Pith. sign in

REVIEW 3 major objections 6 minor 66 references

You Don't Bring Me Flowers: Mitigating Unwanted Recommendations Through Conformal Risk Control

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a post-hoc threshold-and-replace filter can provably keep the expected fraction of flagged, unwanted items in a recommendation list below a user-chosen level, without retraining the recommender.

desk verdict Useful post-hoc method for bounding unwanted content, but the proof of monotonicity for the replacement rule is wrong as written; the advertised guarantee requires the monotonized risk the paper only mentions in a footnote. read the letter →

arxiv 2507.16829 v1 pith:3BIPMOT7 submitted 2025-07-09 cs.IR cs.AI

Giovanni De Toni , Erasmo Purificato , Emilia Gómez , Bruno Lepri , Andrea Passerini , Cristian Consonni This is my paper · ORCID
classification cs.IRcs.AI
keywords conformalriskcontrolrecommendersystemsunwantedcontentbinarynegativefeedbackrepeatconsumptiondistribution-freeguaranteespost-hocfilteringKuaiRanddataset
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 sets out to show that a recommender system can be made safer after training, with no retraining, by post-processing the candidate list with a user-chosen risk level α. The method uses conformal risk control — a distribution-free technique that selects a threshold so that expected loss stays below a bound — to keep the expected fraction of flagged, unwanted items in the top-k list below α, using only binary negative feedback such as a "not interested" click. Because pure thresholding can shrink the list below k items, the paper adds replacements: previously seen items the user did not flag and watched past a watch-time threshold β, which are treated as safe to resurface. Experiments on video-sharing platform data show the procedure keeps empirical risk at or below the target across several ranking models, while preserving more ranking quality than simply removing items. If correct, the method gives users a simple, provable control over unwanted content exposure in any personalized recommender that can supply historical feedback.

What carries the argument

The load-bearing object is the risk function of Equation (5), the fraction of flagged items in a recommendation set, paired with the conformal risk control threshold selection of Theorem 1: choose the smallest threshold λ̂ such that (n/(n+1))R̂(λ) + 1/(n+1) ≤ α on a calibration set of size n, and the expected risk at λ̂ is provably at most α whenever the risk is non-increasing and right-continuous in λ. The replacement mechanism is the safe-item set T_safe = {previously seen, unflagged items with watch-time proportion C > β}, which is unioned with the thresholded pool so the recommender can still return up to k items. The safety of this union rests on Property 1, which asserts zero probability of a second report for items with H1st = 0 and C > β; the paper checks this approximately on Kuaishou watch-time data and explores the trade-off through the β ablation.

What would settle it

Run Algorithm 1 with the Replace strategy on held-out data containing second exposures, and count how often a previously seen, never-reported item with first-exposure watch time above β is later flagged; if the rate is nonzero and the empirical risk R_H(S_λ(U,k)) exceeds the promised α for the chosen β, the guarantee is falsified. The paper's own no-filtering ablation in Fig. 5a already shows the risk overshoot when β filtering is removed.

Watch

Extended reading notes

Core claim

The central claim is that the fraction of unwanted content in a personalized top-k recommendation list can be provably bounded in expectation by a post-hoc threshold-and-replace procedure. For any user-selected α, Algorithm 1 computes a threshold λ̂ from a held-out calibration set using the conformal risk control result of Theorem 1, so that the expected value of the risk R_H(S_λ(U,k)) — the fraction of flagged items in the returned set — is at most α, provided the risk is non-increasing in λ. To avoid returning fewer than k items, candidates below the threshold are replaced by items from the user's own history that were not flagged and whose first-exposure watch time exceeds a threshold β; Property 1 asserts that such items have zero probability of being flagged on a second exposure, which keeps the risk monotone and lets the guarantee survive replacement. The paper tests this by wrapping several pretrained rankers with the postprocessor on KuaiRand data and reports that the empirical reduction in unwanted content meets or exceeds the target, with nDCG and recall degrading more gracefully than under plain removal.

Load-bearing premise

The entire argument hinges on the assumption that an item a user has already seen, never reported, and watched past a watch-time threshold will not be reported if it is shown again.

Editorial extensions

If this is right

  • Any deployed recommender can expose a user to at most an expected α fraction of flagged items in the top-k set without retraining, as long as historical binary feedback and calibration data are available.
  • With a sufficient pool of safe repeated items, the user receives a full top-k list; when the pool is empty, the method degrades to pure removal and may return fewer than k items.
  • The number of items that must be replaced depends on the score function used for thresholding; on KuaiRand, sign-aware rankers replace more items than unsigned ones at equal risk.
  • Stricter safety filtering (larger β) restores the guarantee under distribution shift but shrinks the replacement pool, moving behavior closer to plain removal; no filtering can break the guarantee.

