ALGORITHMS AND WORK IN

ORGANIZATIONS
Agenda

In this session we will…

1. Discuss algorithms (what are they?)
2. Consider how algorithms will change work in organizations
3. Consider what these changes imply for the theories we’ve
covered in the course
4. Discuss responsible algorithmic management in future
organizations
What is an algorithm?

A process or set of rules to be followed in

calculations or other problem-solving
operations, especially by a computer.
What is a learning algorithm?

Learning algorithms to refer to an emergent family of

technologies that build on machine learning, computation,
and statistical techniques, as well as rely on large data sets
to generate responses, classifications, or dynamic
predictions that resemble those of a knowledge worker.

(Faraj et al. 2018)

Why are algorithms transforming our
organizations?

§ Algorithms function by taking in data that has been

structured, formatted, and prepared for processing.
§ Because out of the vast (and constantly increasing)
volume of available data, algorithms can (sometimes)
consider many more factors than humans to make
inference (predictions) and, therefore, perform better on
a range of tasks.

5
Consequential aspects of learning algorithms

1. Black-boxed performance
2. Comprehensive digitization
3. Anticipatory quantification
4. Hidden politics

(Faraj et al. 2018)

Black-boxed performance

§ Algorithmsbehind our personalized search results, ads, or

recommendations cannot be easily explained, redressed, or
adjusted because they are opaque, inaccessible, and/or
unmodifiable.
§ Learning algorithms are especially opaque because they do not
rely on pre-specified instructions, but on evolving weights and
networks of connections that get refined with each additional
data point.
Black-boxed performance
Comprehensive digitization

§ Everything we do becomes digitized and datified.

§ The performance of learning algorithms improves with
the provision of very large and accurate data sets, often
by combining data extracted in one context with data
from other contexts and possibly of different natures.
Anticipatory quantification
§ With most objects being tagged, locations
identified, people's attributes marked,
behavior traced, and interactions mapped, all
aspects of working and living can be digitally
represented and usually quantified.

§ This is based on the assumption that digital

data can stand for social life; that is to say,
the detailed measurement of different
aspects of a person's activities and
utterances get to be constitutive of that
person.
Hidden politics
§ As with other technologies,
algorithms are political in their
design.
§ Algorithms are imbued with the
value choices of their designers,
whether these have been made
implicitly or explicitly. The criteria
included in the algorithmic
processes reflect valuation
schemes, beliefs, and ethical
standards.
§ Politics also play out in the
classification, selection, and pre-
processing of the data that is fed
into a learning algorithm.
Consequences for organizations?
In previous lectures we learned that…
§ An organization is a unit of people that aims to meet individual’s needs and to pursue
collective goals through coordinated action (lecture 1).
§ If people are the key for the organization and in decision making, how will algorithms
affect their opinions, expertise and decisions?

§ Do people get smarter or less smart through interaction with algorithms?

§ Do people more or less polarized through interaction with algorithms?
§ Do the traditional decision-making processes still apply? Or should we find new
governance structures?
In previous lectures we learned that…

§ Organizations can be built as bureaucracies or post-bureaucracies

(lecture 3).

§ Whathappens to division of labor, hierarchy and merit-based pay in an

organization where half (or more) of the workforce is hardware and
software robots?
§ Who is accountable when the decision is made by a learning algorithm?
In previous lectures we learned that…

§ Core technologies influence organizational structure, strategy, etc.

(lecture 4)

§ Are algorithms just another core technology or is it more than that?

§ When does a technology become a co-worker, collaborator, manager?
§ Who is in control when humans and algorithms work together? (lectures 5-
7)
How to do algorithmic management responsibly?

What do human What should the How do human workers

managers need? algorithms do? respond to algorithms taking
the roles of the human
manager?
Human side: What do managers need?
§ Responsible managers need to attain a
level of algorithmic literacy in order to
work with and interpret an algorithm in real-
world decision making scenarios.
§ Managers need to retain decision control:
Algorithmic decision making systems
should accommodate for a human-in-the-
loop if augmentation, rather than
automation, is the intention.

(Burton, Stein and Jensen, 2019)

Human side: What do managers need?

§ Responsible managers need to understand

and engage their intuition, rather than
combat it.
§ Humans and algorithms make use of
different decision processes: Humans are
nonrational (Gigerenzer, 2001), we satisfice
under real-world constraints and
uncertainty. Algorithms are traditionally
rational (Tversky & Kahneman, 1974), they
optimize on the premise of normativity and
calculable risk. One is not better than the
other.

