Contributions to the multidisciplinarity of computer science and IS

Contributions to the multidisciplinarity of
computer science and IS
« Contributions à la multidisciplinarité de
l’informatique et des SI »
Habilitation à Diriger les Recherches
20 janvier 2017
Saïd Assar
Institut Mines Telecom, Ecole de Management

Agenda
 Preliminary thoughts (x4)
 Research works and research path
I. Epistemological analysis : a tentative
II. About methods
III. Research works 1: Model enactment
IV. Research works 2: Defect resolution time prediction
V. Research works 3: Learning style impact in e-learning
context
VI. Conclusion
2
1st period
2nd period
3rd period

Preliminary thoughts (1st) 3
A seminar in the 90’ about “action research” …
Thought 1: Computing can be considered as a
phenomena, as a social subject of study

Preliminary thoughts (2nd)
Programming would be an “artistic” activity?
4
… with no “scientific” basis?
Hoare, C. A. R. (1984). Programming: Sorcery or Science?
IEEE Software, 1(2), 5-16.

Preliminary thoughts (2nd) 5
“art” ≠ “artistic”
Skills ...
 Practice oriented skills (know how – way of doing)
 Methodological skills
Thought 2: Writing computer programs is a
design issue (vs. programs as formal subjects)

Preliminary thoughts (3rd)
Doing research about e-mail ?
6
… a 30 years old artifact …
• Usage problems
• Strong impact
Thought 3: Relevance is
not necessarily related to
the novelty of the artifact
New requirements …

Preliminary thoughts (4th) 7
Exclusivity of research methods in management,
social and organizational sciences?
(Int. Symposium on Empirical SE, 2004)
(Int. Symposium on Empirical SE and
Measurement, 2009)
(Information and Software Tech., 2014)
(Int. Conference on SE – ICSE, 2004)
… not really
(Empirical Softw. Eng., 2011)
B. Kitchenham
Thought 4: Differences among disciplines is
not always related to research methods

Preliminary thoughts … 8
Conclusion …
 Necessity of a holistic, epistemology based, conciliating analysis
… to clarify my research path and position my contributions
o Quid the disciplinary differences in terms of scientific contributions?
o Quid the differences in terms of validity and veracity?
o Quid the different meanings and understandings of the idiom “method”?
o Quid of
 IS Engineering
 Software Eng.
 Management IS
 Empirical SE

Research works and research path 9
A framework for IS research
Source
Lyytinen, K. (1987). Different perspectives on information systems: problems and solutions. ACM
Computing Surveys, 19(1), 5–46.
Topic 2
Topic 1

Topic 2
(IS usage and
impact)
ProAdmin eGov
project
2005-06
Integrating Learning
Styles in eLearning
2005-09
eGov evolution
analysis
2007-14
LS impact on
Mobile Learning
2012-14
A global view of my research activity
Sabbatical at
LTH &
empirical SE
2012 - 15
Topic 1
(IS design and
development)
Model transf. &
code generation
1995
Map engine
design
2002-05
Traceability
meta-modeling
2007-09
Intentional
service desc.
& discovery
2009-11
Generic approach
to model
enactment
2008-14
Topic 1 Topic 2

Topic 2
(IS usage and
impact)
ProAdmin eGov
project
2005-06
Integrating Learning
Styles in eLearning
2005-09
eGov evolution
analysis
2007-14
LS impact on
Mobile Learning
2012-14
Sabbatical at
LTH &
empirical SE
2012 - 15
Topic 1
(IS design and
development)
Model transf. &
code generation
1995
Map engine
design
2002-05
Traceability
meta-modeling
2007-09
Intentional
service desc.
& discovery
2009-11
Generic approach
to model
enactment
2008-14
Works presented

Agenda
II. About methods
context
VI. Conclusion
12
2nd period

I. Epistemological analysis : a tentative 13
Multidisciplinary questioning
 The subject of study [ontology]
 The investigation method [methodology]
Epistemology …
Mario Bunge

To reason on artifacts capable of
concrete and abstract calculations
To understand and explain
phenomena related to the
production and usage of artifacts
To build useful artifacts
that fit with usage
context
“L’informatique”
Science of
the artificial
Formal
science
Factual
science
14
Ontological perspective

II. About methods 15
Method(s) in formal sciences
Mathematical notations and reasoning
“Absolute validity” … as long as
 rules of logic and deduction are respected
 assumptions and hypotheses are holding
The “mathematisation” of a problem …
 a powerful approach to research
 positively perceived in general

Method(s) in factual sciences
Source
Bunge, M. (1967). Scientific Research 1: The Search for System (Vol. I). New York: Springer, (page 9)
Validity ?
Contribution ?