Reading between the lines

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

  • As an extension beyond the paper, per-user or per-group calibration with the same Theorem 1 machinery would likely reduce the conservativeness the authors observe for low-reporting users.
  • A further extension would replace the hard watch-time threshold β with a per-user calibrated quantile or a learned second-exposure report probability, since the data only support Property 1 approximately.
  • The reliance on repeated items suggests a testable boundary: on platforms where repeat exposure is rare, the method collapses to pure removal, so the promised full-list guarantee would not hold without a larger safe-item pool.
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 / 6 minor

Summary. The paper proposes a post-hoc, model-agnostic method, Algorithm 1, that uses conformal risk control to bound the expected fraction of user-flagged ('unwanted') items in a top-k recommendation list. The method thresholds items by a score and then expands the candidate pool with previously consumed, non-flagged items whose watch time exceeds a threshold β, so as to preserve list size. The authors claim a distribution-free, finite-sample guarantee on the fraction of flagged items (Theorem 1 applied after Proposition 3), and they evaluate the approach on the KuaiRand dataset with four ranking models, comparing a 'Replace' strategy against a 'Remove' strategy and ablating the β threshold and user reporting habits.

Significance. The problem is timely and practically important: giving users a provable handle on unwanted content in recommendations is a valuable goal, and a simple, model-agnostic post-processing layer would be a useful contribution. The empirical study is careful, includes multiple ranking models, ablations, and a user-group analysis, and the authors release source code, which strengthens reproducibility. If the theoretical guarantee were correctly established, the method would be a strong practical tool. However, the central guarantee is currently not proven: the monotonicity assumption required by conformal risk control is violated by the proposed replacement set, and the paper's own ablation (Fig. 5a, 'None') shows the guarantee can fail. The contribution is therefore conditional on a substantive revision of the theoretical argument.

major comments (3)
  1. [Section 5.1, Proposition 3 and Section 6] Proposition 3 is false as stated: the risk fraction in Eq. (5) is not non-increasing in λ for the set T_replace_λ defined in Eq. (8). Counterexample: let T_safe = {S} with H(S)=0, and let T_λ contain a flagged item F (score 10) and an unflagged item A (score 9). For λ=5, T_replace_λ = {F,A,S}, so R_H = 1/3; for λ=9.5, T_replace_λ = {F,S}, so R_H = 1/2. The risk increases when an unflagged low-score item is removed, because the numerator and denominator both shrink. Consequently, the assumption in Theorem 1 that R_H(S_λ(U,k)) is non-increasing in λ is not satisfied, and the expectation bound E[R_H(S_λ̂(U,k))] ≤ α does not follow for Algorithm 1 as written. The footnote in Sec. 4.2 about monotonizing R_H is not implemented: Algorithm 1 and Eq. (6) use the raw risk, and no monotone upper envelope is defined, computed, or used in the experiments. Therefore the Section 6 statement that 'the algorithm provably controls the fraction of unwanted content in the final recommendation list (cf. Eq. (5))' is not supported by the proof given.
  2. [Section 5, Proposition 2] The proof of Proposition 2 is incomplete because it does not specify which items are flagged. If the flagged item is D (score 5), then for k=1 the top-1 set has risk 1; but if the flagged item is A (score 1), then for k=1 the set {D} has risk 0, and the claim that only λ > 5 (empty set) achieves risk ≤ 0.1 is false. The proof needs to state the flagged status of each item and show that the impossibility holds for a fixed flagged item regardless of k, or else the proposition as stated is not established.
  3. [Section 7, RQ4 and Fig. 5a] The ablation with 'None' (no watch-time filtering) shows that the empirical reduction in unwanted content falls below the target, meaning risk control fails. The paper attributes this to distribution shift, but this is also exactly the regime where Property 1 is violated and where the monotonicity needed for Theorem 1 is not guaranteed. The paper should explicitly state whether the theoretical guarantee is conditional on Property 1 holding exactly. If so, the experimental results for β>0 demonstrate behavior under an approximate satisfaction of the property, and the conditions under which the guarantee holds should be stated precisely, including what happens when Property 1 is violated.
minor comments (6)
  1. [Section 4.2] The sentence 'by removing items, we cannot increase R_H(S_λ(U,k))' is incorrect for the fraction in Eq. (5); removing an unflagged item can increase the fraction. Please rephrase or qualify this statement.
  2. [Algorithm 1] The item set is denoted I in the text and Y in lines 2-3 of Algorithm 1; please unify the notation.
  3. [Section 5.1 and Algorithm 1] Line 3 of Algorithm 1 uses the condition W%(U,i') > β, while Eq. (7) defines the safe pool via C(I=i') > β; the correspondence between C and W% should be made explicit.
  4. [Figure 3 caption] The caption says 'nDGC @ 20'; this should be 'nDCG @ 20'.
  5. [Section 3.3, Table 2] The report/no-report symbols in Table 2 may not render correctly in all formats; consider spelling out the four behavior combinations in text.
  6. [Section 4.1] The phrase 'the set of items I ∈ {i1,...,iN}' should be 'the set of items I = {i1,...,iN}'.

