What Does 'Human-Centred AI' Mean?

Olivia Guest

REVIEW 3 major objections 5 minor 7 references

What Does 'Human-Centred AI' Mean?

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read AI is any techno-social relationship that outsources part of human cognitive labour to a machine, which makes every AI system human-centred by definition.

desk verdict A useful analytic taxonomy for reframing HCAI, wrapped in a universal claim that the paper's own alarm-clock example quietly contradicts. read the letter →

arxiv 2507.19960 v2 pith:75Y4BBZF submitted 2025-07-26 cs.AI

Olivia Guest This is my paper · ORCID

classification cs.AI

keywords artificialintelligencecognitivesciencesociotechnicalrelationshiplabourneuralnetworktechnologycognitionhuman-centredAI

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

The paper argues that 'human-centred AI' is not one design option among others: if AI is defined as any techno-social relationship in which a machine or algorithm takes over some part, however small, of human cognitive labour, then every AI system is human-centred by construction. It pairs technologies with the cognitive work they offload—abacus versus mental arithmetic, alarm clock versus the knocker-upper, camera versus vision, sweatshop versus tailor—to show that the same kind of relationship can enhance, replace, or displace cognition. The analytical core is a two-step definition: first decide whether an artefact stands in a relationship with human cognition, then classify that relationship as enhancement (beneficial), replacement (neutral), or displacement (harmful). The practical task of human-centred AI therefore changes from making AI 'more human' to making the hidden human labour inside every system visible and judging whether it is harmful. That matters because obfuscated cognition, from clocks to neural networks, distorts critical engagement and allows exploitation to hide behind the machine.

What carries the argument

The load-bearing instrument is a two-step analytical scheme: Step 1 deflates AI into any relationship where a machine, tool, model, or algorithm appears to perform human cognitive labour, and Step 2 reinflates it into three non-exclusive labels—replacement, enhancement, displacement—scored on eight properties (valence, effect on cognition, labour obfuscation, human equivalence, human-in-the-loop, human input, desired output, and specification). The scheme carries the argument by relocating analysis from the artefact's intrinsic capacities to the sociotechnical relationship, forcing attention to hidden human labour and making displacement identifiable case by case. A supporting recurring pattern, called Pygmalion displacement, names the historical form in which machines are humanised while the women they replace are dehumanised.

What would settle it

A documented deployed system that performs a normally human cognitive task—for example, translating novel text—while requiring no human-authored training data, no human-labelled examples, no human oversight in operation, and no human maintenance would settle the question; the paper's view predicts that no such system can be produced.

Watch

Extended reading notes

Core claim

The paper's central claim is that the question 'What does human-centred AI mean?' has a deflationary answer: AI is usefully defined as any techno-social relationship in which it appears that an artefact performs some part, however small, of human cognitive labour. Because every such artefact—an abacus no less than a large language model—exists only through humans who set its goals, supply its inputs, or maintain it, all AI implicates human cognition, no matter what. The redefinition then classifies each relationship as replacement (neutral, cognition unaffected, human-in-the-loop rare), enhancement (beneficial, reskilling), or displacement (harmful, deskilling, labour obfuscated). Contemporary large language models, image generators, and chatbots all fall into the displacement category, hiding sweatshop labour and stolen data behind what the author calls a technological veil. Human-centred AI is thus not an aspiration but a recognition: the ghost in the machine is a literal human-in-the-loop, and centring humans means looking that person in the eyes and rejecting benchmark correlations as evidence that machines think.

Load-bearing premise

The argument rests on assuming that 'cognitive labour' is a separable quantity that artefacts can take over and that no machine can function without ongoing human cognitive work, so that 'all AI implicates human cognition' is a factual truth rather than a definitional choice.

Editorial extensions

If this is right

Human-centred AI's four stated goals—enhancing skills, upholding human-aligned values, using human behaviour as a benchmark, and applying psychological methods to machines—each become directives to spot displacement, treat values as socially negotiated, and reject correlationist benchmarks as evidence of thought.
Because any artefact that offloads cognition counts as AI, the timeline of AI stretches back to abacuses, astrolabes, sextants, and the Antikythera mechanism, dissolving the claim that only neural networks or generative systems are AI.
No score on a benchmark and no behavioural match can establish that a machine is human-like or thinking, since correlations between a system and human behaviour are red herrings, not causal evidence.
On this scheme, current LLMs, image generators, and chatbots are displacement AI across the board: harmful, deskilling, maximally obfuscatory, and sustained by hidden human workers.
The label 'human-centred' stops being a property a system can have by design and becomes a demand to make the always-present human-in-the-loop visible and to act on the harms of displacement.

