Pith. sign in

REVIEW 3 major objections 6 minor 3 cited by

LLM alignment is a form of power, and decentralising it hinges on context, pluralism, and participation.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A position paper arguing that LLM alignment should be decentralised via context-sensitive pluralism and participation, illustrated through voter advice applications and game NPCs.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection A coherent, well-sourced position paper whose real contribution is the worked VAA/NPC contrast, and whose main empirical premise it openly admits is untested. the 3 major comments →

arxiv 2509.08858 v1 pith:Z7IMJY3I submitted 2025-09-09 cs.CY cs.LG

Decentralising LLM Alignment: A Case for Context, Pluralism, and Participation

Oriane Peter , Kate Devlin This is my paper
classification cs.CY cs.LG
keywords LLM alignmentdecentralisationpluralismparticipationcontext-of-useepistemic injusticerepresentational harmLoRA
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper argues that current LLM alignment is not a neutral technical fix but a form of power: it encodes the values of a narrow reference group and lets a few companies control which knowledge a model reproduces. To counter this, the authors propose that decentralising alignment hinges on three characteristics—context, pluralism, and participation—and that each must be tailored to the specific use case. They demonstrate the need for use-case specificity by contrasting a voter advice application, which benefits from centralised, consensus-seeking, consistency-focused RLHF alignment, with an LLM-powered game character, which calls for community-owned, decentralised, diversity-preserving LoRA-based alignment. If the paper is right, alignment becomes a lever for redistributing epistemic power rather than merely a safety knob, though the authors note it cannot substitute for broader societal change. The paper is a conceptual proposal, not an empirical demonstration; its practical force depends on whether communities can realistically train and control their own alignment models.

Core claim

Alignment is not a neutral safety knob: it encodes the values of a narrow reference group into every output, and institutions that control alignment control which knowledge an LLM reproduces. The paper argues that decentralising this power requires three characteristics—context, pluralism, and participation—each tailored to the specific use case. Two contrasting cases carry the argument: a voter advice application, where RLHF's 'mode collapse' is desirable for consistent, centralised, consensus-seeking presentation of political information, and an LLM-powered game character, where community-trained LoRA matrices, steerable pluralism, and decentralised licensing let communities control their

What carries the argument

Central to the paper is a three-part framework—context, pluralism, and participation—each given concrete form in two contrasting use cases: a Voter Advice Application (VAA) and an LLM-powered Non-Playable Character (NPC). Context dictates the algorithm: RLHF, whose 'mode collapse' narrows output diversity, suits the VAA's consistency; Low-Rank Adaptation (LoRA), a compact matrix of community preferences, suits the NPC's accessibility and diversity. Pluralism likewise splits: Overton pluralism (a spectrum of reasonable answers) for the VAA, steerable pluralism (responses indexed to a chosen attribute) for the NPC. Participation ranges from centralised, consensus-seeking oversight for the VAA

Load-bearing premise

The load-bearing premise is that communities can realistically train and control their own alignment models, such as LoRA matrices, once given sufficient infrastructure—a premise the authors explicitly flag as unproven when they note that not all communities have the resources, time, or technical expertise to train a LoRA matrix.

What would settle it

A concrete falsifying test would be a field study in which two under-resourced communities are given LoRA training infrastructure and asked to produce acceptable character representations; if they cannot produce usable matrices, or if the resulting representations are not preferred over studio-designed ones, the paper's decentralisation mechanism collapses.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Voter advice applications should be aligned with RLHF-style consistency, Overton pluralism, and centralised, consensus-seeking participation, rather than generic commercial chatbots.
  • Community representation in games can shift to the communities themselves through training and licensing LoRA matrices, with the power to revoke licences if a storyline is deemed insensitive.
  • Pluralism must be designed per context: Overton (spectrum) for decision-support systems, steerable (attribute-indexed) for character-driven systems; distributional pluralism is unsuitable for both.
  • Decentralising alignment requires public investment in infrastructure and training so that communities without resources can actually participate.
  • Alignment reform alone will not achieve epistemic justice; broader societal changes remain necessary.

Where Pith is reading between the lines

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

  • The same LoRA-licensing model could extend beyond games to other media and cultural products where communities are represented, such as film, animation, or virtual heritage—a direction the paper does not discuss.
  • The framework suggests a concrete evaluation metric: measuring whether community-trained alignment matrices reduce representational harms compared with studio-designed baselines, and whether communities perceive control over their representations.
  • The paper's context-specificity argument implies that generic alignment benchmarks and universal 'safe AI' certifications may be epistemically misleading; evaluations should be tied to declared contexts of use.
  • A testable extension would be to let two communities train LoRA matrices for the same character and measure whether the divergent outputs are acceptable to both groups and to the studio, probing the limits of 'no inter-community consensus requirement'.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. This paper argues that LLM alignment is not a value-neutral technical fix but a power/knowledge mechanism that concentrates epistemic control in a few large companies. It proposes that decentralising alignment requires three characteristics—context, pluralism, and participation—and uses two deliberately contrasting fictional use cases, a Voter Advice Application (VAA) and an LLM-powered Non-Playable Character (NPC), to show that these principles must be adapted to context. For VAA, the paper recommends RLHF-style consistency, Overton pluralism, and centrally organised participation; for NPCs, it recommends LoRA-based community-owned adapters, steerable pluralism, and fully decentralised, community-led licensing. The paper explicitly frames alignment as a potential site of resistance against epistemic injustice while acknowledging that these strategies do not substitute for broader societal change and that empirical testing is needed.