Method(s) in design – the roots
Brunelleschi (1377-1446)
“… recognized to be the first modern
engineer …” [Wikipedia]
Designeo (It) Dessein (Fr)
Dessin (Fr)Design (En)
To design is “to draw intentions”
 Modeling is the language of design
 Design seeks the production artifacts …
 What about the production of knowledge ?

Product (i.e. artifact)
Process
Methods in design – product vs. process
Design is about both the product and the process …
 Knowledge about the product or the process ?
 Idiographic (specific) vs. Nomothetic (generic) knowledge ?
 Process guidance ?

Methods in design – link with science
Multiple terminology …
 Design studies
 Design discipline
 Design research
 Scientific design
 Design science …
Design vs. science
 Design-based learning and discovery
[Source : Nigel Cross, “Designerly way of knowing”, 1982 / 2001]

Methods in design – “Design science” in IS and MIS
Source
Wieringa, R. (2009). Design science as nested problem solving. In Proceedings of the 4th Int. Conf. on Design
Science Research in IS and Technology (DESRIST’09), New York, USA: ACM.
 Epistemological necessity of distinguishing “practical” and “knowledge” problems

My recent publications related to methodological
issues (2012 – 2016)
Topics tackled
‒ Method Engineering [a synthesis]
‒ MDE and Requirements Engineering [a review of ModRE workshops, 2011-13]
‒ Creative approaches in RE [a state of the art ]
‒ Replication in experimental research [experimental study]
‒ Methods for literature review [a synthesis]
‒ Empirical and experimental methods in IS eng. [a synthesis]
‒ Theory development in ERP research [exploratory study]

Agenda
II. About methods
context
VI. Conclusion
22
3rd period

III. Research works 1: Model enactment 23
Modeling in IS design
Source
Wand, Y., & Weber, R. (1993). On the ontological
expressiveness of information systems analysis
and design grammars. Information Systems
Journal, 3(4), 217–237.
Source
Rolland, C., & Prakash, N. (2000). From conceptual modelling to requirements
engineering. Annals of Software Engineering, 10(1-4), 151-176.
 Conceptual gap: requirements vs. systems ?
 Mapping ?
 Transformation ?
 (Semi-) automated ?
 Data vs. process models ?
 Business vs. engineering processes?

What is model enactment?
Direct execution / interpretation of process models by a software
agent:
o Prototyping
o Model validation
o Higher level of abstraction
Software agent Model(s)

RUBIS system (1986-1990)
RUBIS architecture
o Event processor
o Language interpreter
o Forms generation
Limitations
o Implicit semantics
o Product dependant
o Model specific (i.e. Remora)

The MAP notation (1998 – today)
Intentional specification of a process (business or engineering process)
 Process realization through multiples intentions
 Intention realization through multiple strategies
Informal, complex semantics
 Selection of a “candidate section”
 Execution of a strategy
 Product and process trace

A specific MAP engine
Enactment through a dedicated architecture
 Repository structure derived from the meta-model
 Strategies semi-formally specified
 Specification transformed into code
Limitations
o Tentative for making semantics explicit
o Product dependant
o Model specific (i.e. MAP) M-H. Edme. « Proposition pour la modélisation intentionnelle et le guidage
de l’usage des systèmes d’information ». PhD, Université Paris 1, 2005.

Implicit vs. explicit semantics of modeling languages
Source
Mayerhofer, T., Langer, P., Wimmer, M., & Kappel, G. (2013). xMOF: Executable DSMLs Based on fUML. In M.
Erwig et al. (Eds.), Software Language Engineering (SLE’13), p. 56-75, Springer.
 Implicit semantics generate redundancy and incoherence
 Explicit semantics would enable the application of “MDE
techniques for processing language [meta-] specifications”

A generic approach – an exploratory study
What about using meta-modeling tools?
 Based on meta-modeling
 Automatic generation of a CASE tool
 State of the art :
o Structure meta-modeling
o Limited / absent behavior meta-modeling
o Difficulty in handling multiple levels of
instantiations

