Pith. sign in

REVIEW 2 major objections 4 minor 90 references

"Death by a thousand taxonomies?": AI Risk Classification In Practice

T0 review · 2 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read AI risk taxonomies are weakly integrated into AI governance, with two design–use features explaining why: invisible reductive choices and missing causal links to decision points.

desk verdict Solid qualitative study of AI risk taxonomy practices, with a central claim that overreaches from developers' lack of visibility to weak integration. read the letter →

arxiv 2608.06831 v1 pith:PGBWM5UU submitted 2026-08-07 cs.CY

Glen Berman , Ned Cooper , Angel Hsing-Chi Hwang , Wesley Hanwen Deng , Renee Shelby , Ben Hutchinson This is my paper · ORCID
classification cs.CY
keywords AIgovernancesociotechnicaloutcometaxonomiesharmclassificationresponsibleaccountabilityqualitativeinterviewsstandardisationriskassessment
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

AI risk classification schemes — sociotechnical outcome taxonomies (SOT) — are meant to give regulators, firms, and policymakers a structured account of AI harms to act on. Based on 25 interviews with developers and users across industry, academia, civil society, and government, the paper argues that SOT are only weakly integrated into actual AI governance: there is little evidence they inform deployment decisions or post-deployment monitoring. It identifies two design–use features behind this gap: the reductive choices that shape a taxonomy are invisible to downstream users, who treat categories as exhaustive accounts of risk rather than interpretive aids; and SOT enumerate harms without linking them to decision points or responsible actors, so accountability is hard to assign. The paper's recommendations — interoperability, extensibility, and traceability in SOT design, plus ecosystem-level infrastructure such as a shared registry — follow from these findings.

What carries the argument

The central object is the sociotechnical outcome taxonomy (SOT): a structured classification of AI risks, harms, and impacts intended to serve as a boundary object and proto-standard for AI governance. The argument is carried by two identified design–use features that explain weak integration: the invisibility of the reductive choices made during taxonomy construction, which leads users to treat categories as exhaustive; and the absence of causal pathways linking categories to decision points and actors, which makes accountability unassignable. The paper also introduces the 'altitude problem' — choosing the right level of abstraction for categories — as the central design tension SOT developers face, and proposes three design desiderata (interoperability, extensibility, traceability) plus an ecosystem-level registry as the remediation.

What would settle it

A reader could audit a sample of published SOT (for instance those catalogued by the AI Risk Repository) and count whether most include explicit causal pathways naming decision points and responsible actors; if a majority did, the paper's claim that SOT lack causal linkage would be falsified. Alternatively, direct observation of product teams during pre-deployment review could check whether risk categories from SOT are routinely traced to named owners and post-launch monitoring plans.

Watch

Extended reading notes

Core claim

The paper's central finding is that sociotechnical outcome taxonomies are not well integrated into AI governance processes: participants reported proliferation without coordination, no visibility into adoption, and little evidence that enumerated risks are monitored after deployment. The authors argue that two features explain this. First, the design choices through which an SOT reduces the complexity of AI risks are invisible to users, so frameworks intended as interpretive aids are misused as exhaustive descriptions of risk. Second, SOT generally describe harms without analysing how they arise, leaving out the decision points and actors that could be held responsible, which lets classification substitute for substantive action. These are presented as empirically grounded findings from reflexive thematic analysis of interviews, not as a formal evaluation of SOT effectiveness.

Load-bearing premise

The load-bearing assumption is that 25 interviewees, all recruited through prior engagement with SOT and almost all based in North America or Europe, with no Global South representation, give a reliable picture of how SOT are used across AI governance worldwide.