Significance. If accepted as a position paper, the work is valuable: it moves beyond the usual call for more inclusive datasets to a structural argument about who controls alignment, and it demonstrates that abstract pluralism/participation principles have contradictory implications in different deployment contexts. The authors are unusually honest: they identify their own empirical gaps, call for rigorous testing, and acknowledge financing/organisational limits. The paper also offers concrete, falsifiable proposals (e.g., licensable LoRA representations, use of mode collapse in VAA, centralised vs decentralised participation), which is a strength. However, the practical force of the central claim depends on empirical feasibility facts that the paper does not establish; those are the load-bearing points of my review.

major comments (3)
  1. [LLM-Powered NPCs / Contextual Alignment / Participatory Alignment] The concrete decentralisation mechanism for the NPC use case rests on community-trained LoRA matrices. The text says 'This concept can be realised through SFT combined with LoRA' and that 'diverse communities can be empowered to create personalised representations of themselves,' but later concedes 'Rigorous testing is essential to determine whether diverse communities can satisfactorily use LoRA training...' and that questions of organisation and financing 'remain.' This is not a local caveat: if non-expert communities cannot produce adapters that shift the base model's outputs enough to constitute representational control, or can only do so by relying on expert intermediaries, the NPC model becomes a licensing arrangement in which the base developer retains de facto control rather than a transfer of epistemic power. Please reframe these sections as an open research hypothesis with expl
  2. [Table 1 / Consistent VAA] Table 1 classifies VAA contextual alignment as 'RLHF for consistency and reliability,' but the body text says 'iterative investigation is required to determine whether RLHF can provide sufficient guarantees of consistency and accuracy in such a sensitive use case.' The table overstates what the text supports. Because the use-case contrast is central to the paper's contribution, the table's entries should match the epistemic status of the text (e.g., 'RLHF as a candidate method, pending evaluation'), or the text's hedging should be reduced.
  3. [Overton Pluralistic VAA / Participatory Alignment] The VAA proposal assigns curation of the 'source of truth' to 'democratic institutions, trusted third parties, or a consortium of political parties' and allows a centralised authority to resolve disagreements when consensus fails. This is a shift of control away from private companies, but it is not decentralisation in the same sense used for the NPC case. If 'decentralisation' is intended to have a common meaning across both use cases, the paper is equivocating; if it is intentionally context-dependent, that should be stated explicitly and justified. Otherwise the reader cannot tell why the same term covers both community ownership and state-level centralisation.
minor comments (6)
  1. [Abstract] The abstract says the paper 'demonstrates' the importance of context and 'demonstrates nuanced requirements.' Since the support is argumentative and illustrative rather than empirical, 'argues' and 'illustrates' would be more accurate.
  2. [Throughout] The manuscript alternates between 'contextualization' and 'context'. Standardise the terminology, since 'context' is one of the three named characteristics.
  3. [Table 1] There are inconsistent spacing artifacts such as 'V oter Advice Applications' in the table and body text; these should be cleaned up.
  4. [Pluralistic Alignment] Distributional pluralism is mentioned and dismissed in a single sentence. Give a one-sentence definition so the reader can follow why it is unsuitable for VAAs and NPCs.
  5. [Participatory Alignment] The Māori licensing example is cited via Birhane et al.; providing the primary source would strengthen the claim that community-owned data licensing is a workable precedent.
  6. [Consistent VAA] The phrase 'generalisation should be prioritised over creativity' is ambiguous: it is not clear whether 'generalisation' means consistency across users or breadth across contexts. Clarify to avoid confusion with the earlier discussion of RLHF's mode collapse.

Circularity Check

0 steps flagged

No circularity: the argument is conceptual and externally sourced; the proposal is explicitly flagged as untested rather than assumed.

full rationale

This is a position/argument paper, not an empirical derivation. The central claim that decentralised LLM alignment needs context, pluralism, and participation is advanced as a normative framework, not as a prediction derived from fitted parameters or from the paper's own prior results. The paper does not define these characteristics in terms of the conclusion it wants to reach; it argues for them using external sources (Ouyang et al., Sorensen et al., Delgado et al., Birhane et al., Varshney, etc.) and grounds them in two illustrative use cases. The LoRA-based community-alignment proposal is explicitly borrowed from Varshney (2024) and is accompanied by candid caveats: 'Rigorous testing is essential to determine whether diverse communities can satisfactorily use LoRA training...' and 'questions around how these efforts are organised in practice and how they are financed remain.' These are acknowledged limitations, not hidden inputs to the argument. There are no equations, no fitted values, no uniqueness theorems, and no load-bearing self-citations (the authors do not cite their own prior work as evidence). Even the analytical move that alignment centralises power because developers control outputs is a stated theoretical framing, not a circular derivation: it identifies a mechanism rather than assuming the conclusion. Therefore the paper is self-contained as a conceptual contribution and receives a score of 0.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 0 invented entities

