Pith. sign in

REVIEW 3 major objections 6 minor 32 references

AI fairness research is concentrated in a few countries, most of all in the foundational domain that supplies the rest of the field's definitions and benchmarks.

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 ↗

T0 review · grok-4.5

2026-07-11 05:35 UTC pith:BAZHUXIU

load-bearing objection Solid first field-wide map of AI-bias research; concentration is real inside the sampled frame, strongest in the foundational domain, and the sampling caveat is already owned by the authors. the 3 major comments →

arxiv 2607.05574 v1 pith:BAZHUXIU submitted 2026-07-06 cs.CY cs.AIcs.DL

Whose fairness? Structural concentration in AI bias research

Abhash Shrestha , Subigya Gautam , Anu Sapkota , Sanju Tiwari , Tek Raj Chhetri This is my paper
classification cs.CY cs.AIcs.DL
keywords AI biasfairness researchbibliometricsresearch concentrationgeographic inequalitysemantic clusteringcollaboration networksgeneralizability
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.

This paper argues that the research community that produces AI bias and fairness methods is itself structurally concentrated by country, institution, and author, and that the concentration is strongest in the general fairness domain that other application areas inherit from. From a corpus of 692 papers across five themes, the authors show that the United States leads output and collaboration networks in every domain, and most strongly in general fairness and bias mitigation, which is the largest and most-cited domain and which cuts across the semantic clusters. Low- and middle-income countries are largely absent from authorship and collaboration networks, and citations are heavily skewed, so a small fraction of papers shapes the field. Because application areas adopt definitions and benchmarks from this foundational layer, concentration there can propagate local assumptions as if they were universal. The authors release an interactive atlas so the field's structure can be monitored continuously rather than only in retrospect.

Core claim

AI bias research is structurally concentrated, and the concentration is greatest geographically in general fairness and bias mitigation—the largest, most-cited, semantically cross-cutting domain that supplies the definitions, benchmarks, and mitigation frameworks used across health AI, LLMs and NLP, recommender systems, and graph-based fairness. United States authors lead publication output and collaboration networks in every domain and account for more than half of first-authored work in that foundational domain; low- and middle-income countries remain largely absent; and citation influence is highly skewed (median 9, mean 93.5).

What carries the argument

A 692-paper corpus across five manually assigned thematic domains, analyzed with bibliometric measures (all-author and first-author country and institution counts, collaboration networks, Gini coefficients, citation distributions) and semantic clustering of abstracts (Sentence-BERT embeddings, UMAP, HDBSCAN), which identifies general fairness as the cross-cutting foundational domain where concentration is strongest.

Load-bearing premise

The measured concentration is taken to reflect the true production structure of the field, even though the corpus comes mainly from English-language, Global-North-indexed databases and may under-represent work from other languages and venues.

What would settle it

A parallel corpus built with equal coverage of major non-English and Global-South venues that reverses or erases United States dominance and the relative concentration in the general-fairness domain would falsify the central claim about structural concentration in the field as a whole.

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

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. The manuscript presents a bibliometric and semantic analysis of 692 AI-bias publications (2015–2026) across five manually assigned thematic domains, arguing that research production is structurally concentrated by country, institution, and author, and that this concentration is strongest in General Fairness & Bias Mitigation—the largest, most-cited, and semantically cross-cutting domain that supplies definitions and benchmarks to application areas. Using all-author and first-author counts, co-authorship networks, Gini coefficients, non-parametric citation tests, and Sentence-BERT + UMAP + HDBSCAN clustering (with a default-UMAP robustness check), the authors report US leadership (e.g., 58.3% of first-authored General Fairness papers), sparse LMIC participation and collaboration, and highly skewed citations (median 9, mean 93.5). They release an interactive atlas and public data/code for ongoing monitoring.