Editorial extensions

If this is right

  • If the paper is right, SOT currently provide limited governance value; simply producing more taxonomies will not fix the integration gap.
  • SOT design should make the reductive choices visible through metadata, boundary conditions, and documented mappings so users treat categories as interpretive aids rather than exhaustive descriptions.
  • SOT should include lightweight causal pathways and explicit observability requirements for classified risks, so accountability can be assigned and monitoring expectations attached.
  • Premature standardisation across SOT risks freezing the definitional asymmetries and entrenched interests; a shared registry that aggregates SOT along with their documentation and monitoring expectations is a more tractable first step.
  • Future empirical research should directly observe SOT in use across industry, civil society, and regulatory agencies rather than relying on retrospective accounts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the invisibility hypothesis is right, a low-cost intervention is to require published SOT to carry explicit design-rationale metadata (definitions, boundary conditions, evidence base), which would also make the 'altitude' choices contestable.
  • The missing causal-linkage result suggests SOT-based evaluations will keep tracking measurable proxies unless regulators require documented links from categories to decisions; the EU AI Act's risk classification could be a natural testbed.
  • The paper's registry proposal could double as a monitoring infrastructure: attaching reporting expectations to categories at deposit time would turn classification into an enforceable governance relation.
  • The geographic skew of the sample implies the proposed fixes should be stress-tested against how SOT are used in the Global South, where cultural erasure and diffuse societal harms may be more salient than the corporate reputational dynamics described.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. This paper reports a qualitative interview study of how sociotechnical outcome taxonomies (SOT) for AI risks are developed and used, based on 25 semi-structured interviews with researchers and practitioners in industry, academia, civil society, and government. The central claim is that SOT are weakly integrated into AI governance processes, with little evidence that they inform deployment decisions or post-deployment monitoring. The authors identify two design/use features that explain this: (1) the reductive design choices behind SOT are invisible to downstream users, leading categories to be treated as exhaustive descriptions of risk rather than interpretive aids, and (2) SOT enumerate harms without linking them to decision points or responsible actors, making accountability difficult to assign. The paper closes with design recommendations (interoperability, extensibility, traceability) and an ecosystem-level proposal for a registry of SOT that would support monitoring and standardisation only after stronger infrastructural substrate exists.

Significance. If the findings hold, the paper makes a useful empirical contribution to an area that currently lacks direct qualitative evidence. It is one of the first studies to examine SOT development and use in practice, and it offers concrete, design-oriented recommendations with practical relevance to AI governance, RAI tooling, and standardisation efforts. The authors are transparent about their inter-sectoral positions, provide a detailed codebook, and include participant quotes that ground the analysis. The two explanatory mechanisms--invisible reduction and missing causal linkage to decision points--are plausible and connect well to existing STS scholarship on classification and infrastructure. The central evidentiary claim about 'weak integration' is, however, not as strongly supported as its general wording suggests: the interview data largely show that taxonomy developers lack visibility into downstream use, not that use is absent. The paper's value would be strengthened considerably if this distinction were made explicit and if the central finding were reframed accordingly.

major comments (2)
  1. [3.1 Recruitment, Table 1, and 6 Limitations] The central claim that 'SOT are not well integrated into AI governance processes, with little evidence of SOT being used to inform AI deployment decisions or the monitoring of AI impacts' conflates participants' lack of visibility into downstream use with an observed absence of use. The quoted evidence--I20's 'I unfortunately have no evidence whether people have found it useful' and I14's 'insight into how [developers] actually do, or if they actually do, use them'--are statements of ignorance about use, not observations of non-use. The paper's own codebook (Appendix C, T4, 'Specific use cases and applications') contains 42 low-level codes documenting concrete reported uses, including pre-deployment ethics review, evaluation design, and guiding product team deliberation. The current wording of the finding therefore overstates the support in the data. I recommend reframing the central finding as, for example, 'SOT developers lack systematic visibility into downstream use, and no tracking infrastructure exists to assess adoption or effects'--a claim that the interview data do support--or, alternatively, providing direct observational or prevalence data that would justify the original 'weak integration' claim.
  2. [4 Findings, first paragraph] The generalisation 'SOT are weakly integrated into AI governance processes' is stated without geographic or sampling qualification, even though the sample is heavily concentrated in North America (18/25), includes no participants from the Global South, and was recruited entirely through prior engagement with SOT, thereby excluding practitioners who encountered SOT and chose not to use them. The Limitations section acknowledges this, but the abstract, findings, and conclusion all repeat the broad claim. Since the paper's contribution is partly a general empirical statement about the state of SOT adoption, this mismatch between claim and evidence is load-bearing. Please qualify the central claim (e.g., 'among SOT-engaged actors in North America and Europe/UK') or present additional evidence that the pattern extends beyond the sampled population.
minor comments (4)
  1. [1 Introduction] There is a typo: 'We refer to these artefacts associotechnical outcome taxonomies' should read 'as sociotechnical outcome taxonomies'.
  2. [Appendix C, final line] The codebook appendix states the full codebook is 'available at https://doi.org/10.5281/zenodo.21830185' but the final line says it 'will be made available as supplementary material'; please make these statements consistent.
  3. [Appendix A, Table 2] The row labelled '2025' for AGORA cites Arnold et al. 2024, while the reference list gives 2024; please align the year in the table with the actual publication year.
  4. [References] The same paper by Lee et al. is listed twice (2024a and 2024b) with identical titles; please merge into a single reference or distinguish the versions if they are genuinely different.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core findings are inductive interpretations of interview data, with self-citations serving as contextual examples rather than load-bearing premises.

