Pith. sign in

REVIEW 3 major objections 5 minor 42 references

Bias Mitigation for AI-Feedback Loops in Recommender Systems: A Systematic Literature Review and Taxonomy

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A systematic review maps 24 studies that test bias mitigation inside recommender feedback loops

desk verdict A competent SLR with a genuinely useful taxonomy, but its headline claim about 'only six studies' is contradicted by its own Table 2's seven 'Both' entries—fix that before relying on the synthesis. read the letter →

arxiv 2509.00109 v1 pith:44B7VFQT submitted 2025-08-28 cs.IR cs.LG

Theodor Stoecker , Samed Bayer , Ingo Weber This is my paper · ORCID
classification cs.IRcs.LG
keywords systematicliteraturereviewbiasmitigationrecommendersystemsAIfeedbackloopsamplificationtaxonomyfairnessevaluationdynamic
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

This paper tries to establish a reliable map of what is actually known about correcting recommender-system bias when the system keeps learning from its own outputs. It argues that most bias-mitigation research is tested on static data and therefore says little about long-term fairness, while 24 studies do test mitigation under repeated retraining in simulations or live A/B tests. If the map is right, practitioners gain a six-dimensional checklist for choosing mitigation methods and researchers gain a list of the field's most urgent gaps, most notably the scarcity of shared simulators and the fact that only six studies measure fairness and performance together.

What carries the argument

The carrying object is the six-dimensional taxonomy, built with the iterative conceptual-to-empirical method of [30]. It starts from the pre-, in-, and post-processing pipeline stages of [7] and adds recommender-specific classes such as Causal Inference-based, Learning Approach, Learning Problem, and Add-On. The taxonomy's work is to make 24 heterogeneous studies comparable on a common grid and to expose where the field is thin. The central phenomenon it organizes is the ML Model feedback loop, in which a system retrains on the very instances its own predictions caused to be observed, such as only recommended items receiving user feedback.

What would settle it

Run the same inclusion criteria over additional bibliographic databases and include extended abstracts and posters; if this yields more than a handful of additional studies that test bias mitigation under multi-round retraining and report both fairness and performance, the claimed scarcity of 6 out of 24 would be shown to be a search artifact.

Watch

Extended reading notes

Core claim

The central claim is a field-level map: screening 347 records yields 24 primary studies published between 2019 and 2025 that explicitly address ML Model feedback loops and evaluate bias mitigation in multi-round simulation or live A/B tests. The paper organizes these studies along six dimensions—mitigation technique, bias addressed, dynamic testing setup, evaluation focus, application domain, and ML task—and reports that 17 of 24 interventions are in-processing, that A/B tests mostly use performance metrics, and that only six studies report both fairness and performance. This is presented as evidence that the field lacks shared simulators, standardized metrics, and combined fairness-and-performance evaluation.

Load-bearing premise

The map's validity depends on the literature search being complete: if relevant studies live in databases, formats, or venues the search did not cover, including other databases and excluded extended abstracts and posters, then the reported gaps could be artifacts of the search rather than features of the field.

Editorial extensions

If this is right

  • If the taxonomy holds, practitioners can classify any new feedback-loop-aware mitigation study quickly and compare it against the 24 existing studies on six dimensions.
  • The 24 studies become a baseline: claims that a bias mitigation method works under feedback loops should be tested in simulations or A/B tests with multiple retraining rounds, not on static splits.
  • The dominance of in-processing interventions (17 of 24) suggests that future mitigation work will likely concentrate on learning algorithms, loss functions, and reward design rather than on data or output fixes.
  • The absence of shared simulators, if real, means cross-study comparability is currently limited, and the taxonomy can serve as a specification for the shared benchmarking environment the field lacks.
  • Because only six studies report both fairness and performance, current dynamic evaluations do not yet show whether fairness gains come at an acceptable performance cost.