Significance. If the measured concentration is taken as a fair description of the indexed literature that sets fairness standards, the paper makes a useful, field-level contribution: it quantifies who produces the methodological layer that application domains inherit, rather than only documenting underrepresentation within a single application area. Strengths include dual all-author/first-author reporting, appropriate non-parametric statistics with bootstrap CIs, unsupervised clustering validated post-hoc (ARI/NMI) without using domain labels in the pipeline, a default-UMAP robustness check, and public corpus, code, and interactive atlas. The work is timely for cs.CY and fairness research policy, even if the strongest epistemic claim about non-generalizability remains partly inferential.

major comments (3)
  1. [Methods (Corpus construction); Limitations; Abstract; Discussion] The central claim is framed as field-wide structural concentration (Abstract; Introduction; Discussion), but the corpus is built from English-language, Global-North-indexed sources (IEEE Xplore, ACM DL, Scopus, ScienceDirect, Engineering Village, FAccT, snowballing; Methods, Corpus construction). Limitations correctly notes under-representation of non-English and regional venues, yet the paper asserts that broader coverage is “unlikely to reverse” the pattern without a sensitivity check. Because the load-bearing inference is that concentration is greatest precisely in the foundational domain the rest of the field inherits, the manuscript should either (i) re-scope claims explicitly to the well-indexed English literature that currently sets mainstream benchmarks, or (ii) add a concrete sensitivity analysis (e.g., regional venues, non-English indexes, or stratified snowballing from Global-
  2. [Abstract; Discussion (esp. paragraphs on epistemic influence and implications)] The Discussion links geographic concentration in General Fairness to a risk that mitigation methods “may not generalize,” while also stating that the data “do not directly establish” reduced generalizability and citing the authors’ Nepali LLM study [9] as illustration. That hedge is appropriate but under-weighted relative to the Abstract and closing paragraphs, which present non-generalizability as a field-wide concern flowing from the bibliometric pattern. Tighten the causal language: treat concentration as a measured structural fact within the sampled frame, and treat non-generalizability as a motivated hypothesis requiring external validation designs (multi-site benchmarks, reporting of validation populations), not as a demonstrated consequence of the Gini/country shares alone.
  3. [Methods (Corpus construction); Results (Domain distribution); Fig. 3] Domain assignment is manual and load-bearing for the claim that concentration is greatest in the foundational domain (Results, Domain distribution; Fig. 1; Fig. 3). Semantic clustering supports coherence (Fig. 6; ARI/NMI; robustness check), but Methods does not report inter-rater reliability, double-coding fraction, or a decision protocol for multi-domain papers. Given that General Fairness is defined as cross-cutting and is the domain where US dominance is strongest (58.3% first-author), please report how borderline papers were assigned and whether reassignment of a plausible fraction of multi-topic papers would change the domain-wise country shares or the “most concentrated foundational domain” ranking.
minor comments (6)
  1. [Abstract] Abstract: “the AI bias research are structurally concentrated” — subject–verb agreement; also “mean =93.5” needs a space.
  2. [Fig. 2c] Fig. 2c collaboration heatmap is dense; a thresholded network or top-k edges would improve readability without changing the Western-hub message.
  3. [Fig. 4 caption] Institutional heatmaps (Fig. 4) report percentages among the top-20 institutions shown; state this more prominently in the caption so readers do not read them as shares of the full institutional population.
  4. [Results (Country-level); Methods (Statistics)] Chi-square uses 11 countries × 5 domains with Monte Carlo p-values (Results); report the exact country list and the ≥10 first-authored-paper threshold in one place for reproducibility.
  5. [Methods (Author and institution counting)] ORCID-based author deduplication was not applied (Methods); briefly discuss possible inflation of unique-author counts for common names, especially for China/India comparisons.
  6. [Results (Temporal trends)] arXiv version date in header (“6 Jul 2026”) and incomplete 2026 counts are fine if noted; ensure temporal claims in the text always flag incomplete-year status (already partly done in Results).