full rationale

The paper's central claim that "SOT are not well integrated into AI governance processes, with little evidence of SOT being used to inform AI deployment decisions or the monitoring of AI impacts" is an inductive inference from participant interviews and reflexive thematic analysis, not a consequence of any prior taxonomy or fitted parameter. The supporting quotations (e.g., I20: "I unfortunately have no evidence whether people have found it useful"; I14: little "insight into how [developers] actually do, or if they actually do, use them") report visibility gaps, and the inference from those gaps to weak integration is an interpretive step that could be challenged on evidentiary grounds, but it is not circular: the finding is not true by definition of the interview protocol, nor is it derived from an equation that already contains the conclusion. The two explanatory features—invisible design choices and the absence of causal linkage to decision points and actors—are generated from participants' accounts and presented as themes, not as assumptions smuggled in from prior work. Self-citations appear (e.g., Shelby et al. 2023; Berman, Goyal, and Madaio 2024; Rismani et al. 2023, 2025; Mitchell et al. 2022; Raji et al. 2020), and the positionality statement candidly discloses that one author led the development of an SOT and others contributed to SOT work, but none of these citations is used to define the object of study, to justify the coding scheme, or to establish the central finding. The paper's own Limitations section acknowledges that the sample excludes non-users, lacks Global South representation, and relies on retrospective accounts rather than direct observation; these are generalizability concerns, not circularity. No step in the derivation reduces to its inputs by construction, and no self-citation chain is load-bearing. The derivation is therefore self-contained as a qualitative empirical study, and no circularity score above zero is warranted.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No free parameters or invented entities apply to this qualitative study. The analysis assumes the STS premise that classifications are not neutral, the proto-standard framing of SOT, the reliability of retrospective self-reports, and the representativeness of the sample; these are listed as domain assumptions.

assumptions (4)
  • domain assumption Classification is never neutral; categorisation systems stabilise particular problem framings, rendering some phenomena visible and actionable while marginalising others.
    Adopted from Bowker and Star (2000) and used to structure the analysis (Section 1).
  • domain assumption SOT are proto-standards, that is, standardisation efforts that seek to stabilise the problem space of AI harms.
    Conceptual framing introduced in Section 1 and guiding the study's interpretation.
  • domain assumption Participants' retrospective self-reports are reliable evidence of SOT use and governance integration.
    The study relies on retrospective accounts rather than direct observation; the authors acknowledge this in Limitations (second bullet).
  • domain assumption The sample of 25 participants, despite geographic and sector skew, supports claims about the broader SOT landscape.
    Sampling assumption underlying the generalizing framing of findings (Section 3.1 and Limitations).

how reviews work

0 comments
Cite this review

Pith. "Pith review of "Death by a thousand taxonomies?": AI Risk Classification In Practice." pith.science (2026). https://pith.science/paper/PGBWM5UU

@misc{pith2026260806831,
  author       = {Pith},
  title        = {Pith review of: "Death by a thousand taxonomies?": AI Risk Classification In Practice},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PGBWM5UU}},
  note         = {Machine review of arXiv:2608.06831}
}
read the original abstract

The harms in which AI is implicated range in nature and scope from unsafe user interactions through to the societal-wide consequences of AI adoption. Classification of the diverse risks of AI is foundational to AI governance: regulators, technology firms, and policymakers need structured accounts of risk upon which to act. Researchers and practitioners have accordingly developed many Sociotechnical Outcome Taxonomies (SOT). This paper presents an empirical study of SOT development and use, drawing on 25 interviews with researchers and practitioners across industry, academia, civil society, and government. We find SOT are weakly integrated into AI governance processes, and identify two features of SOT design and use that explain why. First, the design choices through which SOT produce structured representations of the complex problem space of AI risks tend to be invisible to downstream taxonomy users. Those users treat the resulting categories as exhaustive accounts of risk rather than as interpretive aids. Second, SOT typically enumerate harms without linking them to decision points or actors implicated in their occurrence, leaving accountability difficult to assign. We close with design recommendations for SOT developers and users, and argue realising the potential of SOT requires governance infrastructure that does not yet exist.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