Reading between the lines

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

  • A natural extension beyond the paper would be to use the six studies that report both fairness and performance as seeds for a standardized evaluation protocol, fixing the number of simulation rounds and the metric set so future mitigation claims can be compared directly.
  • The fact that A/B tests almost always report performance metrics suggests that live industry settings monitor fairness informally or not at all; if that is right, the next bottleneck is measurement practice, not mitigation algorithms.
  • Since the paper notes that applying the 90% static-evaluation estimate from [25] to other biases would exclude most studies, the field's apparent shortfall could shrink as evaluation practice improves; the taxonomy would then need richer classes for non-simulation validation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This manuscript presents a systematic literature review (SLR) of bias mitigation techniques for recommender systems that explicitly account for AI feedback loops and are evaluated in multi-round simulations or live A/B tests. The authors screen 347 records from ACM Digital Library, IEEE Xplore, and arXiv, supplemented by papers from Pagan et al.'s feedback-loop classification, and retain 24 primary studies published between 2019 and 2025. Each study is coded on six dimensions: mitigation technique, biases addressed, dynamic testing set-up, evaluation focus, application domain, and ML task, resulting in a proposed taxonomy (Figure 5 and Table 2). The paper reports that in-processing methods dominate, that A/B tests tend to use performance-only metrics, and that a small number of studies evaluate both fairness and performance. The authors also discuss limitations, including the restricted database choice and the absence of widely used simulation benchmarks.

Significance. If the results are accepted after the inconsistencies below are resolved, the review would make a useful contribution by consolidating a small but growing literature on feedback-loop-aware bias mitigation. Its strengths include a transparent reporting of search strings and access dates, a two-stage screening procedure with a second evaluator, a detailed coding table (Table 2) that allows independent checking of the taxonomy mapping, and the use of an established taxonomy-development method (Nickerson et al.). The paper's descriptive findings, such as the dominance of in-processing interventions and the rarity of studies reporting both fairness and performance, are directly informative for both researchers and practitioners. The internal inconsistencies identified below, however, currently undermine the reliability of the headline numbers and the reproducibility of the selection process, so the contribution is best viewed as provisional until those points are corrected.

major comments (3)
  1. [Abstract and §4.6 / Table 2] The abstract states that 'only six' of the 24 studies report both fairness and performance, but this contradicts the paper's own coding. In Table 2, the Evaluation Focus column is 'Both' for seven studies: [18], [11], [6], [13], [10], [23], and [33]. Section 4.6 reports that 29% of studies investigated both, and 7/24 = 29.2%, which matches Table 2 but not the abstract's 'six'. Since this figure is used as evidence of a research gap and appears in the headline contribution, the abstract must be corrected to 'seven' or the table and percentage must be revised to be consistent with a count of six.
  2. [§3 and Figure 2] The screening counts reported in the text are not consistent with the flowchart. The text says the first screening 'led to 150 papers', the LLM screening 'identified 158 papers' with '93 overlapping with our initial set', and 'This yielded 216 non-duplicate papers'. However, 150 + 158 - 93 = 215, not 216, and the numbers surviving screening stage 1 in Figure 2 are 202 (ACM), 8 (IEEE), 2 (arXiv), and 4 (Pagan et al.), which sum to 216. The relationship among the 150 human-screened papers, the 158 LLM-identified papers, and the 216 papers in Figure 2 is not explained. Please reconcile these numbers or provide a detailed breakdown so the study-selection arithmetic can be independently checked.
  3. [§3, Figure 2, Table 2] The inclusion criteria are applied inconsistently with respect to preprints. The text states that 'Only research papers from conferences, workshops, and journals were considered; extended abstracts, and posters were excluded', yet later an arXiv search is added 'to also include the most up-to-date research', Figure 2 explicitly labels 'Pre-Prints and studies included through other methods', and Table 2 includes arXiv preprints such as [23] and [33]. This contradiction affects the systematic nature of the review; please clarify whether preprints are within scope, and if so, adjust the stated inclusion criteria and discuss the implications for the review's conclusions.
minor comments (5)
  1. [Table 2 caption] The caption says the table maps papers 'across our seven core dimensions', but the paper consistently describes a six-dimensional taxonomy; the Paper identifier column is not a taxonomy dimension. Please change 'seven' to 'six'.
  2. [§4.6] The percentages (54%, 17%, 29%) correspond to 13, 4, and 7 studies, respectively. Reporting the raw counts alongside the percentages would prevent the ambiguity that led to the abstract's 'six' versus 'seven' discrepancy.
  3. [Figure 5] The figure label 'liDimensions' appears to be a typographical artifact and should read 'Dimensions'.
  4. [§3] The sentence 'This led to 150 papers' is ambiguous because it is not clear whether this number refers only to the human screening of the ACM database or to the combined screening of all sources; please clarify.
  5. [§2.1 and Figure 1] The mapping in Figure 1 shows four bias types, but the text says the taxonomy focuses on four of Suresh and Guttag's classifications; the relationship between Figure 1 and the 'Depends on bias' category in Table 2 could be stated more explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the review synthesizes external studies, and the taxonomy is a descriptive classification, not a derived prediction.

