Moving beyond Principles: Identifying Actionable AI Fairness Practices

Christoph Burtscher; Mateusz Dolata

arxiv: 2604.18502 · v1 · submitted 2026-04-20 · 💻 cs.CY

Moving beyond Principles: Identifying Actionable AI Fairness Practices

Christoph Burtscher , Mateusz Dolata This is my paper

Pith reviewed 2026-05-10 03:15 UTC · model grok-4.3

classification 💻 cs.CY

keywords AI fairnessactionable practicesgovernance matrixAI lifecycleorganizational rolesthematic analysisfairness implementation

0 comments

The pith

A matrix organizes actionable AI fairness practices by obligation and role across the full AI lifecycle.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

AI systems now shape many organizational decisions, turning fairness into a practical governance problem rather than a purely ethical one. Most existing guidance stays at the level of broad principles and gives little help for day-to-day decisions during design, deployment, and monitoring. This paper examines sixty academic, policy, and practitioner sources through discourse and thematic analysis to extract concrete practices. It assembles those practices into a single matrix that covers every stage of the AI lifecycle and sorts them according to how binding each action is and which organizational roles should carry it out. A sympathetic reader would care because the matrix turns abstract commitments into role-specific tasks that organizations can actually assign and track.

Core claim

Through discourse and thematic analyses of sixty academic, policy, and practitioner sources, the authors derive a structured set of AI fairness practices presented in a comprehensive matrix that spans the AI lifecycle and is organized by obligation degree and organizational role. The matrix supplies dynamic, role-specific guidance intended to support both the implementation and the ongoing sustainment of AI fairness.

What carries the argument

The AI fairness practices matrix, which compiles practices from the source analyses and arranges them by lifecycle stage, obligation level, and responsible organizational role to turn principles into assignable tasks.

If this is right

Different roles within an organization receive targeted tasks rather than generic advice.
Guidance extends from initial design through deployment and monitoring to long-term maintenance.
The matrix functions as a modular scaffold that organizations can adopt or adapt for their governance processes.
Information-systems research gains an operational tool that moves fairness discussions from principles to documented practices.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Teams could use the matrix as a checklist to audit existing AI workflows for missing fairness steps.
Future applications might test whether the same structure works equally well in regulated industries versus informal settings.
The matrix implicitly connects technical fairness interventions to the distribution of responsibility inside firms.
Longer-term use could reveal which obligation levels are most often ignored and why.

Load-bearing premise

The practices extracted from the sixty sources are both actionable in real organizations and general enough to apply across sectors and cultural settings without major selection or interpretive bias.

What would settle it

A multi-organization trial that applies the matrix to actual AI projects and finds that teams cannot translate its entries into measurable fairness improvements or encounter consistent adaptation failures would challenge the matrix's claimed utility.

read the original abstract

Because artificial intelligence (AI) increasingly mediates organizational work, fairness has become a critical governance challenge. Existing frameworks often prioritize abstract ethical principles rather than fairness-specific ones and lack actionable guidance across the entire AI lifecycle. This study addresses the principles-to-practice gap in AI fairness governance. We develop actionable AI fairness practices and draw on a socio-technical and praxiological lens, conducting discourse and thematic analyses of 60 academic, policy, and practitioner sources. From these analyses, we derive a structured set of AI fairness practices in a comprehensive, AI lifecycle-spanning matrix organized by obligation degree and organizational role. The matrix provides dynamic, role-specific guidance to support implementation and sustainment of AI fairness. By extending the AI fairness beyond abstract principles to operationalized, actionable practices, we contribute to IS scholarship and offer a modular governance scaffold.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper gives a role-differentiated matrix of AI fairness practices but skips key details on how it was built from the sources.

read the letter

The paper's key contribution is a matrix that maps AI fairness practices to specific organizational roles and levels of obligation throughout the AI lifecycle. It comes from analyzing 60 sources. This synthesis does a good job of making fairness more concrete. Instead of stopping at principles, it breaks down actions for people in different positions, like data scientists versus executives, and indicates how mandatory each practice is. That structure could serve as a reference for teams trying to implement fairness in real projects. The main weakness is the thin description of the analysis process. The paper uses discourse and thematic analysis but does not detail source selection criteria, coding procedures, or validation steps. This leaves room for bias in what ended up in the matrix and raises questions about how well the practices apply beyond the chosen sources. The concern about generalizability is fair here. Readers in AI ethics, information systems, or organizational governance will find the matrix useful as a starting framework. It is not a substitute for context-specific work but offers a modular way to think about responsibilities. The work shows honest engagement with the literature on the principles-to-practice gap. I recommend sending it for peer review, with the expectation that referees will ask for more on the methods to confirm the matrix's robustness.