A generic approach – an exploratory study
Results
 Feasible for the “structure” part of a meta-model (i.e. MAP editor)
 Semantics specified in the code generation scripts (i.e. not explicit)
=> Limitation in the meta-modeling languages

A generic approach – a proposal
Transformation
rules
Code
generation
CIM level PIM level
(XML specifications)
PSM level
Abstract syntax
Semantics
Engine
architecture
Structural view
Static meta-model
(UML)
Behavioral view
Dynamic meta-model
(Remora)
Process
enactment
engine
S. Mallouli. « Méta-modélisation du Comportement d’un Modèle de Processus: Une Démarche
de Construction d’un Moteur d’Exécution ». PhD thesis, Univ. Panthéon-Sorbonne-Paris I, 2014.

A generic approach – behavior meta-modeling
stopMapEnact()
selectCandidateSection()
stateMapInst.Old=Selected
stateMapInst.New=Running
M1:
Start
Map
Execution
notifyEndSection()
SectionInstance
Product Instance
SendCandidateSections()
EV4
EV5
EV9
MapActor
M2:Liste of candidates
M4:End of Section Execution
compute
Candidate
Sections()
IntentionInstance
Section
Application
Executor
updateIntention()
updateExecSection()
notifyEndExec()
invokeExec()
EV1
M3: Choice
C1: target intention= Stop
C1
MapInstance
startMapEnact()
stateMapInst.Old=Running
stateMapInst.New=Enacted
notifyEndMapEnact()
M6:Stop
stateSecInst.Old=Candidate
statSecInst.New=Selected
M8:Execution
status
computeCandidateSections()
stateSecInst.Old=Created
stateSecInst.New=Candidate
stateSecInst.Old=Selected
stateSecInst.New=Executed
EV3
EV6
EV8
EV12
Product
Manager
ImplemExe
SectionInstance
invokeExec()
notifyEndExec()
EV5
EV10 stateIntInst.Old=Created
stateIntInst.New=Realised
notifyProduct()
EV11
updateProduct()
M5: Product
M9: NewProduct
Situation
EV7
M7:Execution
ParametersstateImpl.Old=Created
stateImpl.New=Selected
C2
C2: target intention=/= Stop
newProductInstance()
EV2
M1bis:
Start MapInst
newMapInstance()
F1
F1: pour toute sectionInstance retournée par l'algorithme de calcul de candidates

Conclusion
Practical problem
 Actual meta-modeling & CAME tools are limited
 A proposal for the graphical expression of semantics
 Operational semantics = the architecture of an enactment engine
Knowledge problems
 feasibility (conceptual – technical – practical)
 Scalability
MAP for method engineering
 a very powerful conceptual tool
 .. yet difficult to enact in a generic manner (product and process
inseparability)

Agenda
II. About methods
context
VI. Conclusion
34
3rd period

IV. Research works 2: exploiting defect text reports 35
Defects in software
Textual
description

Defect lifecycle and resolution time prediction
How much time will it
take to fix?
=> As much as “similar”
defects …

Prediction using text similarity (Weiss et al., 2007)
Previous defects
Prediction set (K)
New incoming
defect
Similarity level (α)
Defect Resolution Time (DRT) prediction – State of the art
Results
 Good predictions when (α) is high … but limited applicability
 Mostly, ~20% of acceptable predictions
 No optimal value for parameters (α) and (K)

Empirical SE, 18(1), 2013, pp. 117-138.
Defect Resolution Time (DRT) prediction – State of the art
Complete set of
defect reports
cluster1
cluster2
cluster3
cluster4
For all K,J
MRTK and MRTJ are
significantly different
(ANOVA statistic test)
MRTK = Mean Resolution Time for
all defects in cluster K Clustering for DRT
prediction ?

Data preparation
Prediction simulation
Replication
The experimental study
S. Assar, M. Borg, D. Pfahl. “Using Text Clustering to Predict Defect Resolution Time: A Conceptual Replication and an Evaluation of
Prediction Accuracy”. Empirical Software Engineering 22, no 3 (2016): 1-39.