The central argument rests on five assumptions: three domain assumptions drawn from the cited literature (narrowness of alignment, mode collapse, applicability of power/knowledge theory), and two ad hoc assumptions specific to the paper's own proposal (feasibility of community LoRA training and representativeness of the two use cases). No free parameters are fitted since the paper makes no numerical predictions. No invented entities are introduced.

axioms (5)
  • domain assumption Alignment necessarily mirrors a narrow reference group and cannot be universal.
    Foundational to the power-centralisation claim; based on Ouyang et al. 2022 quote that alignment always targets a specific group. Introduction, paragraph 3.
  • domain assumption RLHF causes mode collapse, reducing output diversity compared to non-aligned models.
    Used to justify RLHF for VAA and LoRA for NPC. Cited to Kirk et al. 2024c, Sorensen et al. 2024, O'Mahony et al. 2024; not independently verified. Section 'Contextual Alignment'.
  • domain assumption Foucault's and Jasanoff's analyses of power/knowledge apply to LLM alignment as a knowledge-dissemination technology.
    The paper's theoretical lens; assumed rather than argued. Section 'LLM Alignment and Power Centralisation'.
  • ad hoc to paper Community-based actors can feasibly train, control, and license LoRA matrices for representation.
    The NPC proposal depends on this. The authors acknowledge resource and expertise barriers and call for 'rigorous testing'. Section 'Community-led NPCs' and 'Contextual Alignment'.
  • ad hoc to paper Two fictional use cases (VAA and NPC) provide a sufficient basis for claims about use-case specificity.
    The cases are 'deliberately chosen for their stark contrast', with no representativeness argument. Section 'Use Cases'.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Decentralising LLM Alignment: A Case for Context, Pluralism, and Participation." pith.science (2026). https://pith.science/paper/Z7IMJY3I

@misc{pith2026250908858,
  author       = {Pith},
  title        = {Pith review of: Decentralising LLM Alignment: A Case for Context, Pluralism, and Participation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z7IMJY3I}},
  note         = {Machine review of arXiv:2509.08858}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Large Language Models (LLMs) alignment methods have been credited with the commercial success of products like ChatGPT, given their role in steering LLMs towards user-friendly outputs. However, current alignment techniques predominantly mirror the normative preferences of a narrow reference group, effectively imposing their values on a wide user base. Drawing on theories of the power/knowledge nexus, this work argues that current alignment practices centralise control over knowledge production and governance within already influential institutions. To counter this, we propose decentralising alignment through three characteristics: context, pluralism, and participation. Furthermore, this paper demonstrates the critical importance of delineating the context-of-use when shaping alignment practices by grounding each of these features in concrete use cases. This work makes the following contributions: (1) highlighting the role of context, pluralism, and participation in decentralising alignment; (2) providing concrete examples to illustrate these strategies; and (3) demonstrating the nuanced requirements associated with applying alignment across different contexts of use. Ultimately, this paper positions LLM alignment as a potential site of resistance against epistemic injustice and the erosion of democratic processes, while acknowledging that these strategies alone cannot substitute for broader societal changes.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Positive Alignment: Artificial Intelligence for Human Flourishing

    cs.AI 2026-05 unverdicted novelty 6.0

    Positive Alignment introduces AI systems that support human flourishing pluralistically and proactively while remaining safe, as a necessary complement to traditional safety-focused alignment research.

  2. Positive Alignment: Artificial Intelligence for Human Flourishing

    cs.AI 2026-05 unverdicted novelty 5.0

    Positive Alignment is defined as AI systems that support human flourishing pluralistically while staying safe and cooperative, presented as a necessary complement to existing safety-focused alignment research.

  3. Positive Alignment: Artificial Intelligence for Human Flourishing

    cs.AI 2026-05 unverdicted novelty 4.0

    Positive Alignment is introduced as a distinct AI agenda that supports human flourishing through pluralistic and context-sensitive design, complementing traditional safety-focused alignment.

Reference graph

Works this paper leans on

81 extracted references · 77 canonical work pages · cited by 1 Pith paper

  1. [1]

    Aakanksha ; Ahmadian, A.; Ermis, B.; Goldfarb-Tarrant , S.; Kreutzer, J.; Fadaee, M.; and Hooker, S. 2024. The Multilingual Alignment Prism : Aligning Global and Local Preferences to Reduce Harm . In Al-Onaizan , Y.; Bansal, M.; and Chen, Y.-N., eds., Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , 12027--12049. Mi...

    2024

  2. [2]

    Anthropic. 2023. Collective Constitutional AI : Aligning a Language Model with Public Input

    work page 2023

  3. [3]

    Anthropic . 2024. The Claude 3 Model Family : Opus , Sonnet , Haiku . Technical report, Anthropic

    work page 2024

  4. [4]

    J.; Park, P

    Atari, M.; Xue, M. J.; Park, P. S.; Blasi, D.; and Henrich, J. 2023. Which Humans ?

    work page 2023

  5. [5]

    E.; Fort, S.; Lanham, T.; Telleen-Lawton , T.; Conerly, T.; Henighan, T.; Hume, T.; Bowman, S

    Bai, Y.; Kadavath, S.; Kundu, S.; Askell, A.; Kernion, J.; Jones, A.; Chen, A.; Goldie, A.; Mirhoseini, A.; McKinnon, C.; Chen, C.; Olsson, C.; Olah, C.; Hernandez, D.; Drain, D.; Ganguli, D.; Li, D.; Tran-Johnson , E.; Perez, E.; Kerr, J.; Mueller, J.; Ladish, J.; Landau, J.; Ndousse, K.; Lukosuite, K.; Lovitt, L.; Sellitto, M.; Elhage, N.; Schiefer, N.;...