Referee Report

2 major / 0 minor

Summary. The paper claims that discourse and thematic analyses of 60 academic, policy, and practitioner sources on AI fairness yield a comprehensive, lifecycle-spanning matrix of actionable practices, organized by obligation degree and organizational role. This matrix is positioned as dynamic, role-specific guidance that bridges the principles-to-practice gap in AI governance, extending beyond abstract ethical principles to operationalized steps for implementation and sustainment.

Significance. If the synthesis proves robust, the work would contribute a modular governance scaffold to IS scholarship on AI fairness, offering practical value for organizations seeking role-specific and lifecycle-aware fairness practices rather than high-level principles alone.

major comments (2)

[Abstract / Methods] Abstract and methods description: The central claim that the analyses produce a 'comprehensive' and 'actionable' matrix rests on the representativeness of the 60 sources and the reproducibility of the thematic extraction. However, no source selection criteria, sampling strategy, coding protocol, inter-rater agreement metrics, or validation steps for mapping themes to matrix cells (obligation/role) are provided. This directly affects the generalizability asserted in the abstract and weakens the assertion that the output consists of operational practices rather than restatements of principles.
[Abstract] Abstract: The weakest assumption—that the inductive synthesis yields practices 'sufficiently generalizable across different organizations, sectors, and cultural contexts'—is not supported by any reported checks for selection bias or interpretive subjectivity. Without these, the matrix cannot be confidently presented as a 'modular governance scaffold' applicable beyond the sampled sources.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback on our manuscript. We address each major comment below, indicating revisions where appropriate to enhance methodological transparency and temper claims.

read point-by-point responses

Referee: [Abstract / Methods] Abstract and methods description: The central claim that the analyses produce a 'comprehensive' and 'actionable' matrix rests on the representativeness of the 60 sources and the reproducibility of the thematic extraction. However, no source selection criteria, sampling strategy, coding protocol, inter-rater agreement metrics, or validation steps for mapping themes to matrix cells (obligation/role) are provided. This directly affects the generalizability asserted in the abstract and weakens the assertion that the output consists of operational practices rather than restatements of principles.

Authors: We agree that the abstract and methods section lack these details. The full manuscript describes discourse and thematic analyses of 60 sources but does not specify selection criteria, sampling strategy, coding protocol, inter-rater metrics, or validation for matrix mapping. We will revise the methods section to add this information, including source identification through targeted searches across academic databases, policy repositories, and practitioner reports (2018–2023), application of established thematic analysis phases, and steps for mapping themes to lifecycle, obligation, and role dimensions. This will better demonstrate operationalization and support reproducibility. revision: yes
Referee: [Abstract] Abstract: The weakest assumption—that the inductive synthesis yields practices 'sufficiently generalizable across different organizations, sectors, and cultural contexts'—is not supported by any reported checks for selection bias or interpretive subjectivity. Without these, the matrix cannot be confidently presented as a 'modular governance scaffold' applicable beyond the sampled sources.

Authors: The abstract bases generalizability on the diversity of the 60 sources, but no formal bias checks or subjectivity measures are reported. As an inductive qualitative synthesis, the study did not include such empirical validations. We will revise the abstract to qualify the claim, describing the matrix as adaptable guidance derived from a broad sample rather than universally applicable. A new limitations section will discuss selection bias, interpretive subjectivity, and the value of future context-specific testing. This addresses the concern while preserving the contribution as a modular scaffold. revision: partial