Reading between the lines

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

Applied outward, the definition would classify everyday devices such as GPS navigation, spell-checkers, and calendar reminders as AI relationships, which suggests that 'human-centred AI' regulation would cover far more than machine-learning systems unless a threshold is deliberately chosen.
The tripartite scheme could be tested as a coding instrument: raters could classify descriptions of technologies into replacement, enhancement, and displacement and check whether the categories predict outcomes such as skill retention, wage effects, or user wellbeing.
If the claim that no human-in-the-loop-free machine exists is read as an empirical prediction, then a full supply-chain audit of a supposedly autonomous system that finds no human data-labelling, moderation, or maintenance at any stage would directly test it.
The view implies a reversal of benchmarking: instead of asking how human-like a model is, researchers could measure how much cognitive labour is offloaded, by whom, and under what conditions, making labour visibility itself a scientific variable.

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.

Desk Editor's Note

read the letter

You should know this is a conceptual essay, not an empirical result. The genuinely new part is the tripartite taxonomy: AI as a sociotechnical relationship that offloads cognitive labour, sorted into displacement, enhancement, or replacement, with row-level dimensions like labour obfuscation, human equivalence, and human-in-the-loop. That analytic frame is fresh and does real work. Applying it to low-tech cases—abacus, calculator, alarm clock, camera, sweatshop—is effective and gives HCAI a vocabulary that cuts across the usual LLM-centrism. The insistence that hidden human labour is constitutive, not incidental, of contemporary AI is well supported by the critical STS and feminist literature the paper draws on. Credit where earned: the framework deserves to circulate.

Now the soft spots. The headline claim, 'all AI implicates human cognition; no matter what,' and the Section 4 assertion that there are no human-in-the-loop-free machines, are undercut by the paper's own Table 3, which lists the alarm clock as having 'human-in-the-loop: none.' The paper never disambiguates between hidden labour during operation, any user or designer, and ongoing cognitive input. Under the hidden-labour reading, the alarm clock is a straightforward counterexample. Under the other readings, the claim becomes definitionally true and loses the empirical punch the author wants. This is an internal tension, not just an external objection. Relatedly, the normative mapping—displacement harmful, enhancement beneficial, replacement neutral—is asserted rather than defended, and the universal claims about LLM harms rest on selected examples and secondary sources. Several load-bearing epistemic moves also cite the author's own prior work; some of that work is independently published, but it doesn't ground a universal claim of this scope.

Who is this for? Readers debating what HCAI should be, and critical cognitive scientists, will get value from the taxonomy and the historical framing even if they reject the universal. It deserves peer review, but the author needs to either weaken the universal claim or fix the internal inconsistency with the alarm clock. I'd send it to reviewers with a clear request to address that tension.

Referee Report

3 major / 5 minor

Summary. The paper argues that AI should not be defined by machine capabilities or benchmark performance but as any techno-social relationship in which an artefact appears to perform part of human cognitive labour. It introduces a tripartite taxonomy—replacement (neutral), enhancement (beneficial), and displacement (harmful)—with associated effects on cognition, labour obfuscation, and human involvement (Tables 1–4). It applies this taxonomy to nine examples: abacus/calculator/computer, alarm clock/camera/garment factory, and LLM/image generator/chatbot. The paper concludes that all AI implicates human cognition, that no machine is free of a human-in-the-loop, and that human-centred AI is therefore an analytic feature of AI rather than a design aspiration; the practical task becomes making hidden human labour visible and classifying each AI relationship correctly. It also makes broader metatheoretical claims against correlationist benchmarking and in favour of centring cognitive science in AI research.