90 extracted references · 34 canonical work pages

  1. [1]

    Hutiri, Wiebke and Papakyriakopoulos, Orestis and Xiang, Alice , year = 2024, month = jun, pages =. Not. The 2024

    2024

  2. [2]

    Sociotechnical

    Shelby, Renee and Rismani, Shalaleh and Henne, Kathryn and Moon,. Sociotechnical. Proceedings of the 2023. doi:10.1145/3600211.3604673 , isbn =

    arXiv 2023

  3. [3]

    AI and Ethics , volume=

    AI governance: A systematic literature review , author=. AI and Ethics , volume=. 2025 , publisher=

    2025

  4. [4]

    Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems , articleno =

    Rismani, Shalaleh and Shelby, Renee and Smart, Andrew and Jatho, Edgar and Kroll, Joshua and Moon, AJung and Rostamzadeh, Negar , title =. Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems , articleno =. 2023 , isbn =. doi:10.1145/3544548.3581407 , abstract =

    arXiv 2023

  5. [5]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , volume=

    Measuring what matters: Connecting AI ethics evaluations to system attributes, hazards, and harms , author=. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , volume=

  6. [6]

    Patterns , year=

    The AI risk repository: A meta-review, database, and taxonomy of risks from artificial intelligence , author=. Patterns , year=

  7. [7]

    Taxonomy of

    Weidinger, Laura and Uesato, Jonathan and Rauh, Maribeth and Griffin, Conor and Huang, Po-Sen and Mellor, John and Glaese, Amelia and Cheng, Myra and Balle, Borja and Kasirzadeh, Atoosa and Biles, Courtney and Brown, Sasha and Kenton, Zac and Hawkins, Will and Stepleton, Tom and Birhane, Abeba and Hendricks, Lisa Anne and Rimell, Laura and Isaac, William ...

    arXiv 2022

  8. [8]

    Troubling

    Berman, Glen and Cooper, Ned and Deng, Wesley Hanwen and Hutchinson, Ben , month = oct, year =. Troubling. doi:10.48550/arXiv.2410.22985 , urldate =

    arXiv doi:10.48550/arxiv.2410.22985

Show all 90 references
  1. [9]

    and Sengupta, Ushnish and Suriyawongkul, Arthit and Thelot, Ruby and Vei, Sofia and Waltersdorfer, Laura , month = nov, year =

    Abercrombie, Gavin and Benbouzid, Djalel and Giudici, Paolo and Golpayegani, Delaram and Hernandez, Julio and Noro, Pierre and Pandit, Harshvardhan and Paraschou, Eva and Pownall, Charlie and Prajapati, Jyoti and Sayre, Mark A. and Sengupta, Ushnish and Suriyawongkul, Arthit a...

    doi arXiv

  2. [10]

    A sectoral taxonomy of

    Calvino, Flavio and Dernis, Hélène and Samek, Lea and Ughi, Antonio , year =. A sectoral taxonomy of

  3. [11]

    2024 , month =

    Mary Frances Theofanos and Yee-Yin Choong and Theodore Jensen , title =. 2024 , month =. doi:https://doi.org/10.6028/NIST.AI.200-1 , language =

    2024 doi

  4. [12]

    Feder and Wang, Angelina and Atalla, Chad and Barocas, Solon and Blodgett, Su Lin and Chouldechova, Alexandra and Corvi, Emily and Dow, P

    Wallach, Hanna and Desai, Meera and Cooper, A. Feder and Wang, Angelina and Atalla, Chad and Barocas, Solon and Blodgett, Su Lin and Chouldechova, Alexandra and Corvi, Emily and Dow, P. Alex and Garcia-Gathright, Jean and Olteanu, Alexandra and Pangakis, Nicholas and Reed, Ste...