Circularity Check

0 steps flagged

No significant circularity in inductive synthesis from external sources

full rationale

The paper performs discourse and thematic analysis on 60 external academic, policy, and practitioner sources to derive an organized matrix of AI fairness practices spanning the AI lifecycle. This is a standard inductive synthesis where the output is an aggregation and structuring of independent inputs rather than a self-referential derivation. No equations, fitted parameters, quantitative predictions, or self-citations that bear the load of the central claim are present in the described derivation chain. The matrix represents an organization of findings from the literature and does not reduce to its inputs by construction or via any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that qualitative synthesis from a curated set of sources can yield generalizable, actionable practices; this depends on untested premises about source representativeness and minimal researcher bias in thematic extraction.

axioms (2)

domain assumption Discourse and thematic analysis of academic, policy, and practitioner sources can reliably extract actionable AI fairness practices.
Invoked as the method for deriving the matrix from the 60 sources.
domain assumption A socio-technical and praxiological lens is the appropriate frame for identifying fairness practices.
Used to guide the overall analysis approach.

pith-pipeline@v0.9.0 · 5432 in / 1268 out tokens · 44794 ms · 2026-05-10T03:15:26.780340+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

[1]

treating others as they should or deserve to be treated by adhering to standards of right and wrong

Thirty-Fourth European Conference on Information Systems (ECIS 2026), Milan, Italy 1 MOVING BEYOND PRINCIPLES: IDENTIFYING ACTIONABLE AI FAIRNESS PRACTICES Completed Research Paper Christoph Burtscher, University of Reading, Reading, United Kingdom, c.burtscher@reading.ac.uk Mateusz Dolata, Zeppelin University, Friedrichshafen, Germany, mateusz.dolata@zu....

work page 2026
[2]

Discriminatory non-harm based on equality and non-discrimination (Ryan & Stahl,

and incorporates AI fairness as a key principle (Jobin et al., 2019). Discriminatory non-harm based on equality and non-discrimination (Ryan & Stahl,

work page 2019
[3]

the perception and understanding of fairness in organizations

for equal treatment and outcomes is central to AI fairness (Leslie et al., 2023). Biases underlie unfairness, and fairness mitigates their effects (Li et al., 2023). Fairness and its perceptions are often studied through the lens of organizational justice (Li et al., 2014). It refers to "the perception and understanding of fairness in organizations" (Kals...

work page 2023
[4]

specifying the framework for decision rights and accountabilities to encourage desirable behavior in the use of IT

differentiate it from other responsible AI domains. Additionally, the complex nature of fairness perceptions, as explicated by its dimensions, warrants attention. Overall, it calls for a more adaptive, dynamic, and differentiated approach towards AI fairness and its governance (Chowdhury & Klautzer, 2025). 2.2 Governance Corporate governance provides the ...

work page 2025
[5]

Various frameworks are available for corporate or IT governance, but few focus on AI Governance (Wirtz et al., 2020)

defines AI governance as a "system of rules, practices, processes, and technological tools that are employed to ensure an organization's use of AI technologies aligns with the organization's strategies, objectives, and values; fulfills legal requirements; and meets principles of ethical AI followed by the organization.". Various frameworks are available f...

work page 2020
[6]

Principles, on the other hand, act as guides to action (Whittlestone et al.,

and are persuasive (Jobin et al., 2019). Principles, on the other hand, act as guides to action (Whittlestone et al.,

work page 2019
[7]

2.3 AI Fairness Governance This paper focuses on AI fairness and its governance

and aspirations. 2.3 AI Fairness Governance This paper focuses on AI fairness and its governance. Fairness is a key component of responsible AI. Based on Mäntymäki et al. (2022) AI governance definition, we define AI fairness governance as a system of rules, practices, processes, and technological tools that are employed to ensure an organization's use of...

work page 2022
[8]

Such studies intend to translate norms into operations, pattern-oriented engineering approaches, sociotechnical maturity criteria, and operationalizations

recent works. Such studies intend to translate norms into operations, pattern-oriented engineering approaches, sociotechnical maturity criteria, and operationalizations. The underlying claim is that responsible AI outcomes depend on what actors actually do with tools, rules, and structures over time rather than on declarations of values or isolated model ...

work page 2023
[9]

all forms of talk and texts

that lack actionable detail tailored to specific stakeholders, leaving practitioners without concrete implementation pathways. Third, coverage is fragmented, focusing on �� Short Title (up to 5 words) Thirty-Fourth European Conference on Information Systems (ECIS 2026), Milan, Italy 5 discrete softw...