(Burton, Stein and Jensen, 2019)

Human side: What do managers need?

§ Managers need to be able to cultivate

virtue (personal excellence that moves
the individual towards pursuing internal
goods, practical wisdom and to act
voluntarily)
§ Responsible managers are virtuous in
that they continually try to improve and
excel at something, thereby pushing
forward their own practice as well as
the broader activity of management
itself

(Gal, Jensen and Stein, 2020)

Data side: Design for legibility and informing

§ Making data and analytics algorithms

comprehensible to people
§ What data are they consuming?
§ What methods are they using to draw
inferences?
§ Algorithmic systems designed for
informing can provide reasoning and
context for the insights or decisions they
generate like an expert assistant would.
(Gal, Jensen and Stein, 2020)

https://www.interaction-design.org/literature/article/human-data-interaction-hdi-the-new-information-frontier
Data side: Design for ambiguity
§ Individual understanding and attitudes change
over time.
§ Data systems should reflect:
§ current social norms around data
§ different legal and regulatory frameworks that apply
§ the ways in which data is often ambiguous in its
subject and its meaning
§ Many algorithmic systems are designed to be
over-confident. The probabilistic calculations
are usually hidden from users and results are
presented as facts.
(Gal, Jensen and Stein, 2020)
https://www.interaction-design.org/literature/article/human-data-interaction-hdi-the-new-information-frontier
Data side: Design for agency

§ Giving people the capacity to act

within data systems, to opt-in or to
opt-out, to control, inform and
correct data and inferences.
§ Not all users must or will choose to
continually exercise this capacity,
engaging in detailed, ongoing
control and management of their
data.

https://www.interaction-design.org/literature/article/human-data-interaction-hdi-the-new-information-frontier
 Garbage in-garbage out

Workers’ Perceived Information-Based Vulnerability: a user’s perception that

information entered into a system could be used in an opportunistic manner by
the employer. Based on:
Sensitivity of Trust in system
Perceived control Trust in employer
information (algorithm)

Workers who feel vulnerable reduce the quality of their system use
Algorithm games
Algorithmic decision-making is a dynamic process.

Algorithms attempt to estimate some difficult-to-measure quality about a subject using proxies.

The subjects (e.g., employees) in turn change their behavior in order to game the system and get a better treatment for themselves (or, in
some cases, to protest the system).

These behavioral changes can then prompt the algorithm to make corrections.

As the game escalates, you have adverse effects on algorithm accuracy, distributional fairness, and efficiency, and human autonomy.

A better strategy is to minimize gaming by maximizing trust & meaningful data

Bambauer, J., & Zarsky, T. (2018). The algorithm game. Notre Dame L. Rev., 94, 1.
Take-aways
§ Algorithms can reshape organizational structures and enhancing
operational efficiency.
§ Their 'black-boxed' nature present challenges to transparency
and accountability.
§ Pervasive digitization of activities plays a critical role in vast data
generation for refining algorithmic accuracy and relevance.
§ There are embedded politics in algorithm design, reflecting
values and biases of creators, and their broader societal
implications.
§ Responsible management in the era of algorithms highlights need
for a balance between human intuition and algorithmic assistance
§ An alternative way of creating algorithms is based on principles
of transparency, negotiability, and user agency to foster trust and
minimize misuse or manipulation.

25
References
§ Ali H., and Mancha R. (2018). Coming to Grips with dangerous Algorithms. MIT Sloan Management Review.
§ Boucher Ferguson, R. (2013). Are algorithms influencing your business? MIT Sloan Management Review.
§ Bambauer, J., & Zarsky, T. (2018). The algorithm game. Notre Dame L. Rev., 94, 1.
§ Burton, J. W., Stein, M. K., & Jensen, T. B. (2020). A systematic review of algorithm aversion in augmented decision making.
Journal of Behavioral Decision Making, 33(2), 220-239.
§ Faraj, S., Pachidi, S., & Sayegh, K. (2018). Working and organizing in the age of the learning algorithm. Information and
Organization, 28(1), pp. 62-70.
§ Gal, U., Jensen, T. B., & Stein, M. K. (2020). Breaking the vicious cycle of algorithmic management: A virtue ethics approach to
people analytics. Information and Organization, 30(2), 100301.