Significance. If the proposed framework were accepted, it would reframe human-centred AI from a normative design programme into a descriptive-analytic claim about the nature of AI, connecting critical AI studies, cognitive science, and labour studies. The paper has genuine strengths: the taxonomy in Table 1 is clear and usable; the historical examples (knocker-upper, human computers, Mechanical Turk) are vivid and well chosen; and the paper is explicit that definitions are conventions (Section 2), which is appropriate for a conceptual-engineering contribution. The attention to hidden and obfuscated labour is important and presses a useful corrective to technology-centred definitions of AI. However, the central universal claim is currently secured by stipulation and is internally inconsistent with the paper's own tables, so the contribution needs substantial revision before it can stand as published.

major comments (3)

[Section 4; Table 3, row (f); Table 2, row (f)] The central claim 'There are no Rube Goldberg-like perpetual motion-like human-in-the-loop-free machines. There never will be' (Section 4) is directly contradicted by the paper's own Table 3, column 1, row (f), which classifies the alarm clock's human-in-the-loop as 'none,' and by Table 2, column 2, row (f), which classifies the calculator's human-in-the-loop as 'none.' If 'human-in-the-loop' denotes ongoing hidden human cognitive labour during operation, these entries are counterexamples to the universal negative. If it denotes any human involvement whatsoever (user, designer, maintainer), then the claim is trivially true because every artefact has human origins, and the empirical 'demonstration' in Section 4 collapses into a stipulation. The paper never disambiguates these readings, so the central conclusion is currently internally inconsistent.
[Section 2, Table 1] The definition of AI as 'any techno-social relationship that outsources to machines or algorithms some part, however small, of human cognitive labour' makes the conclusion 'all AI implicates human cognition' true by stipulation. Section 2 itself states that 'definitions have no inherent truth' and are 'agreed conventions,' but the abstract and Section 4 present the universal as a substantive finding ('no matter what'; 'There never will be'). The paper should state explicitly whether the universal claim is analytic (a consequence of the chosen definition) or empirical (a fact about all actual and possible machines). If analytic, the claim is not a discovery about AI but a recommendation about how to use the word 'AI'; if empirical, it needs a non-circular demarcation of cognitive labour and evidence that no machine can operate without human cognitive input.
[Section 3.3, Table 4] Table 4 classifies LLMs, image generators, and chatbots as displacement, harmful, deskilling, and maximally labour-obfuscating, and the text says this is so 'in every case.' These classifications are presented as outcomes of the framework, but they are not derived from any specified procedure; they rest on selected examples and secondary sources, and in the case of 'chatbot versus companionship' the table itself lists the desired output as 'unclear, wellness,' which is not a determinate classification. The paper should either present Table 4 as an illustrative exercise of the taxonomy, explicitly not evidence for universal harm, or provide a systematic argument for why these relationships are always displacement/harmful. This matters because Section 4's first action item ('recognise and act when displacement AI relationships take place') presupposes that such classifications are reliable.

minor comments (5)

[Abstract] The abstract contains formatting and orthographic issues: 'interalia' should be 'inter alia,' and several passages have missing spaces or line-break artefacts that should be corrected in the final version.
[Section 3.1] The thermostat is introduced as 'another great example of no deskilling' but does not appear in any table; either add it to the analysis or remove the aside.
[Section 4] The phrase 'Rube Goldberg-like perpetual motion-like human-in-the-loop-free machines' is grammatically dense and should be rephrased; as written, it is difficult to determine exactly which claim is being made.
[References] Several load-bearing references are to unpublished or in-preparation manuscripts by the author (e.g., Guest, Suarez, et al. 2025; Guest & Martin 2024, 2025; Guest, Scharfenberg, & van Rooĳ 2025); these should be replaced by published, accessible sources or clearly marked as preprints.
[References] The reference to Ryle (1949) gives the publisher as 'Routeledge'; this should read 'Routledge.'

Circularity Check

3 steps flagged · score 8.0 of 10

The paper's central claim that 'all AI implicates human cognition' is an analytic consequence of its own stipulative definition of AI, with additional load-bearing self-citations and an equivocation about 'human-in-the-loop.'

self definitional [Abstract; Section 2, Table 1]
"AI, I argue, is usefully seen as a relationship between technology and humans where it appears that artifacts can perform, to a greater or lesser extent, human cognitive labour. ... To presage the coming analyses, herein AI is any techno-social relationship that outsources to machines or algorithms some part, however small, of human cognitive labour. ... Ultimately, all AI implicates human cognition; no matter what."