The experimental study : replication
K-means clustering
(100% data set)
Statistical analysis
(ANOVA & post-hoc test)
Replication
Data preparation
Replication conditions
‒ Different data sets : 2 open source + 1 proprietary
‒ Different text mining tool : RapidMiner
‒ (Slightly) different data preparation steps
Result
Fully positive

Prediction simulation
Data preparation
K-means clustering
(x% of the data set)
Analysis of
predictive power
x=10,20, …, 100%
Statistical
analysis
The experimental study : testing the claim …
Parameters of the experiment
‒ Data sets: Eclipse, Android, Company A
‒ Prediction error: 25% and 50%
‒ Size of the test set: 10% to 100%
‒ Number of defects used for testing: 1%, 3% and 5%
‒ Number of clusters: 4, 6, 8, 10

Three data sets | K=4 | Pred(0.25) | Number of test points = 1% | Naïve prediction

Three data sets | K=4 | Pred(0.25) | Number of test points (SSF) = 1% - 3% - 5%

Company A| K = 6 – 8 – 10 | Pred(0.25) | Number of test points (SSF) = 3% - 5%
Final result
Claim not
confirmed

IV. Research works 2: exploiting defect text reports
 Limited reliability of text based approaches to DRT prediction
 Need to challenge the theoretical grounding (i.e. “similarity
assumption(s)”)
 Knowledge production in a “factual science” manner with an
empirical and inductive approach :
– Replication is an essential (yet challenging) issue …
– Validity depends on
o Design of the experiment
o Size and quality of the data
o Sophistication and validity of data analysis procedures
o Underlying theory
45
Conclusion

Agenda
II. About methods
context
VI. Conclusion
46
3rd period

V. Research works 3: learning style impact 47
 Individual differences in learning and
acquiring knowledge
 Over 70 learning style models
 Differences in Ls are easier to accommodate
in IT based teaching and learning
Problematic
 Contradictory results concerning the impact of LS
 How to exploit electronic media potentialities in an IT based
learning context

LS and e-media integration (1) Felder-Silverman
LS modelA-L. Franzoni-Velázquez, “A Proposed Method for Adapting
and Integrating Student Learning Style, Teaching Strategies
and Electronic Media”. PhD thesis, 2009.

 Delphi method with 20 participants (univ. teachers)
 Partial implementation
 Test with ~700 students
 Positive correlation
LS and e-media integration (2)

 Kolb LS model
 Technology Acceptance Models (TAM,
UTAUT)
 LS as a moderating factor
 Distinction between “acceptance” and
“continuance to use”
 Multiple experimentations
Mobile learning usage and adoption (1)
Yaneli Cruz, “Learning Styles Effect on Mobile Learning Acceptance: A Continuance
Intention Approach” PhD thesis, 2014.

Acceptance
 Empirical testing (39 valid responses)
 LS moderation effect is modest

Continuance to use
 Empirical testing (51 valid responses)
 “Effort expectancy” and “Social influence” are the only variables that are
moderated by users’ LS

 Differences in learning certainly exist … are they correctly captured
in learning styles?
 Complex typology for e-media … that is highly cited
 Knowledge production in a “factual science” manner and the
hypothetico-deductive approach
– Validity depends on
o Underlying theory
o Size and quality of the data
o Design of the experiment
o Sophistication and validity of data analysis procedures
Conclusion

Agenda
II. About methods
context
VI. Conclusion
54

VI. Conclusion
 An apparent heterogeneous collection of research works
… yet, IT artifact design and usage are at the center
 Importance of understanding “what we know” and “how we know
it”
– Learning about methods …
– Research methods are an important facet of multidisciplinary research
55
So what ?

VI. Conclusion
 Design and usage of DSL for new technologies (IoT ?)
 Enterprise modeling for Digital Transformation
 Text mining in Software Engineering
 Theory development
– In MIS research
– In Design Science research
 Meta-analysis for evidence aggregation
56
Research perspectives

VI. Concluding remarks 57
Source
Ramsin, R., The Engineering of an Object-Oriented Software Development Methodology, PhD thesis,
University of York, UK, 2006.
About methods : terminological / conceptual
confusion?

VI. Concluding remarks 58
About methods: multidisciplinary misunderstandings?
Matthieu Cisel, Utilisation des MOOC : éléments de typologie,
Thèse de Doctorat, ENS Cachan, 08 juillet 2016
 Description vs. Explanation … ?
 Theoretical contribution … ?

Contributions to the multidisciplinarity of computer science and IS

Recommended

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to Contributions to the multidisciplinarity of computer science and IS (20)

More from Saïd Assar (7)

Recently uploaded (20)

Contributions to the multidisciplinarity of computer science and IS