    2025

  5. [13]

    arXiv preprint arXiv:2503.05780 , year=

    AI risk atlas: Taxonomy and tooling for navigating AI risks and resources , author=. arXiv preprint arXiv:2503.05780 , year=

    arXiv

  6. [14]

    Database , volume=

    A simple standard for sharing ontological mappings (SSSOM) , author=. Database , volume=. 2022 , publisher=

    2022

  7. [15]

    Patterns , volume=

    The AI risk repository: A meta-review, database, and taxonomy of risks from artificial intelligence , author=. Patterns , volume=. 2026 , publisher=

    2026

  8. [16]

    Typology of

    Bird, Charlotte and Ungless, Eddie and Kasirzadeh, Atoosa , month = aug, year =. Typology of. Proceedings of the 2023. doi:10.1145/3600211.3604722 , language =

    2023

  9. [17]

    doi:10.48550/ARXIV.2306.06924 , urldate =

    Critch, Andrew and Russell, Stuart , year =. doi:10.48550/ARXIV.2306.06924 , urldate =

    doi arXiv

  10. [18]

    Adversarial machine learning : a taxonomy and terminology of attacks and mitigations , shorttitle =

    Vassilev, Apostol and Oprea, Alina and Fordyce, Alie and Anderson, Hyrum , month = jan, year =. Adversarial machine learning : a taxonomy and terminology of attacks and mitigations , shorttitle =. doi:10.6028/NIST.AI.100-2e2023 , pages =

    doi

  11. [19]

    Golpayegani, Delaram and Hovsha, Joshua and Rossmaier, Leon W. S. and Saniei, Rana and Misic, Jana , month = sep, year =. Towards a. 2022. doi:10.1109/TransAI54797.2022.00007 , urldate =

    2022

  12. [21]

    Evaluating the

    Solaiman, Irene and Talat, Zeerak and Agnew, William and Ahmad, Lama and Baker, Dylan and Blodgett, Su Lin and Chen, Canyu and Daumé, Hal and Dodge, Jesse and Duan, Isabella and Evans, Ellie and Friedrich, Felix and Ghosh, Avijit and Gohar, Usman and Hooker, Sara and Jernite, ...

    doi arXiv

  13. [22]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =

    Introducing the. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =. 2024 , pages =. doi:10.1609/aies.v7i1.31615 , urldate =

    2024 doi

  14. [23]

    Cui, Tianyu and Wang, Yanling and Fu, Chuanpu and Xiao, Yong and Li, Sijia and Deng, Xinhao and Liu, Yunpeng and Zhang, Qinglin and Qiu, Ziyi and Li, Peiyang and Tan, Zhixing and Xiong, Junwu and Kong, Xinyu and Wen, Zujie and Xu, Ke and Li, Qi , year =. Risk. doi:10.48550/ARX...

    doi arXiv

  15. [24]

    Social Indicators Research , author =

    A. Social Indicators Research , author =. 2022 , pages =. doi:10.1007/s11205-021-02768-7 , language =

    2022 doi

  16. [25]

    Puscas, Ioana , year =

  17. [26]

    2023 , author =

    Taxonomy of Human Rights Risks Connected to Generative AI , url =. 2023 , author =

    2023

  18. [28]

    Harms from

    Chan, Alan and Salganik, Rebecca and Markelius, Alva and Pang, Chris and Rajkumar, Nitarshan and Krasheninnikov, Dmitrii and Langosco, Lauro and He, Zhonghao and Duan, Yawen and Carroll, Micah and Lin, Michelle and Mayhew, Alex and Collins, Katherine and Molamohammadi, Maryam ...

    2023

  19. [29]

    , year =

    Yampolskiy, Roman V. , year =. Taxonomy of

  20. [30]

    Yale JL & Tech

    Algorithms and economic justice:. Yale JL & Tech. , author =. 2020 , note =

    2020

  21. [31]

    Future of artificial intelligence: language, ethics, technology workshop , author =

    A typology of ethical risks in language technology with an eye towards where transparent documentation can help , volume =. Future of artificial intelligence: language, ethics, technology workshop , author =

  22. [32]

    Big Data & Society , author =

    On the controversiality of. Big Data & Society , author =. 2025 , pages =. doi:10.1177/20539517251383870 , language =