AI is stipulated to be a techno-social relationship that outsources human cognitive labour; therefore the universal claim that AI implicates human cognition is already contained in the definition. The 'demonstration' via Tables 2-4 only applies this definition to examples and cannot confirm a universal. The abstract presents the analytic consequence ('cannot be any other way') as though it were an independent discovery, satisfying the self-definitional circularity pattern: the predicted conclusion is identical to the defining property.
self citation load bearing [Section 2, n.1; Section 4]
"In general, correlationism may be unproblematic, but in a setting where computation is involved correlations cannot function as a useful guide, serving more as red herring than anything else (Guest & Martin, 2023, 2024, 2025; Guest, Scharfenberg, & van Rooĳ, 2025). ... no amount of high scores on benchmarks, or any other correlationary evidence, can ever pile up high enough to graduate to a causal claim."

The normative conclusion that benchmarks are meaningless and correlationism is a red herring is load-bearing for the paper's recommendation to 'roundly reject benchmarks as meaningful' (Section 4, point 3) and for its diagnosis that current AI discourse 'pervert[s] cognitive science.' The only cited support for this epistemic premise is prior work by the same author (Guest & Martin; Guest, Scharfenberg, & van Rooij), not an argument reproduced here. Since those works are not machine-checked or independently grounded within the paper, the argument at this point reduces to self-citation rather than external evidence.

1 more flagged steps

other [Table 3, col. 1, row f; Section 4]
"f) human-in-the-loop none ... There are no Rube Goldberg-like perpetual motion-like human-in-the-loop-free machines. There never will be."

The universal negative requires every AI system to have a human-in-the-loop, but Table 3 classifies the alarm clock as having 'none' in that category. The claim can be saved only by expanding 'human-in-the-loop' to include the user or designer, in which case the claim is trivially true of any artifact; under the narrow hidden-labour meaning, the alarm clock is a counterexample. The demonstration thus either overreaches empirically or becomes analytic, so the proof of the central thesis equivocates between the two readings.

full rationale

The paper's headline claim—'all AI implicates human cognition; no matter what'—is not an empirical discovery but an unpacking of its own stipulative definition: AI is defined as a relationship that outsources human cognitive labour to machines. Any system satisfying that definition trivially implicates human cognition, so the abstract's 'cannot be any other way' is analytic. The typologies in Tables 2-4 apply the definition to historical and contemporary examples and these applications are informative, but they do not independently ground the universal. Separately, several load-bearing epistemic claims (correlationism as red herring; benchmarks as meaningless) are supported by citations to the author's own prior work (Guest & Martin 2023-2025; Guest, Scharfenberg, & van Rooij 2025), and the Section 4 universal 'no human-in-the-loop-free machines' conflicts with Table 3's alarm-clock row f ('none') unless 'human-in-the-loop' is read so broadly that the claim becomes definitional. The paper is self-aware about proposing a redefinition—'definitions have no inherent truth'—but the presentation of the analytic consequence as a demonstration means the central claim reduces by construction to its input. The independent examples and political analysis prevent this from being a 10; score 8.

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

The central claim is a stipulative redefinition, so the paper carries no numeric free parameters. The key axioms are the definition itself, the normative valence mapping, the rejection of correlationist evidence, and the universal claim that no machine runs without hidden human cognitive labour. These are asserted rather than derived. No new physical entities are introduced; the 'AI relationship' is a redefinition of an existing term.

assumptions (4)

ad hoc to paper AI is any techno-social relationship in which an artefact appears to perform part of human cognitive labour.
Stipulated in Section 2 and Table 1; the conclusion that all AI implicates human cognition follows directly from this definition.
ad hoc to paper Displacement is harmful, enhancement is beneficial, and replacement is neutral, with corresponding effects on cognition (deskilling, reskilling, unaffected).
Normative mapping asserted in Table 1, Step 2, and applied in Tables 2-4 without a defended theory of harm.
domain assumption Correlations and benchmark performance cannot support causal or human-likeness claims about machines.
The anti-correlationist premise is inherited from the author's prior work (Guest & Martin 2023, 2024, 2025) and is load-bearing for rejecting HCAI themes 3 and 4.
ad hoc to paper There are no human-in-the-loop-free machines; every apparently autonomous system is sustained by human cognitive labour.
Universal empirical and metaphysical claim asserted in Section 4 with no supporting evidence.