full rationale

This is a systematic literature review and taxonomy, not a derivation of a predictive result from first principles. The 24 primary studies are external works; the selection procedure (Section 3) and coding (Table 2) are descriptive. The taxonomy is built iteratively with Nickerson et al.'s method from the coded studies and prior classifications; there is no fitted parameter renamed as a prediction. The only self-citation, [2] (Bass, Lu, Weber, Zhu), appears in generic statements about ML models learning from outputs and is not load-bearing for the review's central taxonomy or gap claims. Other cited frameworks (Pagan et al., Suresh & Guttag, Chen et al., Caton & Haas) are external and are used as classification scaffolding, not as proof of the findings. The closest issue is an internal consistency discrepancy—the abstract says 'only six' studies report both fairness and performance, while Table 2 codes seven Evaluation Focus 'Both' entries and Section 4.6 reports 29% (7/24)—but this is a data-coding/consistency problem, not a circular step. No quantity claimed as a finding is defined in terms of itself, and no analysis result is statistically forced by the input criteria.

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

This is a literature review; no free parameters are fit and no new entities are postulated. The taxonomy rests on prior bias and feedback-loop frameworks (Suresh-Guttag, Chen et al., Pagan et al.) and on methodological choices in the screening process, which are listed as axioms.

assumptions (4)
  • domain assumption Bias taxonomies of Suresh and Guttag (seven types) and Chen et al. (seven RS bias classes) are valid and adequate for classifying feedback-loop biases.
    The paper maps every reported bias to these frameworks (Section 4.4 and Figure 4) and uses this mapping as the bias-type dimension of the taxonomy; if these frameworks are wrong or incomplete, the taxonomy's bias dimension inherits the error.
  • domain assumption Pagan et al.'s five-way classification of feedback loops, and the focus on ML Model feedback loops, captures the relevant loop types.
    Section 2.2 adopts this classification and restricts the SLR to ML Model feedback loops; if the classification omits other feedback loops relevant to bias mitigation, the survey's scope is misaligned with its claim of covering AI feedback loops generally.
  • ad hoc to paper LLM-assisted screening (gpt-4o-mini) used for the ACM database produces a reliable superset of relevant papers.
    Section 3: 'we utilised a large language model ... which identified 158 papers in this screening step (93 overlapping with our initial set).' The reliability of this step is not evaluated; the final decision was human, but the LLM-assessed set bounds what humans saw.
  • domain assumption Requiring dynamic testing (multi-round simulation or live A/B) is a necessary and sufficient criterion for evaluating bias mitigation under AI feedback loops.
    This is the paper's core inclusion criterion (Section 3); if static evaluation can sometimes capture feedback-loop effects, the review excludes relevant evidence by construction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bias Mitigation for AI-Feedback Loops in Recommender Systems: A Systematic Literature Review and Taxonomy." pith.science (2026). https://pith.science/paper/44B7VFQT

@misc{pith2026250900109,
  author       = {Pith},
  title        = {Pith review of: Bias Mitigation for AI-Feedback Loops in Recommender Systems: A Systematic Literature Review and Taxonomy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/44B7VFQT}},
  note         = {Machine review of arXiv:2509.00109}
}
read the original abstract

Recommender systems continually retrain on user reactions to their own predictions, creating AI feedback loops that amplify biases and diminish fairness over time. Despite this well-known risk, most bias mitigation techniques are tested only on static splits, so their long-term fairness across multiple retraining rounds remains unclear. We therefore present a systematic literature review of bias mitigation methods that explicitly consider AI feedback loops and are validated in multi-round simulations or live A/B tests. Screening 347 papers yields 24 primary studies published between 2019-2025. Each study is coded on six dimensions: mitigation technique, biases addressed, dynamic testing set-up, evaluation focus, application domain, and ML task, organising them into a reusable taxonomy. The taxonomy offers industry practitioners a quick checklist for selecting robust methods and gives researchers a clear roadmap to the field's most urgent gaps. Examples include the shortage of shared simulators, varying evaluation metrics, and the fact that most studies report either fairness or performance; only six use both.

