The Forgotten Margins of AI Ethics

Birhane, Abeba; Ruane, Elayne; Thomas, L.; Brown, Matthew S.; Flowers, Johnathan; Ventresque, Anthony; Dancy, Christopher L.

doi:10.1145/3531146.3533157

Cited by 58 publications

(34 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Here we quote Birhane et. al [11] who measure the scale of this neglect across four years of proceedings in FAccT and the closely related AI, Ethics, and Society conference:…”

Section: Discussionmentioning

confidence: 99%

“…Marketing materials produced by Pymetrics and those produced by other companies offering algorithmic employee selection tools, such as HireVue, suggest that their products are tools for addressing employment discrimination. 11 Following on this market positioning, Pymetrics calibrates their model to meet a specific employment discrimination standard. In the United States, legal protections for job applicants against adverse impact in employee selection are found in the Uniform Guidelines on Employee Selection Procedures (1978), which defines "adverse impact" as a "selection rate for any race, sex, or ethnic group which is less than four-fifths (4/5) (or eighty percent) of the rate for the group with the highest rate... " The highest rate is generally assumed to be cisgender men of European descent but includes anyone who does not fall into a category described by one of the legally protected classes.…”

Section: Publishing and Certifying Corporate Apologiamentioning

confidence: 99%

“…In this paper, we argue that the enthusiastic participation of technology firms in the ACM FAccT conference is in service of certifying their commitment to ethics [28,50,60,64,65] and add that corporate entanglement forecloses the opportunity for these firms to be confronted by narratives of the real-world consequences of AI harms (c.f. particularly Birhane et al [11], who provide robust evidence for this neglect at both FAccT and the closely related conference on AI, Ethics, and Society; and Greene et al [28] who find that critiques of AI/ML are co-opted by firms into a technological solutionistic, expert lens). Instead of interrogating AI from a political perspective, FAccT contributions frequently limit their gaze to the "technical" part of "sociotechnical"-the level of data, metadata, or models [24,28].…”

mentioning

confidence: 99%

See 2 more Smart Citations

Confronting Power and Corporate Capture at the FAccT Conference

Young

Katell

Krafft

2022

2022 ACM Conference on Fairness, Accountability, and Transparency

View full text Add to dashboard Cite

Fields such as medicine and public health attest to deep conflict of interest concerns present when private companies fund evaluation of their own products and services. We draw on these lessons to consider corporate capture of the ACM Fairness, Accountability, and Transparency (FAccT) conference. We situate our analysis within scholarship on the entanglement of industry and academia and focus on the silences it produces in the research record. Our analysis of the institutional design at FAccT indicates the conference's neglect of those people most negatively impacted by algorithmic systems. We focus on a 2021 paper by Wilson et al., "Building and auditing fair algorithms: A case study in candidate screening" as a key example of conflicted research accepted via peer review at FAccT. We call on the conference to (1) lead on models for how to manage conflicts of interest in the field of computing beyond individual disclosure of funding sources, (2) hold space for advocates and activists able to speak directly to questions of algorithmic harm, and (3) reconstitute the conference with attention to fostering agonistic dissensus-un-making the present manufactured consensus and nurturing challenges to power. These changes will position our community to contend with the political dimensions of research on AI harms. CCS CONCEPTS• Social and professional topics → Codes of ethics; Funding; Political speech; Governmental regulations.

show abstract

“…Here we quote Birhane et. al [11] who measure the scale of this neglect across four years of proceedings in FAccT and the closely related AI, Ethics, and Society conference:…”

Section: Discussionmentioning

confidence: 99%

Section: Publishing and Certifying Corporate Apologiamentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Confronting Power and Corporate Capture at the FAccT Conference