    Pith/arXiv arXiv 2022

  6. [6]

    Bergman, S.; Marchal, N.; Mellor, J.; Mohamed, S.; Gabriel, I.; and Isaac, W. 2024. STELA : A Community-Centred Approach to Norm Elicitation for AI Alignment. Scientific Reports, 14(1): 6616

    work page 2024

  7. [7]

    C.; Gabriel, I.; and Mohamed, S

    Birhane, A.; Isaac, W.; Prabhakaran, V.; Diaz, M.; Elish, M. C.; Gabriel, I.; and Mohamed, S. 2022. Power to the People ? Opportunities and Challenges for Participatory AI . In Equity and Access in Algorithms , Mechanisms , and Optimization , 1--8. Arlington VA USA: ACM. ISBN 978-1-4503-9477-2

    work page 2022

  8. [8]

    L.; Barocas, S.; Daum \'e III, H.; and Wallach, H

    Blodgett, S. L.; Barocas, S.; Daum \'e III, H.; and Wallach, H. 2020. Language ( Technology ) Is Power : A Critical Survey of `` Bias '' in NLP . In Jurafsky, D.; Chai, J.; Schluter, N.; and Tetreault, J., eds., Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , 5454--5476. Online: Association for Computational Linguistics

    work page 2020

  9. [9]

    A.; Kumar, A.; Jurafsky, D.; and Liang, P

    Bommasani, R.; Creel, K. A.; Kumar, A.; Jurafsky, D.; and Liang, P. 2022. Picking on the Same Person: Does Algorithmic Monoculture Lead to Outcome Homogenization? In Proceedings of the 36th International Conference on Neural Information Processing Systems , NIPS '22, 3663--3678. Red Hook, NY, USA: Curran Associates Inc. ISBN 978-1-7138-7108-8

    work page 2022

  10. [10]

    K.; Scheurer, J.; Rando, J.; Freedman, R.; Korbak, T.; Lindner, D.; Freire, P.; Wang, T

    Casper, S.; Davies, X.; Shi, C.; Gilbert, T. K.; Scheurer, J.; Rando, J.; Freedman, R.; Korbak, T.; Lindner, D.; Freire, P.; Wang, T. T.; Marks, S.; Segerie, C.-R.; Carroll, M.; Peng, A.; Christoffersen, P.; Damani, M.; Slocum, S.; Anwar, U.; Siththaranjan, A.; Nadeau, M.; Michaud, E. J.; Pfau, J.; Krasheninnikov, D.; Chen, X.; Langosco, L.; Hase, P.; Biy...

    work page 2023

  11. [11]

    Cave, S.; and Dihal, K. 2020. The Whiteness of AI . Philosophy & Technology, 33(4): 685--703

    work page 2020

  12. [12]

    Chen, Z.; Cano, A. H.; Romanou, A.; Bonnet, A.; Matoba, K.; Salvi, F.; Pagliardini, M.; Fan, S.; K \"o pf, A.; Mohtashami, A.; Sallinen, A.; Sakhaeirad, A.; Swamy, V.; Krawczuk, I.; Bayazit, D.; Marmet, A.; Montariol, S.; Hartley, M.-A.; Jaggi, M.; and Bosselut, A. 2023. MEDITRON-70B : Scaling Medical Pretraining for Large Language Models . arXiv:2311.16079

    Pith/arXiv arXiv 2023

  13. [13]

    N.; Li, T.; Li, D.; Zhu, B.; Zhang, H.; Jordan, M

    Chiang, W.-L.; Zheng, L.; Sheng, Y.; Angelopoulos, A. N.; Li, T.; Li, D.; Zhu, B.; Zhang, H.; Jordan, M. I.; Gonzalez, J. E.; and Stoica, I. 2024. Chatbot arena: an open platform for evaluating LLMs by human preference. In Proceedings of the 41st International Conference on Machine Learning , volume 235 of ICML '24 , 8359--8388. Vienna, Austria: JMLR.org

    work page 2024

  14. [14]

    Coeckelbergh, M. 2025. LLMs , Truth , and Democracy : An Overview of Risks . Science and Engineering Ethics, 31(1): 4

    work page 2025

  15. [15]

    R.; and Ooi, W

    Cox, S. R.; and Ooi, W. T. 2024. Conversational Interactions with NPCs in LLM-Driven Gaming : Guidelines from a Content Analysis of Player Feedback . In F lstad, A.; Araujo, T.; Papadopoulos, S.; Law, E. L.-C.; Luger, E.; Goodwin, M.; Hobert, S.; and Brandtzaeg, P. B., eds., Chatbot Research and Design , 167--184. Cham: Springer Nature Switzerland. ISBN 9...

    work page 2024

  16. [16]

    Crenshaw, K. 1998. Demarginalizing the Intersection of Race and Sex : A Black Feminist Critique of Antidiscrimination Doctrine , Feminist Theory , and Antiracist Politics . In Phillips, A., ed., Feminism And Politics : Oxford Readings In Feminism . Oxford University Press. ISBN 978-0-19-878206-3

    work page 1998

  17. [17]