work page 2026
[10]

AI Fairness

Data Corpus and Search Details. The unit of analysis is an object, i.e., a publication of a newspaper article, an academic manuscript, or an interview. Each object is analyzed for two content types: what and how it is communicated. All data corpus articles are incorporated in their original form and not summarized or corrected (Gill, 2000). We combined th...

work page 2000
[11]

These were documented in the updated thematic map to reflect analytical clarity and thematic coherence

Define and Name Themes We then named and described the final themes, categories, and sub-categories, including definitions, narrative synopses, and relational mappings. These were documented in the updated thematic map to reflect analytical clarity and thematic coherence. 1 The details of the final data corpus can be accessed on https://osf.io/r732y/overv...

work page 2026
[12]

Train users/employees

Thematic Analysis Phases and Implemented Activities. Connections between sub-categories and sub-categories to categories were established based on associations between codes, and then hierarchically ordered as part of step three in the thematic analysis. For example, the code "Train users/employees" describes the STS element of a task part of the sub-cate...

work page 2022
[13]

and inform decisions (Chatterjee et al., 2024), their proliferation (Floridi & Cowls, 2021; Morley et al.,

work page 2024
[14]

This gap stems from vague, practitioner-inadequate, and non-actionable guidance (Baldassarre et al., 2024; Morley et al., 2023)

has not closed the translation gap, i.e., the failure to convert principles into practices (Ali et al., 2023; Mittelstadt, 2019). This gap stems from vague, practitioner-inadequate, and non-actionable guidance (Baldassarre et al., 2024; Morley et al., 2023). Principles are implemented through practices (Seppälä et al., 2021). We propose AI fairness practi...

work page 2023
[15]

regimes of a mediated object-oriented performance of an organized set of sayings and doings

states that activities unfold through materially mediated practices (Schatzki, 2001). Practices are "regimes of a mediated object-oriented performance of an organized set of sayings and doings" (Nicolini, 2017, p. 5), representing technology as material objects, as social relations between agents and technology, and as institutional properties (Nicolini e...

work page 2001
[16]

of collaboration engineering (Briggs et al., 2006), of tools, configuration, and scripts (Briggs & De Vreede, 2009; Schwabe et al.,

work page 2006
[17]

are covered. STS serves as our lens for examining how AI fairness practices are structured inclusively, addressing gaps such as the overemphasis on technical guidance and the underrepresentation of context and agent influence on AI fairness. Originating with Trist and Bamforth’s (1951) study, STS has, over time, become a cohesive axis of interest for info...

work page 1951
[18]

data are used in a manner most useful to stakeholders

AI Fairness Governance Thematic Map. Various interactions were identified and depicted as relationships. The theme of AI Fairness Practices encompasses practices for each lifecycle phase and foundational practices (i.e., sub-categories). For example, the identified AI fairness definition is related to the category AI Ethics, as AI fairness forms one part ...

work page 2019
[19]

extensive internal testing to validate the effectiveness of any AI

and "extensive internal testing to validate the effectiveness of any AI" (King, 2024, p. 2). Each component and its definition of an AI fairness practice are based on the codes allocated in the data corpus. They include the STS elements context, task, tool, and actor, which are interdependent as per STS, but were amended with other identified components, ...

work page 2024
[20]

doing of fairness

ensures concreteness and operationalization of AI fairness practices. Each effective practice is intended to be evidence-based (grounded in empirical data), context-sensitive, legitimate, embedded, inclusive, and aligned with the goal of AI fairness. The structure of practice definitions is influenced first by the context-mechanism-outcome (CMO) structure...

work page 1997
[21]

Phase Development Name 4.3 Testing Definition Outcome The Testing practice validates AI and ensures all system components are fair and compliant

This table not only details, based on the coding, hierarchical ordering, and relational connections, the AI fairness practices, but especially specifies the STS elements. Phase Development Name 4.3 Testing Definition Outcome The Testing practice validates AI and ensures all system components are fair and compliant. Subsystems It integrates adversarial tes...

work page 2026
[22]