Circularity Check

0 steps flagged

No circularity: empirical bibliometric measurement of an external 692-paper corpus; manual domains validated post-hoc by unsupervised clustering that never used domain labels as input.

full rationale

The paper’s load-bearing chain is observational, not definitional or fitted-then-predicted. Corpus construction (keyword search + snowballing + FAccT + manual screening) yields 692 papers; country/institution/first-author counts, Gini coefficients, collaboration edges, Kruskal–Wallis/Dunn citation tests, and chi-square domain-by-country tests are computed directly from that external metadata. Semantic clustering encodes abstracts with Sentence-BERT, projects with UMAP, and clusters with HDBSCAN; domain labels are applied only after clustering to compute ARI/NMI, so the four clusters are not forced by the five manual categories. The sole self-citation ([9], Nepali LLM bias) appears only as an illustrative risk example in Discussion/Introduction and is not an input to any count, Gini, network, or cluster. No parameter is fitted to a subset and then re-presented as a prediction; no uniqueness theorem or ansatz is imported from the authors’ prior work to force the result. Sampling-frame limitations (English/Global-North databases) affect external validity of the concentration claim but do not make the reported statistics circular by construction. Honest non-finding: score 0.

Axiom & Free-Parameter Ledger

2 free parameters · 5 axioms · 2 invented entities

The central claims rest on standard bibliometric and clustering practice plus a set of domain-construction choices. No free parameters are fitted to produce the concentration statistics themselves; clustering hyperparameters were selected by silhouette but the main geographic results do not depend on them. Invented entities are limited to the five thematic domains used as an analytic scaffold.

free parameters (2)
  • UMAP n_neighbors / min_dist and HDBSCAN min_cluster_size / min_samples = n_neighbors=30, min_dist=0.0, min_cluster_size=20, min_samples=10
    Chosen by grid search maximizing silhouette; a default-parameter robustness check is reported, but the exact cluster count (4 vs 3) is hyperparameter-sensitive even though domain-overlap conclusions hold.
  • Country inclusion threshold for chi-square and Spearman = ≥10 papers
    Analyses restricted to countries with ≥10 first-authored papers (11 countries) to avoid sparse cells; threshold is conventional but hand-chosen.
axioms (5)
  • domain assumption Papers retrieved from English-language Global-North databases plus FAccT and snowballing constitute a usable sample of AI-bias research for measuring structural concentration.
    Stated in Methods (corpus construction) and Limitations; authors note coverage bias cannot be fully disentangled from production gaps.
  • domain assumption Manual assignment of each paper to one of five primary thematic domains is a valid analytic partition.
    Methods and Results; supported post-hoc by HDBSCAN alignment (ARI/NMI above chance) but embedding silhouette is modest.
  • domain assumption First-author country/institution is a reasonable proxy for research leadership; all-author counts proxy collaborative participation.
    Explicitly used throughout Results to distinguish leadership from participation.
  • standard math Standard bibliometric indicators (publication counts, co-authorship edges, OpenAlex citation counts, Gini) capture the field's documented structure.
    Conventional scientometric practice; non-parametric tests chosen for skewed data.
  • domain assumption Concentration of production in the general-fairness domain raises a generalizability risk for methods adopted elsewhere.
    Discussion; presented as a concern illustrated by the authors' prior Nepali LLM study, not as a direct causal proof from the bibliometric data.
invented entities (2)
  • Five thematic domains (General Fairness & Bias Mitigation, Health & Clinical AI, LLMs & NLP, Recommender Systems, Graph-Based Fairness) no independent evidence
    purpose: Analytic scaffold for measuring where concentration is strongest and how foundational work relates to applications.
    Manually defined then validated by unsupervised clustering; useful but overlapping categories rather than natural kinds.
  • Interactive Bias Atlas (biasatlas.cair-nepal.org) independent evidence
    purpose: Continuous public monitoring of the field's geographic, thematic, and citation structure.
    New artifact released with the paper; independent of the concentration claims themselves.