how reviews work

0 comments

Cite this review

Pith. "Pith review of What Does 'Human-Centred AI' Mean?." pith.science (2026). https://pith.science/paper/75Y4BBZF

@misc{pith2026250719960,
  author       = {Pith},
  title        = {Pith review of: What Does 'Human-Centred AI' Mean?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/75Y4BBZF}},
  note         = {Machine review of arXiv:2507.19960}
}

read the original abstract

While it seems sensible that human-centred artificial intelligence (AI) means centring "human behaviour and experience," it cannot be any other way. AI, I argue, is usefully seen as a relationship between technology and humans where it appears that artifacts can perform, to a greater or lesser extent, human cognitive labour. This is evinced using examples that juxtapose technology with cognition, inter alia: abacus versus mental arithmetic; alarm clock versus knocker-upper; camera versus vision; and sweatshop versus tailor. Using novel definitions and analyses, sociotechnical relationships can be analysed into varying types of: displacement (harmful), enhancement (beneficial), and/or replacement (neutral) of human cognitive labour. Ultimately, all AI implicates human cognition; no matter what. Obfuscation of cognition in the AI context -- from clocks to artificial neural networks -- results in distortion, in slowing critical engagement, perverting cognitive science, and indeed in limiting our ability to truly centre humans and humanity in the engineering of AI systems. To even begin to de-fetishise AI, we must look the human-in-the-loop in the eyes.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

7 extracted references · 4 canonical work pages

[1]

ChatGPT Psychosis

Adler, P. S. (1990). Marx, Machines, and Skill.Technology and Culture,31(4). Agar, J. (2003).The government machine: a revolutionary history of the computer. MIT press. 15 O. Guest Alač, M., Gluzman, Y., Aflatoun, T., Bari, A., Jing, B., & Mozqueda, G. (2020). Talking to a toaster: how everyday interactions with digital voice assistants resist a return to...

arXiv 1990
[4]

(2024).Governed By Affect: Hot Cognition and the End of Cold War Psychology

Perrigo,B.(2023).Exclusive:OpenAIusedKenyanworkersonlessthan$2perhourtomake ChatGPT less toxic.Time Magazine, 18, 2023.https://time.com/6247678/openai-c hatgpt-kenya-workers/ Pettit, M. (2024).Governed By Affect: Hot Cognition and the End of Cold War Psychology. Oxford University Press. Pfaffenberger, B. (1988). Fetishised objects and humanised nature: to...

arXiv 2023
[14]

S., Atle Mikkola Kjøsen

Nick Dyer-Witheford, J. S., Atle Mikkola Kjøsen. (2019).Inhuman Power: Artificial Intelligence And The Future Of Capitalism. Pluto Press. Novack, G. (1968).Empiricism and Its Evolution: A Marxist View. Pathfinder Press. Ojewale,V.,Steed,R.,Vecchione,B.,Birhane,A.,&Raji,I.D.(2025).TowardsAIaccountability infrastructure:GapsandopportunitiesinAIaudittooling....

work page 2019
[80]

Ponomareva, V. (1998). The Human Factor in Space Exploration: Soviet and American Ap- proaches. Proceedings of the 1998 annual conference of the Institute for the History of Natural Science and Technology], Moscow, 614–18. Powell, B. (1970). Descartes’ Machines.Proceedings of the Aristotelian Society,71, 209–222. Raji, I. D., Kumar, I. E., Horowitz, A., &...

work page 1998
[204]

Rasenberg,M.,Amha,A.,Coler,M.,VanKoppen,M.,vanMiltenburg,E.,DeRĳk,L.,Stommel, W., & Dingemanse, M. (2023). Reimagining language: Towards a better understand- ing of language by including our interactions with non-humans.Linguistics in the Netherlands,40(1), 309–317. Reader,T.M.P.(2024).SiliconValley’sObsessionWithAILooksaLotLikeReligion. TheMIT Press Read...

work page arXiv 2023
[913]

On discriminative vs generative classifiers: A comparison of logistic regression and naive bayes