Figures

Figures reproduced from arXiv: 2509.00109 by the authors.

Figure 1
Figure 1. Offline evaluation detects bias in static data, whereas online (or simulated) evaluation reveals how feedback loops [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Flowchart of study selection. From 347 records iden￾tified (ACM 318; IEEE Xplore 13; ArXiv 11; Pagan et al. 5), titles and abstracts were first screened for AI-feedback loops and bias mitigation; full texts were then assessed for applied mitigation strategies tested via simulation or A/B testing, yielding 24 studies. Our procedure to select relevant papers as outlined in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Number of included studies per publication venue [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Biases mentioned and their mapping to Suresh and Guttag’s framework. Data and algorithmic bias depend on the [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Taxonomy for recommender systems bias mitigation evaluated in dynamic environments based on six dimensions. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Metrics used to evaluate a mitigation type in sim [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 12 canonical work pages

  1. [18]

    Cuize Han, Pablo Castells, Parth Gupta, Xu Xu, and Vamsi Salaka. 2022. Ad- dressing Cold Start in Product Search via Empirical Bayes. en. InProceedings of the 31st ACM International Conference on Information & Knowledge Management. ACM. doi: 10.1145/3511808.3557066

    arXiv 2022

  2. [11]

    Khalil Damak, Sami Khenissi, and Olfa Nasraoui. 2022. Debiasing the Cloze Task in Sequential Recommendation with Bidirectional Transformers. en. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. ACM. doi: 10.1145/3534678.3539430

    arXiv 2022

  3. [6]

    Gökhan Çapan, İlker Gündoğdu, Ali Caner Türkmen, and Ali Taylan Cemgil

    work page

  4. [10]

    Minmin Chen, Alex Beutel, Paul Covington, Sagar Jain, Francois Belletti, and Ed H. Chi. 2019. Top-K Off-Policy Correction for a REINFORCE Recommender System. en. In Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining . ACM. doi: 10.1145/3289600.3290999

    arXiv 2019

  5. [23]

    Sami Khenissi and Olfa Nasraoui. 2020. Modeling and counteracting exposure bias in recommender systems. (2020). https://arxiv.org/abs/2001.04832 arXiv: 2001.04832 [cs.IR]

    arXiv 2020

  6. [33]

    Yining She, Sumon Biswas, Christian Kästner, and Eunsuk Kang. 2025. FairSense: Long-Term Fairness Analysis of ML-Enabled Systems. en. (2025). doi: 10.48550 /arXiv.2501.01665

    arXiv doi:10.48550/arxiv.2501.01665 2025

  7. [1]

    Nil-Jana Akpinar, Cyrus DiCiccio, Preetam Nandy, and Kinjal Basu. 2022. Long- term Dynamics of Fairness Intervention in Connection Recommender Systems. en. In Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society . ACM. doi: 10.1145/3514094.3534173

    arXiv 2022

  8. [2]

    Len Bass, Qinghua Lu, Ingo Weber, and Liming Zhu. 2025. Engineering AI Systems: Architecture and DevOps Essentials . Addison-Wesley Professional

    work page 2025

Show all 42 references
  1. [3]

    Christine Bauer, Eva Zangerle, and Alan Said. 2024. Exploring the landscape of recommender systems evaluation: practices and perspectives. en. ACM Transactions on Recommender Systems , 2, 1. doi: 10.1145/3629170

    2024 doi

  2. [4]

    Christopher M Bishop. 2006. Pattern recognition and machine learning . Num- ber 4. Vol. 4. Springer

    2006

  3. [5]

    Jennifer Brennan, Yahu Cong, Yiwei Yu, Lina Lin, Yajun Peng, Changping Meng, Ningren Han, Jean Pouget-Abadie, and David M. Holtz. 2025. Reducing Symbiosis Bias through Better A/B Tests of Recommendation Algorithms. en. In Proceedings of the ACM on Web Conference 2025 . ACM. do...