    Delgado, F.; Yang, S.; Madaio, M.; and Yang, Q. 2023. The Participatory Turn in AI Design : Theoretical Foundations and the Current State of Practice . In Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms , Mechanisms , and Optimization , EAAMO '23, 1--23. New York, NY, USA: Association for Computing Machinery. ISBN 979-8-4007-0381-2

    work page 2023

  18. [18]

    D.; Rosen, R.; Baker, D

    Diaz, M.; Kivlichan, I. D.; Rosen, R.; Baker, D. K.; Amironesei, R.; Prabhakaran, V.; and Denton, E. 2022. CrowdWorkSheets : Accounting for Individual and Collective Identities Underlying Crowdsourced Dataset Annotation . In 2022 ACM Conference on Fairness Accountability and Transparency , 2342--2351

    work page 2022

  19. [19]

    Durmus, E.; Nyugen, K.; Liao, T.; Schiefer, N.; Askell, A.; Bakhtin, A.; Chen, C.; Hatfield-Dodds , Z.; Hernandez, D.; Joseph, N.; Lovitt, L.; McCandlish, S.; Sikder, O.; Tamkin, A.; Thamkul, J.; Kaplan, J.; Clark, J.; and Ganguli, D. 2023. Towards Measuring the Representation of Subjective Global Opinions in Language Models . ArXiv

    work page 2023

  20. [20]

    Foucault, M. 1975. Surveiller et Punir : Naissance de Ia Prison . Editions Gallimard,Paris

    work page 1975

  21. [21]

    Gabriel, I. 2020. Artificial Intelligence , Values , and Alignment . Minds and Machines, 30(3): 411--437

    work page 2020

  22. [22]

    Gallotta, R.; Todd, G.; Zammit, M.; Earle, S.; Liapis, A.; Togelius, J.; and Yannakakis, G. N. 2024. Large Language Models and Games : A Survey and Roadmap

    work page 2024

  23. [23]

    I.; and Moss, E

    Ganesh, M. I.; and Moss, E. 2022. Resistance and Refusal to Algorithmic Harms: Varieties of `Knowledge Projects'. Media International Australia, 183(1): 90--106

    work page 2022

  24. [24]

    Gebru, T.; and Torres, \'E . P. 2024. The TESCREAL Bundle: Eugenics and the Promise of Utopia through Artificial General Intelligence. First Monday

    work page 2024

  25. [25]

    Germann, M.; Mendez, F.; and Gemenis, K. 2023. Do Voting Advice Applications Affect Party Preferences ? Evidence from Field Experiments in Five European Countries . Political Communication, 40(5): 596--614

    work page 2023

  26. [26]

    Gillespie, T. 2024. Generative AI and the Politics of Visibility. Big Data & Society, 11(2): 20539517241252131

    work page 2024

  27. [27]

    L.; and Suri, S

    Gray, M. L.; and Suri, S. 2019. Ghost Work : How to Stop Silicon Valley from Building a New Global Underclass . Boston New York NY: Harper Business, illustrated edition edition. ISBN 978-1-328-56624-9

    work page 2019

  28. [28]

    Heaven, W. D. 2023. The inside Story of How ChatGPT Was Built from the People Who Made It. MIT Technology Review

    work page 2023

  29. [29]

    u ller, A.; Spielkamp, M.; Schiller, A. L.; Kesler, W.; Omalar, M.; Th \

    Helming, C.; M \"u ller, A.; Spielkamp, M.; Schiller, A. L.; Kesler, W.; Omalar, M.; Th \"u mmler, M.; Zimmermann, M.; Sanchez, I.; Kimel, A.; Pannatier, E.; Urech, T.; Sorie, D.; Loi, M.; Felder, A.; Romano, S.; Kerby, N.; Angius, R.; Robutti, S.; Schueler, M.; Faddoul, M.; and C etin, R. B. 2023. Generative AI and Elections: Are Chatbots a Reliable Sour...

    work page 2023

  30. [30]

    J.; and Norenzayan, A

    Henrich, J.; Heine, S. J.; and Norenzayan, A. 2010. The Weirdest People in the World? Behavioral and Brain Sciences, 33(2-3): 61--83

    work page 2010

  31. [31]

    J.; Shen, Y.; Wallis, P.; Allen-Zhu , Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu , Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2021. LoRA : Low-Rank Adaptation of Large Language Models . In International Conference on Learning Representations

    work page 2021

  32. [32]

    I.; Durmus, E.; Tamkin, A.; and Ganguli, D

    Huang, S.; Siddarth, D.; Lovitt, L.; Liao, T. I.; Durmus, E.; Tamkin, A.; and Ganguli, D. 2024. Collective Constitutional AI : Aligning a Language Model with Public Input . In Proceedings of the 2024 ACM Conference on Fairness , Accountability , and Transparency , FAccT '24, 1395--1417. New York, NY, USA: Association for Computing Machinery. ISBN 979-8-40...

    work page 2024

  33. [33]

    Jakesch, M.; Bhat, A.; Buschek, D.; Zalmanson, L.; and Naaman, M. 2023. Co- Writing with Opinionated Language Models Affects Users ' Views . In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems , CHI '23, 1--15. New York, NY, USA: Association for Computing Machinery. ISBN 978-1-4503-9421-5

    work page 2023

  34. [34]