    2025 doi

  23. [33]

    and Star, Susan Leigh , year =

    Bowker, Geoffrey C. and Star, Susan Leigh , year =. Sorting Things Out: Classification and its Consequences , isbn =

  24. [34]

    Star, Susan Leigh , year =. The. Distributed. doi:10.1016/B978-1-55860-092-8.50006-X , pages =

    doi

  25. [35]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Where. Proceedings of the ACM on Human-Computer Interaction , author =. 2021 , pages =. doi:10.1145/3449081 , language =

    2021 doi

  26. [36]

    and Stark, Luke and Wortman Vaughan, Jennifer and Wallach, Hanna , month = apr, year =

    Madaio, Michael A. and Stark, Luke and Wortman Vaughan, Jennifer and Wallach, Hanna , month = apr, year =. Co-. doi:10.1145/3313831.3376445 , urldate =

  27. [37]

    Science, Technology, & Human Values , author =

    This is. Science, Technology, & Human Values , author =. 2010 , pages =. doi:10.1177/0162243910377624 , language =

    2010 doi

  28. [38]

    Nature Machine Intelligence , volume=

    Principles alone cannot guarantee ethical AI , author=. Nature Machine Intelligence , volume=. 2019 , publisher=

    2019

  29. [39]

    Science and Engineering Ethics , author =

    From. Science and Engineering Ethics , author =. 2020 , pages =. doi:10.1007/s11948-019-00165-5 , language =

    2020 doi

  30. [40]

    Nature Machine Intelligence , author =

    The global landscape of. Nature Machine Intelligence , author =. 2019 , pages =. doi:10.1038/s42256-019-0088-2 , language =

    2019 doi

  31. [41]

    Greene, Daniel and Hoffmann, Anna Lauren and Stark, Luke , year =. Better,. Hawaii. doi:10.24251/HICSS.2019.258 , url =. 10125/59651 , eprinttype =

    2019 doi

  32. [45]

    Social Studies of Science , author =

    Institutional. Social Studies of Science , author =. 1989 , pages =. doi:10.1177/030631289019003001 , language =

    1989 doi

  33. [46]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =

    Toward a. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , author =. 2025 , pages =. doi:10.1609/aies.v8i3.36745 , number =

    2025 doi

  34. [47]

    , year =

    Hovy, Dirk and Spruit, Shannon L. , year =. The social impact of natural language processing , url =. The 54th

  35. [48]

    The problem with bias:

    Barocas, Solon and Crawford, Kate and Shapiro, Aaron and Wallach, Hanna , year =. The problem with bias:. 9th

  36. [49]

    Annual Review of Sociology , author =

    A. Annual Review of Sociology , author =. 2010 , pages =. doi:10.1146/annurev.soc.012809.102629 , language =

    2010

  37. [50]

    Standardization in

    Timmermans, Stefan and Berg, Marc , date =. Standardization in. 1997 , journal =. doi:10.1177/030631297027002003 , langid =

    1997 doi

  38. [51]

    Treading the

    Klaus Scheuerman, Morgan and Hutiri, Wiebke and Rahmattalabi, Aida and Matthews, Victoria and Xiang, Alice and Andrews, Jerone , date =. Treading the. Proceedings of the 2026. 2026 , pages =. doi:10.1145/3772318.3790364 , eventtitle =

    2026

  39. [52]

    Understanding and

    Gehrmann, Sebastian and Huang, Claire and Teng, Xian and Yurovski, Sergei and Bhorkar, Arjun and Thomas, Naveen and Doucette, John and Rosenberg, David and Dredze, Mark and Rabinowitz, David , date =. Understanding and. Proceedings of the 2025. 2025 , pages =. doi:10.1145/3715...

    2025

  40. [53]

    Li, Megan and Bickersteth, Wendy and Tang, Ningjing and Cranor, Lorrie and Hong, Jason and Shen, Hong and Heidari, Hoda , date =. A. 2025 , booktitle =

    2025

  41. [54]

    Improving

    Holstein, Kenneth and Vaughan, Jennifer Wortman and Daumé, Hal and Dudík, Miroslav and Wallach, Hanna , year =. Improving. doi:10.1145/3290605.3300830 , booktitle =

  42. [55]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Assessing the. Proceedings of the ACM on Human-Computer Interaction , author =. 2022 , keywords =. doi:10.1145/3512899 , language =

    2022 doi

  43. [56]

    Mitchell, Margaret and Pistilli, Giada and Jernite, Yacine and Ozoani, Ezinwanne and Gerchick, Marissa and Rajani, Nazneen and Luccioni, Alexandra Sasha and Solaiman, Irene and Masoud, Maraim and Nikpoor, Somaieh and Ferrandis, Carlos Muñoz and Bekman, Stas and Akiki, Christop...