    2025 doi

  4. [7]

    Simon Caton and Christian Haas. 2024. Fairness in machine learning: a survey. en. ACM Comput. Surv., 56, 7. doi: 10.1145/3616865

    2024 doi

  5. [8]

    Bo Chang et al. 2024. Cluster Anchor Regularization to Alleviate Popularity Bias in Recommender Systems. en. In Companion Proceedings of the ACM Web Conference 2024. ACM. doi: 10.1145/3589335.3648312

    2024

  6. [9]

    Jiawei Chen, Hande Dong, Xiang Wang, Fuli Feng, Meng Wang, and Xiang- nan He. 2023. Bias and debias in recommender system: a survey and future directions. en. ACM Trans. Inf. Syst., 41, 3. doi: 10.1145/3564284

    2023 doi

  7. [12]

    Ekstrand, and Christine Bauer

    Andres Ferraro, Michael D. Ekstrand, and Christine Bauer. 2024. It’s Not You, It’s Me: The Impact of Choice Models and Ranking Strategies on Gender Imbal- ance in Music Recommendation. en. In 18th ACM Conference on Recommender Systems. ACM. doi: 10.1145/3640457.3688163

    2024

  8. [14]

    Dalin Guo, Sofia Ira Ktena, Pranay Kumar Myana, Ferenc Huszar, Wenzhe Shi, Alykhan Tejani, Michael Kneier, and Sourav Das. 2020. Deep Bayesian Bandits: Exploring in Online Personalized Recommendations. en. In Fourteenth ACM Conference on Recommender Systems . ACM. doi: 10.1145...

    2020

  9. [15]

    Huifeng Guo, Jinkai Yu, Qing Liu, Ruiming Tang, and Yuzhou Zhang. 2019. PAL: a position-bias aware learning framework for CTR prediction in live recom- mender systems. en. InProceedings of the 13th ACM Conference on Recommender Systems. ACM. doi: 10.1145/3298689.3347033

    2019

  10. [16]

    Huizhong Guo, Zhu Sun, Dongxia Wang, Tianjun Wei, Jinfeng Li, and Jie Zhang. 2025. Enhancing new-item fairness in dynamic recommender systems. en. arXiv preprint arXiv:2504.21362. (2025). doi: 10.48550/arXiv.2504.21362

    2025 doi arXiv

  11. [17]

    Priyanka Gupta, Ankit Sharma, Pankaj Malhotra, Lovekesh Vig, and Gautam Shroff. 2021. CauSeR: Causal Session-based Recommendations for Handling Popularity Bias. en. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management . ACM. doi: 10.114...

    2021

  12. [19]

    Zhang, Mark Harman, and Federica Sarro

    Max Hort, Zhenpeng Chen, Jie M. Zhang, Mark Harman, and Federica Sarro

  13. [20]

    2025.AI Engineering: Building Applications with Foundation Models

    Chip Huyen. 2025.AI Engineering: Building Applications with Foundation Models. O’Reilly Media

    2025

  14. [21]

    Olivier Jeunen and Bart Goethals. 2023. Pessimistic Decision-Making for Rec- ommender Systems. en. ACM Transactions on Recommender Systems , 1. doi: 10.1145/3568029

    2023 doi

  15. [22]

    Olivier Jeunen and Bart Goethals. 2021. Pessimistic reward models for off-policy learning in recommendation. en. In Proceedings of the 15th ACM Conference on Recommender Systems. Association for Computing Machinery. doi: 10.1145/34 60231.3474247

    2021

  16. [24]

    Anastasiia Klimashevskaia, Mehdi Elahi, Dietmar Jannach, Lars Skjærven, Astrid Tessem, and Christoph Trattner. 2023. Evaluating The Effects of Cali- brated Popularity Bias Mitigation: A Field Study. en. In Proceedings of the 17th ACM Conference on Recommender Systems. ACM. doi...

    2023

  17. [25]

    Anastasiia Klimashevskaia, Dietmar Jannach, Mehdi Elahi, and Christoph Trattner. 2024. A survey on popularity bias in recommender systems. en. User Modeling and User-Adapted Interaction, 34, 5. doi: 10.1007/s11257-024-09406-0

    2024 doi

  18. [26]

    Yongkang Li, Xingyu Zhu, Yuheng Wu, Wenxu Zhao, and Xiaona Xia. 2025. A survey on causal inference-driven data bias optimization in recommendation systems: principles, opportunities and challenges. en. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery , 15, ...