pith-pipeline@v1.1.0-grok45 · 24494 in / 3290 out tokens · 27517 ms · 2026-07-11T05:35:01.752429+00:00 · methodology

0 comments
read the original abstract

Artificial intelligence increasingly mediates consequential decisions in healthcare, law, and public services, and the field has responded with an extensive methodology for measuring and mitigating bias. Yet the fairness definitions, benchmarks, and debiasing frameworks on which this methodology rests are treated as universal while being produced by a research community whose composition has never been characterized. We show that the AI bias research are structurally concentrated, and that this concentration is greatest, geographically, in precisely the domain the rest of the field inherits from. Analyzing 692 publications spanning five thematic domains, combining bibliometric analysis with semantic clustering, we find that research activity is dominated by a small set of countries, institutions, and authors, with the United States leading publication output and collaboration networks across every domain and most strongly in general fairness and bias mitigation, the largest, most-cited domain with meaningful representation across all four semantic clusters. Low- and middle-income countries remain largely absent from the community and its collaboration networks, and citation influence is highly skewed (median = 9; mean =93.5 ), indicating that a small fraction of publications disproportionately shapes the field. Because the general-fairness domain supplies the definitions and benchmarks that application areas apply, concentration of research effort in this foundational domain propagates across AI bias research as a whole - raising the concern that mitigation methods developed and validated within a narrow set of contexts may not generalize to all populations and settings where AI is deployed. We provide an interactive atlas for continuous monitoring of the field's structure.

discussion (0)

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

Reference graph

Works this paper leans on

32 extracted references · 11 canonical work pages

  1. [1]

    Microchemical Journal205, 111307 (2024) https://doi.org/10.1016/j.microc.2024.111307

    Ahmed Taha, B., Kadhim, A.C., Addie, A.J., Haider, A.J., Azzahrani, A.S., Raizada, P., Rustagi, S., Chaudhary, V., Arsad, N.: Advancing cancer diagnostics through multifaceted optical biosensors supported by nanomaterials and artifi- cial intelligence: A panoramic outlook. Microchemical Journal205, 111307 (2024) https://doi.org/10.1016/j.microc.2024.111307

    work page doi:10.1016/j.microc.2024.111307 2024

  2. [2]

    Drug Discovery Today29(6), 103992 (2024) https://doi.org/10.1016/j

    Gangwal, A., Lavecchia, A.: Unleashing the power of generative ai in drug dis- covery. Drug Discovery Today29(6), 103992 (2024) https://doi.org/10.1016/j. drudis.2024.103992

    doi:10.1016/j 2024

  3. [3]

    Science Advances4(1), 5580 (2018) https://doi.org/10.1126/sciadv.aao5580 16

    Dressel, J., Farid, H.: The accuracy, fairness, and limits of predicting recidivism. Science Advances4(1), 5580 (2018) https://doi.org/10.1126/sciadv.aao5580 16

    doi:10.1126/sciadv.aao5580 2018

  4. [4]

    Sci6(1), 3 (2023) https://doi.org/10.3390/ sci6010003

    Ferrara, E.: Fairness and bias in artificial intelligence: A brief survey of sources, impacts, and mitigation strategies. Sci6(1), 3 (2023) https://doi.org/10.3390/ sci6010003

    2023

  5. [5]

    WIREs Data Mining and Knowledge Discovery10(3), 1356 (2020) https://doi.org/10.1002/widm.1356

    Ntoutsi, E., Fafalios, P., Gadiraju, U., Iosifidis, V., Nejdl, W., Vidal, M.- E., Ruggieri, S., Turini, F., Papadopoulos, S., Krasanakis, E., Kompatsiaris, I., Kinder-Kurlanda, K., Wagner, C., Karimi, F., Fernandez, M., Alani, H., Berendt, B., Kruegel, T., Heinze, C., Broelemann, K., Kasneci, G., Tiropanis, T., Staab, S.: Bias in data-driven artificial in...