AI Fairness Practice Matrix Example Testing Practice. 5 Discussion This study examined socio-technical practices for AI fairness and operationalized AI fairness through actionable practices that integrate technical and social dimensions. The thematic map illustrates the interconnected nature of AI fairness and provides details on its context-dependence, m...

work page 2023
[23]

while also incorporating organizational and processual aspects, such as the audience and applicability scope. Given that governance guidance frequently remains vague in terms of who should apply it and when (Floridi & Cowls, 2021), the combination of relatively specific scripts with those aspects provides essential advantages for companies that want to em...

work page 2021
[24]

fair to whom?

by introducing new accountability demands within AI governance (Birkstedt et al., 2023). This aligns with institutional logic (Kostova & Zaheer, 1999), which guides behavior through practices (Thornton & Ocasio, 1999, p. 804). The different identified audiences of the top management team, the manager, and the engineer could perceive tensions between role-...

work page 2023
[25]

T., Gigante, D., Kalinowski, M., & Ragone, A

Baldassarre, M. T., Gigante, D., Kalinowski, M., & Ragone, A. (2024). POLARIS: A framework to guide the development of trustworthy AI systems. Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering-Software Engineering for AI, Bargh, M. S., Choenni, S., & Ter Braak, F. (2025). Algorithmic fairness as sociotechnical system. 26th Annual ...

work page arXiv 2024
[26]

Christian, B. (2021). The alignment problem: How can machines learn human values? Atlantic Books Colquitt, J. A. (2012). Organizational justice. In S. W. J. Kozlowski (Ed.), The Oxford Handbook of Organizational Psychology (pp. 526-547). Oxford University Press Colquitt, J. A., Greenberg, J., & Zapata-Phelan, C. P. (2005). What is organizational justice? ...

work page 2021
[27]

Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389-399. Jones Suzuki, G., & Namireddy, M. (2024). Guidance for Implementing Responsible AI in Legal and Business Practice. R. Gray. https://www.ropesgray.com/en/insights/alerts/2024/05/guidance-for-implementing-responsible-ai-in-leg...

work page doi:10.36227/techrxiv.170905835.50964792/v1 2019
[28]

Whittlestone, J., Nyrup, R., Alexandrova, A., & Cave, S. (2019). The role and limits of principles in AI ethics: Towards a focus on tensions ACM Conference on AI Ethics and Society, Wirtz, B. W., Weyerer, J. C., & Sturm, B. J. (2020). The dark sides of Artificial Intelligence: An integrated AI governance framework for public administration. International ...

work page 2019

[1] [1]

treating others as they should or deserve to be treated by adhering to standards of right and wrong

Thirty-Fourth European Conference on Information Systems (ECIS 2026), Milan, Italy 1 MOVING BEYOND PRINCIPLES: IDENTIFYING ACTIONABLE AI FAIRNESS PRACTICES Completed Research Paper Christoph Burtscher, University of Reading, Reading, United Kingdom, c.burtscher@reading.ac.uk Mateusz Dolata, Zeppelin University, Friedrichshafen, Germany, mateusz.dolata@zu....

work page 2026

[2] [2]

Discriminatory non-harm based on equality and non-discrimination (Ryan & Stahl,

and incorporates AI fairness as a key principle (Jobin et al., 2019). Discriminatory non-harm based on equality and non-discrimination (Ryan & Stahl,

work page 2019

[3] [3]

the perception and understanding of fairness in organizations

for equal treatment and outcomes is central to AI fairness (Leslie et al., 2023). Biases underlie unfairness, and fairness mitigates their effects (Li et al., 2023). Fairness and its perceptions are often studied through the lens of organizational justice (Li et al., 2014). It refers to "the perception and understanding of fairness in organizations" (Kals...

work page 2023

[4] [4]

specifying the framework for decision rights and accountabilities to encourage desirable behavior in the use of IT

differentiate it from other responsible AI domains. Additionally, the complex nature of fairness perceptions, as explicated by its dimensions, warrants attention. Overall, it calls for a more adaptive, dynamic, and differentiated approach towards AI fairness and its governance (Chowdhury & Klautzer, 2025). 2.2 Governance Corporate governance provides the ...

work page 2025

[5] [5]