    Jasanoff, S. 2006. States of Knowledge : The Co-Production of Science and the Social Order . London: Routledge, 1st edition edition. ISBN 978-0-415-40329-0

    work page 2006

  35. [35]

    F.; and Kersting, K

    Jentzsch, S. F.; and Kersting, K. 2023. ChatGPT Is Fun, but It Is Not Funny! Humor Is Still Challenging Large Language Models . In 61st Annual Meeting of the Association for Computational Linguistics , ACL 2023 . Toronto, Canada. ISBN 978-1-959429-87-6

    work page 2023

  36. [36]

    Jha, P.; Jain, R.; Mandal, K.; Chadha, A.; Saha, S.; and Bhattacharyya, P. 2024. MemeGuard : An LLM and VLM-based Framework for Advancing Content Moderation via Meme Intervention . In Ku, L.-W.; Martins, A.; and Srikumar, V., eds., Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics ( Volume 1: Long Papers ) , 8084--810...

    work page 2024

  37. [37]

    Kamoen, N.; McCartan, T.; and Liebrecht, C. 2022. Conversational Agent Voting Advice Applications : A Comparison Between a Structured , Semi-structured , and Non-structured Chatbot Design for Communicating with Voters About Political Issues . In F lstad, A.; Araujo, T.; Papadopoulos, S.; Law, E. L.-C.; Luger, E.; Goodwin, M.; and Brandtzaeg, P. B., eds., ...

    work page 2022

  38. [38]

    L.; Laird, K.; Wallach, H.; and Barocas, S

    Katzman, J.; Wang, A.; Scheuerman, M.; Blodgett, S. L.; Laird, K.; Wallach, H.; and Barocas, S. 2023. Taxonomizing and Measuring Representational Harms: A Look at Image Tagging. In Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Sy...

    work page 2023

  39. [39]

    Kirk, H.; Vidgen, B.; Rottger, P.; and Hale, S. 2023. The Empty Signifier Problem : Towards Clearer Paradigms for Operationalising " Alignment '' in Large Language Models . In Socially Responsible Language Modelling Research

    work page 2023

  40. [40]

    R.; Vidgen, B.; R \"o ttger, P.; and Hale, S

    Kirk, H. R.; Vidgen, B.; R \"o ttger, P.; and Hale, S. A. 2024 a . The Benefits, Risks and Bounds of Personalizing the Alignment of Large Language Models to Individuals. Nature Machine Intelligence, 6(4): 383--392

    work page 2024

  41. [41]

    R.; Whitefield, A.; R \"o ttger, P.; Bean, A

    Kirk, H. R.; Whitefield, A.; R \"o ttger, P.; Bean, A. M.; Margatina, K.; Mosquera, R.; Ciro, J. M.; Bartolo, M.; Williams, A.; He, H.; Vidgen, B.; and Hale, S. A. 2024 b . The PRISM Alignment Dataset : What Participatory , Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models . ...

    work page 2024

  42. [42]

    Kirk, R.; Mediratta, I.; Nalmpantis, C.; Luketina, J.; Hambro, E.; Grefenstette, E.; and Raileanu, R. 2024 c . Understanding the Effects of RLHF on LLM Generalisation and Diversity . The Twelfth International Conference on Learning Representations

    work page 2024

  43. [43]

    Kniaz, R. 2023. The Incoming Tidal Wave Of Data Pollution In AI . Forbes

    work page 2023

  44. [44]

    Kuo, T.-S.; Halfaker, A.; Cheng, Z.; Kim, J.; Wu, M.-H.; Wu, T.; Holstein, K.; and Zhu, H. 2024. Wikibench: Community-Driven Data Curation for AI Evaluation on Wikipedia . In Proceedings of the CHI Conference on Human Factors in Computing Systems , 1--24

    work page 2024

  45. [45]

    Latour, B. 1988. Science in Action : How to Follow Scientists and Engineers Through Society . Cambridge (Mass.): Harvard University Press, revised ed. edition edition. ISBN 978-0-674-79291-3

    work page 1988

  46. [46]

    Le Ludec, C.; Cornet, M.; and Casilli, A. A. 2023. The Problem with Annotation. Human Labour and Outsourcing between France and Madagascar . Big Data & Society, 10(2): 20539517231188723

    work page 2023

  47. [47]

    Lorde, A. 2007. Sister Outsider: Essays and Speeches. [ The Crossing Press Feminist Series]. Berkeley [California: Crossing Press, revised edition. edition. ISBN 978-1-58091-186-3

    work page 2007

  48. [48]

    Mamykina, L.; Manoim, B.; Mittal, M.; Hripcsak, G.; and Hartmann, B. 2011. Design Lessons from the Fastest Q&a Site in the West. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems , CHI '11, 2857--2866. New York, NY, USA: Association for Computing Machinery. ISBN 978-1-4503-0228-9

    work page 2011

  49. [49]

    Miceli, M.; and Posada, J. 2022. The Data-Production Dispositif . Proceedings of the ACM on Human-Computer Interaction, 6(CSCW2): 460:1--460:37

    work page 2022

  50. [50]

    Miceli, M.; Schuessler, M.; and Yang, T. 2020. Between Subjectivity and Imposition : Power Dynamics in Data Annotation for Computer Vision . Proceedings of the ACM on Human-Computer Interaction, 4(CSCW2): 115:1--115:25

    work page 2020

  51. [51]

