft step aroused Kirsteen at this moment from
her dreams; but she was set at ease by the sight of a great whiteness which
she at once identified as Marg’ret’s apron coming slowly round the corner
of the house. “I just thought I would find you here,” said Marg’ret. “It’s
natural in me after that warm kitchen and a’ the pots and pans, to want a
breath of air—but what are you doing here with your bare neck, and
nothing on your head? I’m just warning you for ever, you’ll yet your death
of cold.”
“I could not bear it any longer,” said Kirsteen, “the talking and all the
faces and the smell of the toddy.”
“Hoot,” said Marg’ret, “what ails ye at the smell of the toddy? In
moderation it’s no an ill thing—and as for the faces, you wouldna have folk
without faces, you daft bairn; that’s just a silly speech from the like of you.”
“There’s no law against being silly,” Kirsteen said.
“Oh, but that’s true. If there was, the jails would be ower full: though no
from you, my bonnie dear. But I ken weel what it is,” said Marg’ret, putting
her arm round the girl’s shoulder. “Your bit heart’s a’ stirred up, and ye
dinna ken how ye feel. Tak’ comfort, my dear bairn, they’ll come back.”
Kirsteen shed a few more ready tears upon Marg’ret’s shoulder, then she
gave that vigorous arm a push, and burst from its hold with a laugh,
“There’s one of Robbie’s handkerchiefs lost or stolen,” she said. “Where do
ye think he’ll ever find it? and R. D. worked upon it with a thread of my
hair.”
“Bless me!” said Marg’ret with alarm, “who would meddle with the
laddie’s linen? but you’re meaning something mair than meets the eye,” she
added, with a pat upon the girl’s shoulder; “I’ll maybe faddom it by and by.
Gang away ben, the ladies will be wondering where ye are, and it’s eerie
out here in the white moonlight.”
“Not eerie at all: ye mean soft and sweet,” said Kirsteen, “the kind of
light for thinking in; and the moon is this minute up. She’s come for you
and not for me.”
“I cannot faddom you the nicht any more than I can faddom what ye
say,” said Marg’ret. “There’s mair in it than Robbie and his handkerchief.
But I maun go in and fasten up the straps and put his keys in his pocket, or
he’ll forget them. Laddies are a great handful, they’re aye forgetting. But
they’re like the man’s wife, they’re ill to have, but worse to want. Gang in,
gang in out of the night air,” said Marg’ret with a faint sob, softly pushing
Kirsteen before her. The smell of the peat fires, which was pleasant, and of
the smoke of the candles, which was not, and of the penetrating fumes of
the toddy again filled Kirsteen’s nostrils as she came in. She had no right to
be fastidious, for she had been brought up in the habit and knowledge of all
these odours. When she entered, another scent, that of the tea with which
the ladies were concluding the evening, added its more subtle perfume. In
those days people were not afraid of strong tea, mixed with a great deal of
green to modify the strong black Congou, and it had been “masking” for
half an hour before the fire: they were not afraid of being “put off their
sleep.”
“And do ye mean to say, Christina, that there’s nobody coming about the
house that would do for your girls?”
“Oh, for mercy’s sake, Eelen, say not a word about that: we’ve had
trouble enough on that subject,” said Mrs. Douglas in her injured voice.
“Are you meaning Anne? Well, I mind Drumcarro’s vow, but there is no
doubt that was a missalliance. I’m meaning men in their own position of
life.”
“Where are they to see men in their own position, or any men?” said the
mother shaking her head. “Bless me, Kirsteen, is that you? I don’t like
people to go gliding about the house like that, so that ye never can hear
them. When your aunt and me were maybe talking—what was not meant
for the like of you.”
“Hoot, there was no hairm in it,” said Aunt Eelen, “if all the lasses in the
town had been here.”
“But it’s an ill custom,” said Mrs. Douglas. “However, as you’re here ye
may just get me my stocking, Kirsteen, and take up a stitch or two that I let
fall. Na, na, no strangers ever come here. And now that my Robbie’s going,
there will be fewer than ever. I wish that your father would not keep that
laddie out of his bed, and him starting so early. And, eh, me, to think that
I’m his mother, and most likely will never see him in this world again!”

CHAPTER IV.
Robbie went away next morning very early, before the October day was
fairly afloat in the skies. They had no carriage at Drumcarro except “the
gig,” and it was perched up in this high conveyance, looking very red with
tears and blue with cold, that the household, all standing round the door,
saw the last of the boy mounted beside his father, with a large portmanteau
standing uncomfortably between them. His other baggage had been sent off
in the cart in the middle of the night, Jock as a great favour accompanying
the carter, to the great envy and wrath of Jamie, who thought it hard that he
should miss such a “ploy,” and could see no reason why his brother should
be preferred because he was two years older. Jamie stood at the horse’s
head looking as like a groom as he could make himself, while his father
made believe to hold in the steady honest mare who knew the way as well
as he did, and was as little troubled by any superfluous fun or friskiness.
Mrs. Douglas had remained in bed dissolved in tears, and her boy had taken
his leave of her in those congenial circumstances. “Be a good lad, Robbie,
and sometimes think upon your poor mother, that will never live to see you
again.” “Oh, mother, but I’ll be back long before that,” he cried vaguely,
doing his best to behave like a man, but breaking out in a great burst of a
sob, as she fell back weeping upon her pillows. The girls at the door were in
different developments of sorrow, Mary using her handkerchief with
demonstration, Kirsteen with her eyes lucid and large with unshed tears,
through which everything took an enlarged, uncertain outline, and little
Jeanie by turns crying and laughing as her attention was distracted from
Robbie going away to Jamie standing with his little legs wide apart at the
mare’s respectable head. Robbie was not at all sorry to go away, his heart
was throbbing with excitement and anticipation of all the novelties before
him; but he was only eighteen, and it was also full for the moment of softer
emotions. Marg’ret stood behind the girls, taller than any of them, with her
apron to her eyes. She was the last person upon whom his look rested as his
father called out, “Stand away from her head,” as if honest Mally had been
a hunter, and with a friendly touch of the whip stirred the mare into motion.
Robbie looked back at the gray house, the yellow birches waving in the
winds, the hillside beyond, and the group round the door, and waved his
hand and could not speak. But he was not sorry to go away. It was the aim
of all his breeding, the end looked forward to for many years. “It’s me the
next,” said Jock, who was waiting at Inveralton, from which place by
fishing-smack and coach Robbie was to pursue his way to Glasgow and the
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com