Weizenbaum,J.(1966).ELIZA—acomputerprogramforthestudyofnaturallanguagecom- munication between man and machine.Communications of the ACM,9(1), 36–45. Weizenbaum, J. (1976). Computer power and human reason: From judgment to calculation. Wendling, A. (2002). Partial Liberations.International Studies in Philosophy, 34(2), 169–185. Widder,D.G.,Whittaker,M.,&We...

work page 1966
[5821]

child machine,

Geyer, W., Weisz, J., Pinhanez, C. S., & Daly, E. (2022, August). What is human-centered AI? https://research.ibm.com/blog/what-is-human-centered-ai Goetze,T.S.(2024).AIartistheft:Labour,extraction,andexploitation:Or,onthedangersof stochastic Pollocks.Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, 186–196. Grier, D. ...

arXiv 2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

ChatGPT Psychosis

Adler, P. S. (1990). Marx, Machines, and Skill.Technology and Culture,31(4). Agar, J. (2003).The government machine: a revolutionary history of the computer. MIT press. 15 O. Guest Alač, M., Gluzman, Y., Aflatoun, T., Bari, A., Jing, B., & Mozqueda, G. (2020). Talking to a toaster: how everyday interactions with digital voice assistants resist a return to...

arXiv 1990

[2] [4]

(2024).Governed By Affect: Hot Cognition and the End of Cold War Psychology

Perrigo,B.(2023).Exclusive:OpenAIusedKenyanworkersonlessthan$2perhourtomake ChatGPT less toxic.Time Magazine, 18, 2023.https://time.com/6247678/openai-c hatgpt-kenya-workers/ Pettit, M. (2024).Governed By Affect: Hot Cognition and the End of Cold War Psychology. Oxford University Press. Pfaffenberger, B. (1988). Fetishised objects and humanised nature: to...

arXiv 2023

[3] [14]

S., Atle Mikkola Kjøsen

Nick Dyer-Witheford, J. S., Atle Mikkola Kjøsen. (2019).Inhuman Power: Artificial Intelligence And The Future Of Capitalism. Pluto Press. Novack, G. (1968).Empiricism and Its Evolution: A Marxist View. Pathfinder Press. Ojewale,V.,Steed,R.,Vecchione,B.,Birhane,A.,&Raji,I.D.(2025).TowardsAIaccountability infrastructure:GapsandopportunitiesinAIaudittooling....

work page 2019

[4] [80]

Ponomareva, V. (1998). The Human Factor in Space Exploration: Soviet and American Ap- proaches. Proceedings of the 1998 annual conference of the Institute for the History of Natural Science and Technology], Moscow, 614–18. Powell, B. (1970). Descartes’ Machines.Proceedings of the Aristotelian Society,71, 209–222. Raji, I. D., Kumar, I. E., Horowitz, A., &...

work page 1998

[5] [204]

Rasenberg,M.,Amha,A.,Coler,M.,VanKoppen,M.,vanMiltenburg,E.,DeRĳk,L.,Stommel, W., & Dingemanse, M. (2023). Reimagining language: Towards a better understand- ing of language by including our interactions with non-humans.Linguistics in the Netherlands,40(1), 309–317. Reader,T.M.P.(2024).SiliconValley’sObsessionWithAILooksaLotLikeReligion. TheMIT Press Read...

work page arXiv 2023

[6] [913]

On discriminative vs generative classifiers: A comparison of logistic regression and naive bayes

Weizenbaum,J.(1966).ELIZA—acomputerprogramforthestudyofnaturallanguagecom- munication between man and machine.Communications of the ACM,9(1), 36–45. Weizenbaum, J. (1976). Computer power and human reason: From judgment to calculation. Wendling, A. (2002). Partial Liberations.International Studies in Philosophy, 34(2), 169–185. Widder,D.G.,Whittaker,M.,&We...

work page 1966

[7] [5821]

child machine,

Geyer, W., Weisz, J., Pinhanez, C. S., & Daly, E. (2022, August). What is human-centered AI? https://research.ibm.com/blog/what-is-human-centered-ai Goetze,T.S.(2024).AIartistheft:Labour,extraction,andexploitation:Or,onthedangersof stochastic Pollocks.Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, 186–196. Grier, D. ...

arXiv 2024