    Z.; Holtermann, C.; Kaffee, L.-A.; Laud, T.; Lauscher, A.; Lopez-Davila , R

    Mitchell, M.; Attanasio, G.; Baldini, I.; Clinciu, M.; Clive, J.; Delobelle, P.; Dey, M.; Hamilton, S.; Dill, T.; Doughman, J.; Dutt, R.; Ghosh, A.; Forde, J. Z.; Holtermann, C.; Kaffee, L.-A.; Laud, T.; Lauscher, A.; Lopez-Davila , R. L.; Masoud, M.; Nangia, N.; Ovalle, A.; Pistilli, G.; Radev, D.; Savoldi, B.; Raheja, V.; Qin, J.; Ploeger, E.; Subramoni...

    work page 2025

  52. [52]

    Mittelstadt, B. 2019. Principles Alone Cannot Guarantee Ethical AI . Nature Machine Intelligence, 1(11): 501--507

    work page 2019

  53. [53]

    Mohamed, S.; Png, M.-T.; and Isaac, W. 2020. Decolonial AI : Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence . Philosophy & Technology, 33(4): 659--684

    work page 2020

  54. [54]

    Muldoon, J.; Cant, C.; Graham, M.; and Ustek Spilda, F. 2023. The Poverty of Ethical AI : Impact Sourcing and AI Supply Chains. AI & SOCIETY

    work page 2023

  55. [55]

    Muldoon, J.; and Wu, B. A. 2023. Artificial Intelligence in the Colonial Matrix of Power . Philosophy & Technology, 36(4): 80

    work page 2023

  56. [56]

    Munzert, S.; and Ramirez-Ruiz , S. 2021. Meta- Analysis of the Effects of Voting Advice Applications . Political Communication

    work page 2021

  57. [57]

    O'Mahony, L.; Grinsztajn, L.; Schoelkopf, H.; and Biderman, S. 2024. Attributing Mode Collapse in the Fine-Tuning of Large Language Models . In ICLR 2024 Workshop on Mathematical and Empirical Understanding of Foundation Models

    work page 2024

  58. [58]

    Ouyang, L.; Wu, J.; Jiang, X.; Almeida, D.; Wainwright, C. L.; Mishkin, P.; Zhang, C.; Agarwal, S.; Slama, K.; Ray, A.; Schulman, J.; Hilton, J.; Kelton, F.; Miller, L.; Simens, M.; Askell, A.; Welinder, P.; Christiano, P.; Leike, J.; and Lowe, R. 2022. Training Language Models to Follow Instructions with Human Feedback. In Proceedings of the 36th Interna...

    work page 2022

  59. [59]

    Padhi, I.; Natesan Ramamurthy, K.; Sattigeri, P.; Nagireddy, M.; Dognin, P.; and Varshney, K. R. 2024. Value Alignment from Unstructured Text . In Dernoncourt, F.; Preo t iuc-Pietro , D.; and Shimorina, A., eds., Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing : Industry Track , 1083--1095. Miami, Florida, US: Associ...

    work page 2024

  60. [60]

    Politools. 2019. Methodenbeschreibung -- Smartvote Wahlempfehlung . Technical report, Politools

    work page 2019

  61. [61]

    Pols, A.; and Spahn, A. 2015. Design for the Values of DemocracyDemocracyand JusticeJustice . In van den Hoven , J.; Vermaas, P. E.; and van de Poel , I., eds., Handbook of Ethics , Values , and Technological Design : Sources , Theory , Values and Application Domains , 335--363. Dordrecht: Springer Netherlands. ISBN 978-94-007-6970-0

    work page 2015

  62. [62]

    Thick Evaluations

    Qadri, R.; Diaz, M.; Wang, D.; and Madaio, M. 2025. The Case for " Thick Evaluations " of Cultural Representation in AI . arXiv:2503.19075

    arXiv 2025

  63. [63]

    M.; Hanna, A.; and Paullada, A

    Raji, D.; Denton, E.; Bender, E. M.; Hanna, A.; and Paullada, A. 2021. AI and the Everything in the Whole Wide World Benchmark . Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks, 1

    work page 2021

  64. [64]

    Rose, K.; Kroner Dale, M.; and Jusko, K. 2025. From Community Input to AI Alignment : Incorporating Global Perspectives in AI Development . In Participatory AI Research & Practice Symposium . Paris

    work page 2025

  65. [65]

    V.; and Xiao, Z

    Sharma, N.; Liao, Q. V.; and Xiao, Z. 2024. Generative Echo Chamber ? Effect of LLM-Powered Search Systems on Diverse Information Seeking . In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems , CHI '24, 1--17. New York, NY, USA: Association for Computing Machinery. ISBN 979-8-4007-0330-0

    work page 2024

  66. [66]

    Sloane, M.; Moss, E.; Awomolo, O.; and Forlano, L. 2022. Participation Is Not a Design Fix for Machine Learning . In Proceedings of the 2nd ACM Conference on Equity and Access in Algorithms , Mechanisms , and Optimization , EAAMO '22, 1--6. New York, NY, USA: Association for Computing Machinery. ISBN 978-1-4503-9477-2

    work page 2022

  67. [67]