    doi:10.1002/widm.1356 2020

  6. [6]

    IEEE Trans- actions on Neural Networks and Learning Systems35(1), 117–131 (2024) https: //doi.org/10.1109/TNNLS.2022.3172365

    Giloni, A., Grolman, E., Hagemann, T., Fromm, R., Fischer, S., Elovici, Y., Shabtai, A.: Benn: Bias estimation using a deep neural network. IEEE Trans- actions on Neural Networks and Learning Systems35(1), 117–131 (2024) https: //doi.org/10.1109/TNNLS.2022.3172365

    work page doi:10.1109/tnnls.2022.3172365 2024

  7. [7]

    The Lancet Digital Health6(1), 12–22 (2024) https://doi.org/10.1016/S2589-7500(23)00225-X

    Zack, T., Lehman, E., Suzgun, M., Rodriguez, J.A., Celi, L.A., Gichoya, J., Juraf- sky, D., Szolovits, P., Bates, D.W., Abdulnour, R.-E.E., Butte, A.J., Alsentzer, E.: Assessing the potential of gpt-4 to perpetuate racial and gender biases in health care: a model evaluation study. The Lancet Digital Health6(1), 12–22 (2024) https://doi.org/10.1016/S2589-7...

    doi:10.1016/s2589-7500(23)00225-x 2024

  8. [8]

    Association for Computational Linguistics, Abu Dhabi, UAE (2025)

    Lin, L., Wang, L., Guo, J., Wong, K.-F.: Investigating Bias in LLM-Based Bias Detection: Disparities betwceen LLMs and Human Perception. Association for Computational Linguistics, Abu Dhabi, UAE (2025). https://aclanthology.org/ 2025.coling-main.709/

    2025

  9. [9]

    https://arxiv.org/abs/2603.07792

    Pandey, A., Chhetri, T.R.: Dual-Metric Evaluation of Social Bias in Large Lan- guage Models: Evidence from an Underrepresented Nepali Cultural Context (2026). https://arxiv.org/abs/2603.07792

    work page arXiv 2026

  10. [10]

    Artificial Intelligence and Law (2024) https://doi.org/10.1007/s10506-024-09389-8

    Engel, C., Linhardt, L., Schubert, M.: Code is law: how compas affects the way the judiciary handles the risk of recidivism. Artificial Intelligence and Law (2024) https://doi.org/10.1007/s10506-024-09389-8

    work page doi:10.1007/s10506-024-09389-8 2024

  11. [11]

    In: Proceedings of the Conference on Fairness, Accountability, and Transparency

    Hutchinson, B., Mitchell, M.: 50 years of test (un)fairness: Lessons for machine learning. In: Proceedings of the Conference on Fairness, Accountability, and Transparency. FAT* ’19, pp. 49–58. Association for Computing Machin- ery, New York, NY, USA (2019). https://doi.org/10.1145/3287560.3287600 . https://doi.org/10.1145/3287560.3287600

    doi:10.1145/3287560.3287600 2019

  12. [12]

    ACM Computing Surveys54(6), 1–35 (2021) https://doi.org/10.1145/3457607

    Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., Galstyan, A.: A survey on bias and fairness in machine learning. ACM Computing Surveys54(6), 1–35 (2021) https://doi.org/10.1145/3457607

    doi:10.1145/3457607 2021

  13. [13]

    ACM Computing 17 Surveys56(7), 1–38 (2024) https://doi.org/10.1145/3616865

    Caton, S., Haas, C.: Fairness in machine learning: A survey. ACM Computing 17 Surveys56(7), 1–38 (2024) https://doi.org/10.1145/3616865

    doi:10.1145/3616865 2024

  14. [14]

    ACM Journal on Responsible Computing1(2), 1–52 (2024) https://doi.org/10.1145/3631326