Various frameworks are available for corporate or IT governance, but few focus on AI Governance (Wirtz et al., 2020)

defines AI governance as a "system of rules, practices, processes, and technological tools that are employed to ensure an organization's use of AI technologies aligns with the organization's strategies, objectives, and values; fulfills legal requirements; and meets principles of ethical AI followed by the organization.". Various frameworks are available f...

work page 2020

[6] [6]

Principles, on the other hand, act as guides to action (Whittlestone et al.,

and are persuasive (Jobin et al., 2019). Principles, on the other hand, act as guides to action (Whittlestone et al.,

work page 2019

[7] [7]

2.3 AI Fairness Governance This paper focuses on AI fairness and its governance

and aspirations. 2.3 AI Fairness Governance This paper focuses on AI fairness and its governance. Fairness is a key component of responsible AI. Based on Mäntymäki et al. (2022) AI governance definition, we define AI fairness governance as a system of rules, practices, processes, and technological tools that are employed to ensure an organization's use of...

work page 2022

[8] [8]

Such studies intend to translate norms into operations, pattern-oriented engineering approaches, sociotechnical maturity criteria, and operationalizations

recent works. Such studies intend to translate norms into operations, pattern-oriented engineering approaches, sociotechnical maturity criteria, and operationalizations. The underlying claim is that responsible AI outcomes depend on what actors actually do with tools, rules, and structures over time rather than on declarations of values or isolated model ...

work page 2023

[9] [9]

all forms of talk and texts

that lack actionable detail tailored to specific stakeholders, leaving practitioners without concrete implementation pathways. Third, coverage is fragmented, focusing on �� Short Title (up to 5 words) Thirty-Fourth European Conference on Information Systems (ECIS 2026), Milan, Italy 5 discrete softw...

work page 2026

[10] [10]

AI Fairness

Data Corpus and Search Details. The unit of analysis is an object, i.e., a publication of a newspaper article, an academic manuscript, or an interview. Each object is analyzed for two content types: what and how it is communicated. All data corpus articles are incorporated in their original form and not summarized or corrected (Gill, 2000). We combined th...

work page 2000

[11] [11]

These were documented in the updated thematic map to reflect analytical clarity and thematic coherence

Define and Name Themes We then named and described the final themes, categories, and sub-categories, including definitions, narrative synopses, and relational mappings. These were documented in the updated thematic map to reflect analytical clarity and thematic coherence. 1 The details of the final data corpus can be accessed on https://osf.io/r732y/overv...

work page 2026

[12] [12]

Train users/employees

Thematic Analysis Phases and Implemented Activities. Connections between sub-categories and sub-categories to categories were established based on associations between codes, and then hierarchically ordered as part of step three in the thematic analysis. For example, the code "Train users/employees" describes the STS element of a task part of the sub-cate...

work page 2022

[13] [13]

and inform decisions (Chatterjee et al., 2024), their proliferation (Floridi & Cowls, 2021; Morley et al.,

work page 2024

[14] [14]

This gap stems from vague, practitioner-inadequate, and non-actionable guidance (Baldassarre et al., 2024; Morley et al., 2023)

has not closed the translation gap, i.e., the failure to convert principles into practices (Ali et al., 2023; Mittelstadt, 2019). This gap stems from vague, practitioner-inadequate, and non-actionable guidance (Baldassarre et al., 2024; Morley et al., 2023). Principles are implemented through practices (Seppälä et al., 2021). We propose AI fairness practi...

work page 2023

[15] [15]

regimes of a mediated object-oriented performance of an organized set of sayings and doings

states that activities unfold through materially mediated practices (Schatzki, 2001). Practices are "regimes of a mediated object-oriented performance of an organized set of sayings and doings" (Nicolini, 2017, p. 5), representing technology as material objects, as social relations between agents and technology, and as institutional properties (Nicolini e...

work page 2001

[16] [16]

of collaboration engineering (Briggs et al., 2006), of tools, configuration, and scripts (Briggs & De Vreede, 2009; Schwabe et al.,

work page 2006

[17] [17]

are covered. STS serves as our lens for examining how AI fairness practices are structured inclusively, addressing gaps such as the overemphasis on technical guidance and the underrepresentation of context and agent influence on AI fairness. Originating with Trist and Bamforth’s (1951) study, STS has, over time, become a cohesive axis of interest for info...