    L.; Mireshghallah, N.; Rytting, C

    Sorensen, T.; Moore, J.; Fisher, J.; Gordon, M. L.; Mireshghallah, N.; Rytting, C. M.; Ye, A.; Jiang, L.; Lu, X.; Dziri, N.; Althoff, T.; and Choi, Y. 2024. Position: A Roadmap to Pluralistic Alignment . In Proceedings of the 41st International Conference on Machine Learning , 46280--46302. PMLR

    work page 2024

  68. [68]

    Stadelmann-Steffen , I.; Rajski, H.; and Ruprecht, S. 2023. The Role of Vote Advice Application in Direct-Democratic Opinion Formation: An Experiment from Switzerland . Acta Politica, 58(4): 792--818

    work page 2023

  69. [69]

    Stockinger, E.; Maas, J.; Talvitie, C.; and Dignum, V. 2024. Trustworthiness of voting advice applications in Europe . Ethics and Information Technology, 26(3): 55

    work page 2024

  70. [70]

    Suchman, L.; and Suchman, L. A. 2007. Human- Machine Reconfigurations : Plans and Situated Actions . Cambridge University Press. ISBN 978-0-521-67588-8

    work page 2007

  71. [71]

    L.; Bhargava, R.; Cruxen, I.; Cuba, A

    Suresh, H.; Movva, R.; Dogan, A. L.; Bhargava, R.; Cruxen, I.; Cuba, A. M.; Taurino, G.; So, W.; and D'Ignazio, C. 2022. Towards Intersectional Feminist and Participatory ML : A Case Study in Supporting Feminicide Counterdata Collection . In Proceedings of the 2022 ACM Conference on Fairness , Accountability , and Transparency , FAccT '22, 667--678. New Y...

    work page 2022

  72. [72]

    Suresh, H.; Tseng, E.; Young, M.; Gray, M.; Pierson, E.; and Levy, K. 2024. Participation in the age of foundation models. In Proceedings of the 2024 ACM Conference on Fairness , Accountability , and Transparency , FAccT '24, 1609--1621. New York, NY, USA: Association for Computing Machinery. ISBN 979-8-4007-0450-5

    work page 2024

  73. [73]

    Varshney, K. R. 2024. Decolonial AI Alignment : Openness , Visesa-Dharma , and Including Excluded Knowledges . In Proceedings of the AAAI / ACM Conference on AI , Ethics , and Society , volume 7, 1467--1481

    work page 2024

  74. [74]

    Victor, D. 2016. Microsoft Created a Twitter Bot to Learn From Users . It Quickly Became a Racist Jerk . The New York Times

    work page 2016

  75. [75]

    Wallis, D. 2023. Appropriation or Erasure ? Imagining Indigenous Futures in Games . Journal of Games Criticism, Volume 5(Bonus Issue A)

    work page 2023

  76. [76]

    Wang, D.; Prabhat, S.; and Sambasivan, N. 2022. Whose AI Dream ? In Search of the Aspiration in Data Annotation. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems , CHI '22, 1--16. New York, NY, USA: Association for Computing Machinery. ISBN 978-1-4503-9157-3

    work page 2022

  77. [77]

    A.; Isaac, W.; Legassick, S.; Irving, G.; and Gabriel, I

    Weidinger, L.; Mellor, J.; Rauh, M.; Griffin, C.; Uesato, J.; Huang, P.-S.; Cheng, M.; Glaese, M.; Balle, B.; Kasirzadeh, A.; Kenton, Z.; Brown, S.; Hawkins, W.; Stepleton, T.; Biles, C.; Birhane, A.; Haas, J.; Rimell, L.; Hendricks, L. A.; Isaac, W.; Legassick, S.; Irving, G.; and Gabriel, I. 2021. Ethical and Social Risks of Harm from Language Models

    work page 2021

  78. [78]

    A.; Rimell, L.; Isaac, W.; Haas, J.; Legassick, S.; Irving, G.; and Gabriel, I

    Weidinger, L.; Uesato, J.; Rauh, M.; Griffin, C.; Huang, P.-S.; Mellor, J.; Glaese, A.; Cheng, M.; Balle, B.; Kasirzadeh, A.; Biles, C.; Brown, S.; Kenton, Z.; Hawkins, W.; Stepleton, T.; Birhane, A.; Hendricks, L. A.; Rimell, L.; Isaac, W.; Haas, J.; Legassick, S.; Irving, G.; and Gabriel, I. 2022. Taxonomy of Risks Posed by Language Models . In Proceedi...

    work page 2022

  79. [79]

    C.; Dailisan, D.; Korecki, M.; Hausladen, C

    Yang, J. C.; Dailisan, D.; Korecki, M.; Hausladen, C. I.; and Helbing, D. 2025. LLM Voting : Human Choices and AI Collective Decision - Making . In Proceedings of the 2024 AAAI / ACM Conference on AI , Ethics , and Society , AIES '24, 1696--1708. San Jose, California, USA: AAAI Press

    work page 2025

  80. [80]

    Young, M.; Ehsan, U.; Singh, R.; Tafesse, E.; Gilman, M.; Harrington, C.; and Metcalf, J. 2024. Participation versus Scale: Tensions in the Practical Demands on Participatory AI . First Monday

    work page 2024

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.