    2025 doi

  19. [27]

    Xiao Lin, Xiaokai Chen, Linfeng Song, Jingwei Liu, Biao Li, and Peng Jiang. 2023. Tree based Progressive Regression Model for Watch-Time Prediction in Short- video Recommendation. en. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . ACM...

    2023

  20. [28]

    Jiaqi Ma, Zhe Zhao, Xinyang Yi, Ji Yang, Minmin Chen, Jiaxi Tang, Lichan Hong, and Ed H. Chi. 2020. Off-policy Learning in Two-stage Recommender Systems. en. In Proceedings of The Web Conference 2020 . ACM. doi: 10.1145/336 6423.3380130

    2020

  21. [29]

    Masoud Mansoury, Himan Abdollahpouri, Mykola Pechenizkiy, Bamshad Mobasher, and Robin Burke. 2020. Feedback loop and bias amplification in recommender systems. en. In Proceedings of the 29th ACM International Con- ference on Information & Knowledge Management . Association for...

    2020

  22. [30]

    Robert C Nickerson, Upkar Varshney, and Jan Muntermann. 2013. A method for taxonomy development and its application in information systems. en. European Journal of Information Systems , 22, 3. doi: 10.1057/ejis.2012.26

    2013 doi

  23. [31]

    Nicolò Pagan, Joachim Baumann, Ezzat Elokda, Giulia De Pasquale, Saverio Bolognani, and Anikó Hannák. 2023. A classification of feedback loops and their relation to biases in automated decision-making systems. en. In Proceedings of the 3rd ACM Conference on Equity and Access i...

    2023

  24. [32]

    Yi Ren, Hongyan Tang, and Siwen Zhu. 2022. Unbiased Learning to Rank with Biased Continuous Feedback. en. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management . ACM. doi: 10.1145/3511 808.3557483

    2022 doi

  25. [34]

    Chi, and Minmin Chen

    Yi Su, Haokai Lu, Yuening Li, Liang Liu, Shuchao Bi, Ed H. Chi, and Minmin Chen. 2024. Multi-Task Neural Linear Bandit for Exploration in Recommender Systems. en. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . ACM. doi: 10.1145/363752...

    2024

  26. [35]

    Harini Suresh and John Guttag. 2021. A framework for understanding sources of harm throughout the machine learning life cycle. en. In Proceedings of the 1st ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization. Association for Computing Machinery. do...

    2021 doi

  27. [36]

    Wei Tang, Chien-Ju Ho, and Yang Liu. 2021. Bandit Learning with Delayed Impact of Actions. en. doi: 10.5555/3540261.3542314

    2021

  28. [37]

    Xiangmeng Wang, Qian Li, Dianer Yu, and Guandong Xu. 2022. Off-policy Learning over Heterogeneous Information for Recommendation. en. In Pro- ceedings of the ACM Web Conference 2022 . ACM. doi: 10.1145/3485447.3512072

    2022

  29. [38]

    Tao Wu et al. 2020. Zero-Shot Heterogeneous Transfer Learning from Recom- mender Systems to Cold-Start Search Retrieval. en. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management . ACM. doi: 10.1145/3340531.3412752

    2020

  30. [39]

    Hyunsik Yoo et al. 2024. Ensuring user-side fairness in dynamic recommender systems. en. In Proceedings of the ACM Web Conference 2024 . Association for Computing Machinery. doi: 10.1145/3589334.3645536

    2024

  31. [40]

    Yang Zhang, Fuli Feng, Xiangnan He, Tianxin Wei, Chonggang Song, Guohui Ling, and Yongdong Zhang. 2021. Causal intervention for leveraging popularity bias in recommendation. en. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Inform...

    2021

  32. [41]

    Yuqi Zhou, Sunhao Dai, Liang Pang, Gang Wang, Zhenhua Dong, Jun Xu, and Ji-Rong Wen. 2025. Exploring the escalation of source bias in user, data, and recommender system feedback loop. en. arXiv preprint arXiv:2405.17998. (2025). doi: 10.48550/arXiv.2405.17998

    2025 doi arXiv

  33. [2022]

    Dirichlet–Luce choice model for learning from interactions. en. User Modeling and User-Adapted Interaction, 32, 4. doi: 10.1007/s11257-022-09331-0

    doi

  34. [2024]

    Bias mitigation for machine learning classifiers: a comprehensive survey. en. ACM J. Responsib. Comput., 1, 2. doi: 10.1145/3631326

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.