  44. [57]

    Communications of the ACM , volume=

    Datasheets for Datasets , author=. Communications of the ACM , volume=. 2021 , publisher=

    2021

  45. [59]

    An. Harv. J. Law Technol. , author =. 2021 , keywords =

    2021

  46. [60]

    Costanza-Chock, Sasha and Raji, Inioluwa Deborah and Buolamwini, Joy , month = jun, year =. Who. 2022. doi:10.1145/3531146.3533213 , language =

    2022

  47. [61]

    Berman, Glen and Goyal, Nitesh and Madaio, Michael , month = may, year =. A. Proceedings of the. doi:10.1145/3613904.3642398 , language =

  48. [62]

    Computer Law Review International , author =

    Demystifying the. Computer Law Review International , author =. 2021 , pages =. doi:10.9785/cri-2021-220402 , language =

    2021 doi

  49. [63]

    European Journal of Risk Regulation , author =

    European. European Journal of Risk Regulation , author =. 2025 , pages =. doi:10.1017/err.2025.10032 , language =

    2025

  50. [64]

    International Data Privacy Law , author =

    Algorithmic impact assessments under the. International Data Privacy Law , author =. 2021 , pages =. doi:10.1093/idpl/ipaa020 , language =

    2021 doi

  51. [65]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Seeing. Proceedings of the ACM on Human-Computer Interaction , author =. 2023 , pages =. doi:10.1145/3579621 , language =

    2023 doi

  52. [66]

    Bietti, Elettra , year =. From. doi:10.1145/3351095.3372860 , booktitle =

  53. [67]

    Power to the

    Birhane, Abeba and Isaac, William and Prabhakaran, Vinodkumar and Diaz, Mark and Elish, Madeleine Clare and Gabriel, Iason and Mohamed, Shakir , month = oct, year =. Power to the. doi:10.1145/3551624.3555290 , urldate =

  54. [68]

    Cooper, Ned and Horne, Tiffanie and Hayes, Gillian R and Heldreth, Courtney and Lahav, Michal and Holbrook, Jess and Wilcox, Lauren , month = apr, year =. A. doi:10.1145/3491102.3517716 , language =

  55. [69]

    Design justice: community-led practices to build the worlds we need , isbn =

    Costanza-Chock, Sasha , year =. Design justice: community-led practices to build the worlds we need , isbn =

  56. [70]

    Fairness,

    Kasy, Maximilian and Abebe, Rediet , year =. Fairness,. doi:10.1145/3442188.3445919 , booktitle =

  57. [71]

    Passi, Samir and Barocas, Solon , month = jan, year =. Problem. Proceedings of the. doi:10.1145/3287560.3287567 , language =

  58. [72]

    2025 , isbn =

    Zhang, Renwen and Li, Han and Meng, Han and Zhan, Jinyuan and Gan, Hongyuan and Lee, Yi-Chieh , title =. 2025 , isbn =. doi:10.1145/3706598.3713429 , booktitle =

    2025

  59. [73]

    2024 , eprint=

    Should agentic conversational AI change how we think about ethics? Characterising an interactional ethics centred on respect , author=. 2024 , eprint=

    2024

  60. [74]

    2025 , isbn =

    Chandra, Mohit and Naik, Suchismita and Ford, Denae and Okoli, Ebele and De Choudhury, Munmun and Ershadi, Mahsa and Ramos, Gonzalo and Hernandez, Javier and Bhattacharjee, Ananya and Warreth, Shahed and Suh, Jina , title =. 2025 , isbn =. doi:10.1145/3715275.3732063 , booktitle =

    2025

  61. [75]

    2024 , isbn =

    Hutiri, Wiebke and Papakyriakopoulos, Orestis and Xiang, Alice , title =. 2024 , isbn =. doi:10.1145/3630106.3658911 , booktitle =

    2024

  62. [76]