    Hort, M., Chen, Z., Zhang, J.M., Harman, M., Sarro, F.: Bias mitigation for machine learning classifiers: A comprehensive survey. ACM Journal on Responsible Computing1(2), 1–52 (2024) https://doi.org/10.1145/3631326

    work page doi:10.1145/3631326 2024

  15. [15]

    ACM Comput

    Pessach, D., Shmueli, E.: A review on fairness in machine learning. ACM Comput. Surv.55(3) (2022) https://doi.org/10.1145/3494672

    doi:10.1145/3494672 2022

  16. [16]

    ACM Computing Surveys55(13s), 1–37 (2023) https://doi.org/10.1145/3597199

    Tang, Z., Zhang, J., Zhang, K.: What-is and how-to for fairness in machine learn- ing: A survey, reflection, and perspective. ACM Computing Surveys55(13s), 1–37 (2023) https://doi.org/10.1145/3597199

    work page doi:10.1145/3597199 2023

  17. [17]

    arXiv (2023)

    Li, Y., Du, M., Song, R., Wang, X., Wang, Y.: A Survey on Fairness in Large Language Models. arXiv (2023). https://doi.org/10.48550/ARXIV.2308.10149

    Pith/arXiv arXiv doi:10.48550/arxiv.2308.10149 2023

  18. [18]

    Computational Linguistics50(3), 1097–1179 (2024) https://doi.org/10

    Gallegos, I.O., Rossi, R.A., Barrow, J., Tanjim, M.M., Kim, S., Dernoncourt, F., Yu, T., Zhang, R., Ahmed, N.K.: Bias and fairness in large language models: A survey. Computational Linguistics50(3), 1097–1179 (2024) https://doi.org/10. 1162/coli a 00524

    2024

  19. [19]

    ACM SIGKDD Explorations Newsletter26(1), 34–48 (2024) https://doi

    Chu, Z., Wang, Z., Zhang, W.: Fairness in large language models: A taxonomic survey. ACM SIGKDD Explorations Newsletter26(1), 34–48 (2024) https://doi. org/10.1145/3682112.3682117

    doi:10.1145/3682112.3682117 2024

  20. [20]

    In: Proceedings of the 30th International Conference on Intelligent User Interfaces, pp

    Dudy, S., Tholeti, T., Ramachandranpillai, R., Ali, M., Li, T.J.-J., Baeza-Yates, R.: Unequal Opportunities: Examining the Bias in Geographical Recommen- dations by Large Language Models. In: Proceedings of the 30th International Conference on Intelligent User Interfaces, pp. 1499–1516. ACM, Cagliari Italy (2025). https://doi.org/10.1145/3708359.3712111

    doi:10.1145/3708359.3712111 2025

  21. [21]

    Journal of Nonverbal Behavior48(2), 323– 344 (2024) https://doi.org/10.1007/s10919-024-00454-z

    Pang, H.T., Zhou, X., Chu, M.: Cross-cultural differences in using nonverbal behaviors to identify indirect replies. Journal of Nonverbal Behavior48(2), 323– 344 (2024) https://doi.org/10.1007/s10919-024-00454-z

    work page doi:10.1007/s10919-024-00454-z 2024

  22. [22]

    PLOS ONE10(2), 1–15 (2015) https://doi.org/10.1371/journal

    Uono, S., Hietanen, J.K.: Eye contact perception in the west and east: A cross- cultural study. PLOS ONE10(2), 1–15 (2015) https://doi.org/10.1371/journal. pone.0118094

    doi:10.1371/journal 2015

  23. [23]

    https://arxiv.org/abs/2102.01265

    Chan, A., Okolo, C.T., Terner, Z., Wang, A.: The Limits of Global Inclusion in AI Development (2021). https://arxiv.org/abs/2102.01265

    Pith/arXiv arXiv 2021

  24. [24]

    Perc, M.: The matthew effect in empirical data. Journal of The Royal Society Interface11(98), 20140378 (2014) https://doi.org/ 10.1098/rsif.2014.0378 https://royalsocietypublishing.org/rsif/article- pdf/doi/10.1098/rsif.2014.0378/192325/rsif.2014.0378.pdf 18

    work page doi:10.1098/rsif.2014.0378 2014

  25. [25]