work page 1951

[18] [18]

data are used in a manner most useful to stakeholders

AI Fairness Governance Thematic Map. Various interactions were identified and depicted as relationships. The theme of AI Fairness Practices encompasses practices for each lifecycle phase and foundational practices (i.e., sub-categories). For example, the identified AI fairness definition is related to the category AI Ethics, as AI fairness forms one part ...

work page 2019

[19] [19]

extensive internal testing to validate the effectiveness of any AI

and "extensive internal testing to validate the effectiveness of any AI" (King, 2024, p. 2). Each component and its definition of an AI fairness practice are based on the codes allocated in the data corpus. They include the STS elements context, task, tool, and actor, which are interdependent as per STS, but were amended with other identified components, ...

work page 2024

[20] [20]

doing of fairness

ensures concreteness and operationalization of AI fairness practices. Each effective practice is intended to be evidence-based (grounded in empirical data), context-sensitive, legitimate, embedded, inclusive, and aligned with the goal of AI fairness. The structure of practice definitions is influenced first by the context-mechanism-outcome (CMO) structure...

work page 1997

[21] [21]

Phase Development Name 4.3 Testing Definition Outcome The Testing practice validates AI and ensures all system components are fair and compliant

This table not only details, based on the coding, hierarchical ordering, and relational connections, the AI fairness practices, but especially specifies the STS elements. Phase Development Name 4.3 Testing Definition Outcome The Testing practice validates AI and ensures all system components are fair and compliant. Subsystems It integrates adversarial tes...

work page 2026

[22] [22]

AI Fairness Practice Matrix Example Testing Practice. 5 Discussion This study examined socio-technical practices for AI fairness and operationalized AI fairness through actionable practices that integrate technical and social dimensions. The thematic map illustrates the interconnected nature of AI fairness and provides details on its context-dependence, m...

work page 2023

[23] [23]

while also incorporating organizational and processual aspects, such as the audience and applicability scope. Given that governance guidance frequently remains vague in terms of who should apply it and when (Floridi & Cowls, 2021), the combination of relatively specific scripts with those aspects provides essential advantages for companies that want to em...

work page 2021

[24] [24]

fair to whom?

by introducing new accountability demands within AI governance (Birkstedt et al., 2023). This aligns with institutional logic (Kostova & Zaheer, 1999), which guides behavior through practices (Thornton & Ocasio, 1999, p. 804). The different identified audiences of the top management team, the manager, and the engineer could perceive tensions between role-...

work page 2023

[25] [25]

T., Gigante, D., Kalinowski, M., & Ragone, A

Baldassarre, M. T., Gigante, D., Kalinowski, M., & Ragone, A. (2024). POLARIS: A framework to guide the development of trustworthy AI systems. Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering-Software Engineering for AI, Bargh, M. S., Choenni, S., & Ter Braak, F. (2025). Algorithmic fairness as sociotechnical system. 26th Annual ...

work page arXiv 2024

[26] [26]

Christian, B. (2021). The alignment problem: How can machines learn human values? Atlantic Books Colquitt, J. A. (2012). Organizational justice. In S. W. J. Kozlowski (Ed.), The Oxford Handbook of Organizational Psychology (pp. 526-547). Oxford University Press Colquitt, J. A., Greenberg, J., & Zapata-Phelan, C. P. (2005). What is organizational justice? ...

work page 2021

[27] [27]

Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389-399. Jones Suzuki, G., & Namireddy, M. (2024). Guidance for Implementing Responsible AI in Legal and Business Practice. R. Gray. https://www.ropesgray.com/en/insights/alerts/2024/05/guidance-for-implementing-responsible-ai-in-leg...

work page doi:10.36227/techrxiv.170905835.50964792/v1 2019

[28] [28]

Whittlestone, J., Nyrup, R., Alexandrova, A., & Cave, S. (2019). The role and limits of principles in AI ethics: Towards a focus on tensions ACM Conference on AI Ethics and Society, Wirtz, B. W., Weyerer, J. C., & Sturm, B. J. (2020). The dark sides of Artificial Intelligence: An integrated AI governance framework for public administration. International ...

work page 2019