    Symposium On Usable Privacy and Security , year=

    Youth-Centered GAI Risks (YAIR): A Taxonomy of Generative AI Risks from Empirical Data , author=. Symposium On Usable Privacy and Security , year=

  63. [77]

    2024 , isbn =

    Lee, Hao-Ping (Hank) and Yang, Yu-Ju and Von Davier, Thomas Serban and Forlizzi, Jodi and Das, Sauvik , title =. 2024 , isbn =. doi:10.1145/3613904.3642116 , booktitle =

    2024

  64. [78]

    2025 , eprint=

    Simulating Psychological Risks in Human-AI Interactions: Real-Case Informed Modeling of AI-Induced Addiction, Anorexia, Depression, Homicide, Psychosis, and Suicide , author=. 2025 , eprint=

    2025

  65. [79]

    European Journal of Marketing , author=

    More than just a chat: A taxonomy of consumers’ relationships with conversational AI agents and their well-being implications , volume=. European Journal of Marketing , author=. 2024 , month=feb, pages=. doi:10.1108/EJM-01-2023-0037 , number=

    2024 doi

  66. [80]

    2024 , eprint=

    User Privacy Harms and Risks in Conversational AI: A Proposed Framework , author=. 2024 , eprint=

    2024

  67. [81]

    and Shah, Darshi and Schwab-Reese, Laura M

    Zhou, Jiawei and Chen, Amy Z. and Shah, Darshi and Schwab-Reese, Laura M. and De Choudhury, Munmun , title =. 2025 , issue_date =. doi:10.1145/3757544 , journal =

    2025 doi

  68. [82]

    2026 , eprint=

    Understanding Risk and Dependency in AI Chatbot Use from User Discourse , author=. 2026 , eprint=

    2026

  69. [83]

    In the Wild

    Towards Trustworthy AI: Characterizing User-Reported Risks across LLMs "In the Wild" , author=. 2025 , eprint=

    2025

  70. [84]

    2025 , eprint=

    MinorBench: A hand-built benchmark for content-based risks for children , author=. 2025 , eprint=

    2025

  71. [85]

    Fairness in

    Binns, Reuben , year =. Fairness in. Proceedings of the 2018

    2018

  72. [86]

    Layers of

    Star, Susan Leigh and Strauss, Anselm , date =. Layers of. 1999 , journal =. doi:10.1023/A:1008651105359 , langid =

    1999 doi

  73. [87]

    , year =

    Bowker, Geoffrey C. , year =. Biodiversity. Social Studies of Science , shortjournal =. doi:10.1177/030631200030005001 , langid =

    doi

  74. [88]

    and Mayernik, Matthew S

    Edwards, Paul N. and Mayernik, Matthew S. and Batcheller, Archer L. and Bowker, Geoffrey C. and Borgman, Christine L. , year =. Science Friction:. Social Studies of Science , shortjournal =. doi:10.1177/0306312711413314 , langid =

    doi

  75. [89]

    , year =

    Ribes, David and Finholt, Thomas A. , year =. The. Journal for the Association of Information Systems (JAIS): Special issue on eInfrastructures , shortjournal =. 10822/557392 , eprinttype =

  76. [90]

    2014 , journal =

    Rethinking Repair , author =. 2014 , journal =

    2014

  77. [91]

    2002 , journal =

    Located Accountabilities in Technology Production , author =. 2002 , journal =

    2002

  78. [92]

    Translation Tutorial: 21 Fairness Definitions and Their Politics , shorttitle =. Proc. Conf. Fairness Accountability Transp., New York, Usa , author =. 2018 , volume =

    2018

  79. [93]

    2025 , eprint=

    Harmful Traits of AI Companions , author=. 2025 , eprint=

    2025

  80. [94]

    One size fits all?

    Braun, Virginia and Clarke, Victoria , year =. One size fits all?. Qualitative Research in Psychology , publisher =. doi:10.1080/14780887.2020.1769238 , issn =

    2020

  81. [95]

    2006 , journal =

    Using thematic analysis in psychology , author =. 2006 , journal =. doi:10.1191/1478088706qp063oa , issn =

    2006 doi

  82. [96]

    A worked example of

    Byrne, David , year =. A worked example of. Quality & Quantity , publisher =. doi:10.1007/s11135-021-01182-y , issn =

    doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.