Young

Katell

Krafft

2022

2022 ACM Conference on Fairness, Accountability, and Transparency

View full text Add to dashboard Cite

show abstract

“…This intersection is often characterised by the 'FAccT' (Fairness, Accountability and Transparency) agenda, which sees the technical community trying to develop ways and means of tackling principled ethical concerns [e.g., 43,60,61,62]. However, Birhane et al [6] analysis of the ACM FAccT Conference series papers finds ethics is often treated in an abstract manner e.g., fairness as a mathematical or philosophical problem, whilst neglecting the situated historical and societal power structures, actual harms, and the lived experiences of marginalised groups impacted by AI systems. Fairness, accountability, and transparency are not only ethical and technical concepts 6 ; they are key to good governance which requires governments and industry treat people equally without discrimination, ensure ongoing answerability for decision-making, and provide openness around decisions [13].…”

Section: The Legal Concept Of Trustworthinessmentioning

confidence: 99%

Nordic Human-Computer Interaction Conference

2022

View full text Add to dashboard Cite

We propose a series of legal provocations emerging from the proposed European Union Artificial Intelligence Act 2021 (AIA) and explore how they open up new possibilities for HCI in the design and development of trustworthy autonomous systems. The AIA continues the 'by design' trend seen in recent EU regulation of emerging technologies. The AIA targets AI developments that pose risks to society and citizens' fundamental rights, introducing mandatory design and development requirements for high-risk AI systems (HRAIS). These requirements regulate different stages of the AI development cycle including ensuring data quality and governance strategies, mandating testing of systems, ensuring appropriate risk management, designing for human oversight, and creating technical documentation. These requirements open up new opportunities for HCI that reach beyond established concerns with the ethics and explainability of AI and situate AI development in human-centered processes and methods of design to enable compliance with regulation and foster societal trust in AI. CCS CONCEPTS • Human computer interaction (HCI) • Artificial intelligence • Applied computing → Law Additional Keywords and Phrases: European Union AI Act (AIA), trustworthy AI, mandatory design and development requirements.ACM Reference Format: This block will be automatically generated when manuscripts are processed after acceptance."The proposal is without prejudice and complements the General Data Protection Regulation and the Law Enforcement Directive with a set of harmonised rules applicable to the design, development and use of certain highrisk AI systems" (European Commission, Proposal for an AI Act).

show abstract

“…fizierung und Abstrahierung von Fairness kann davon ablenken, wie KI-Systeme in strukturellen Diskriminierungsdynamiken verankert sind(John-Mathews et al 2022) und wie diese sich konkret im Leben von unterprivilegierten Personen niederschlagen(Birhane et al 2022). Das tangiert auch andere Voraussetzungen für Fairness wie Transparenz und Erklärbarkeit, da Betroffene der Entscheidungen von Hochrisiko-Systemen z.…”

unclassified

Algorithmische Fairness in der polizeilichen Ermittlungsarbeit:Ethische Analyse von Verfahren des maschinellen Lernens zur Gesichtserkennung

Brandner¹,

Hirsbrunner²

2023

TATuP

View full text Add to dashboard Cite

This article discusses fairness in artificial intelligence (AI) based policing procedures using facial recognition as an example. Algorithmic decisions based on discriminatory dynamics can (re)produce and automate injustice. AI fairness here concerns not only the creation and sharing of datasets or the training of models but also how systems are deployed in the real world. Quantifying fairness can distract rom how discrimination and oppression translate concretely into social phenomena. Integrative approaches can help actively incorporate ethical, legal, social, and economic factors into technology development to more holistically assess the consequences of deployment through continuous interdisciplinary collaboration.

show abstract

The Forgotten Margins of AI Ethics

Cited by 58 publications

References 33 publications

Confronting Power and Corporate Capture at the FAccT Conference

Confronting Power and Corporate Capture at the FAccT Conference

Nordic Human-Computer Interaction Conference

Algorithmische Fairness in der polizeilichen Ermittlungsarbeit:Ethische Analyse von Verfahren des maschinellen Lernens zur Gesichtserkennung

Contact Info

Product

Resources

About