    Science 159(3810), 56–63 (1968) https://doi.org/10.1126/science.159.3810.56 https://www.science.org/doi/pdf/10.1126/science.159.3810.56

    Merton, R.K.: The matthew effect in science. Science 159(3810), 56–63 (1968) https://doi.org/10.1126/science.159.3810.56 https://www.science.org/doi/pdf/10.1126/science.159.3810.56

    doi:10.1126/science.159.3810.56 1968

  26. [26]

    Nature Communications 15(1), 7527 (2024) https://doi.org/10.1038/s41467-024-51714-x

    Schmallenbach, L., B¨ arnighausen, T.W., Lerchenmueller, M.J.: The global geog- raphy of artificial intelligence in life science research. Nature Communications 15(1), 7527 (2024) https://doi.org/10.1038/s41467-024-51714-x

    work page doi:10.1038/s41467-024-51714-x 2024

  27. [27]

    PLOS Global Public Health4(1), 0002513 (2024) https://doi.org/10.1371/journal.pgph.0002513

    Alberto, I.R.I., Alberto, N.R.I., Altinel, Y., Blacker, S., Binotti, W.W., Celi, L.A., Chua, T., Fiske, A., Griffin, M., Karaca, G., Mokolo, N., Naawu, D.K.N., Patscheider, J., Petushkov, A., Quion, J.M., Senteio, C., Taisbak, S., Tırnova, ˙I., Tokashiki, H., Velasquez, A., Yaghy, A., Yap, K.: A scientometric analysis of fairness in health ai literature. ...

    work page doi:10.1371/journal.pgph.0002513 2024

  28. [28]

    Science366(6464), 447–453 (2019) https://doi.org/10.1126/science.aax2342

    Obermeyer, Z., Powers, B., Vogeli, C., Mullainathan, S.: Dissecting racial bias in an algorithm used to manage the health of populations. Science366(6464), 447–453 (2019) https://doi.org/10.1126/science.aax2342

    doi:10.1126/science.aax2342 2019

  29. [29]

    FAccT ’21, pp

    Bender, E.M., Gebru, T., McMillan-Major, A., Shmitchell, S.: On the dangers of stochastic parrots: Can language models be too big? In: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. FAccT ’21, pp. 610–623. ACM, . (2021). https://doi.org/10.1145/3442188.3445922

    doi:10.1145/3442188.3445922 2021

  30. [30]

    In: Proceedings of the 3rd Innovations in Theoretical Computer Sci- ence Conference

    Dwork, C., Hardt, M., Pitassi, T., Reingold, O., Zemel, R.: Fairness through awareness. In: Proceedings of the 3rd Innovations in Theoretical Computer Sci- ence Conference. ITCS ’12, pp. 214–226. ACM, New York, NY, USA (2012). https://doi.org/10.1145/2090236.2090255

    doi:10.1145/2090236.2090255 2012

  31. [31]

    Cambridge University Press, Cambridge (2023)

    Oliveira, G., Baggs, E.: Psychology’s WEIRD Problems. Cambridge University Press, Cambridge (2023). https://doi.org/10.1017/9781009303538

    work page doi:10.1017/9781009303538 2023

  32. [32]

    1: Domain distribution and temporal dynamics.(a)Distribution of 692 papers across five thematic domains.(b)Annual publication volume by domain (2015–2026)

    Henrich, J., Heine, S.J., Norenzayan, A.: The weirdest people in the world? Behavioral and Brain Sciences33(2-3), 61–83 (2010) https://doi.org/10.1017/ s0140525x0999152x 19 (a) (b) Fig. 1: Domain distribution and temporal dynamics.(a)Distribution of 692 papers across five thematic domains.(b)Annual publication volume by domain (2015–2026). The post-2022 s...

    2010