Circularity Check

1 steps flagged · score 2.0 of 10

Main conformal-guarantee chain is independent of the paper's own outputs; minor circularity enters in choosing the replacement-safety threshold β from the same Kuaishou data later used to demonstrate risk control.

  1. fitted input called prediction [Section 5.1, Property 1 / Eq. (7) / Fig. 2; Section 7 RQ4, Fig. 5a]
    "In the case of Kuaishou, we can pick as C the watch time of videos. Then, we can simply pick videos on which the users have spent time W%(U, I) above a certain threshold. For example, Fig. 2 shows the distribution of P (H = 1 | H1st = 0, W%). Indeed, the plot on the left shows that the videos that are reported in the second view have a much shorter watch time. Moreover, the right plot of Fig. 2 shows that by filtering repeated videos with a thresholding β, we can reduce the likelihood of picking a harmful video almost to zero, globally."

    The algorithm's replacement pool (Eq. (7)) is T(safe) = {i' : H(U,i') = 0 ∧ C(i') > β}, and Property 1 is the load-bearing assumption that such items are never flagged on second exposure. The paper selects β from the empirical Kuaishou conditional probability P(H = 1 | H1st = 0, W% > β) shown in Fig. 2, using the same dataset on which the later experiments measure the flagged fraction RH. Thus the empirical claim that β-filtering 'restores risk control' (Fig. 5a, in contrast to the failing 'None' curve) is partly a consequence of choosing the threshold on the outcome it is used to certify, rather than an independent prediction. This is a minor circularity in the empirical validation.

full rationale

The central derivation is not circular. Section 4.2 imports Theorem 1 directly from Angelopoulos et al. (2023), an external result; Eq. (6) is exactly that theorem's threshold rule, and Eq. (5) is plugged into it as the risk function. No step redefines the target α as an output of fitting, and the finite-sample guarantee would hold for any scorer if the theorem's monotonicity condition were met. The replacement idea (Eqs. (7)–(8)) is the paper's own contribution, and Proposition 3 is an attempted verification of the condition, not a self-citation or a renaming of a known result. The one circular element is the practical choice of β: Property 1 is stated as an existence assumption, but Section 5.1 justifies it by inspecting the very Kuaishou data used later for evaluation, selecting the watch-time threshold that drives the conditional probability to near zero. Fig. 5a then treats the resulting risk reduction as evidence that the method works. That is a fitted input called a prediction in the empirical section, which the formal theory does not fully inoculate because the theory is conditional on Property 1. I therefore set score 2, not higher: the CRC theorem and its proof chain remain independent, and the paper includes an honest ablation showing that without β the guarantee fails. Separately, the reviewer's monotonicity objection is a correctness risk, not a circularity: the claim that removing items 'cannot increase RH' is suspect for the proportion in Eq. (5), and the footnote's proposed monotonization is not implemented in Algorithm 1; this does not enter the circularity score.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The paper's theoretical core rests on the imported conformal risk control theorem plus the domain assumption Property 1, which is only validated empirically on one dataset. The β threshold is a hand-chosen free parameter, and the empirical claim of risk control is sensitive to it.

free parameters (1)
  • β (watch-time threshold for safe replacements) = 0%, 50%, 100% in ablations (and 'None')
    Chosen by the authors via exploratory analysis of the same KuaiRand data (Fig. 2) and evaluated with ablations; it controls which previously seen items are considered safe replacements and whether Property 1 approximately holds.
assumptions (3)
  • standard math Conformal risk control theorem (Theorem 1 of Angelopoulos et al. 2023)
    Imported as a black-box result; the paper states it as Theorem 1 and relies on it for the threshold choice and the expectation guarantee.
  • domain assumption Property 1: P(H=1 | H1st=0, C>β)=0 for some C and β
    Needed for the monotonicity of the risk over the replaced set (Proposition 3); only approximately satisfied in the data.
  • domain assumption Exchangeability of calibration and test data
    Required for the conformal risk control guarantee; the paper relies on random splitting of single interactions and acknowledges distribution shift as a limitation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of You Don't Bring Me Flowers: Mitigating Unwanted Recommendations Through Conformal Risk Control." pith.science (2026). https://pith.science/paper/3BIPMOT7

@misc{pith2026250716829,
  author       = {Pith},
  title        = {Pith review of: You Don't Bring Me Flowers: Mitigating Unwanted Recommendations Through Conformal Risk Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3BIPMOT7}},
  note         = {Machine review of arXiv:2507.16829}
}
read the original abstract

Recommenders are significantly shaping online information consumption. While effective at personalizing content, these systems increasingly face criticism for propagating irrelevant, unwanted, and even harmful recommendations. Such content degrades user satisfaction and contributes to significant societal issues, including misinformation, radicalization, and erosion of user trust. Although platforms offer mechanisms to mitigate exposure to undesired content, these mechanisms are often insufficiently effective and slow to adapt to users' feedback. This paper introduces an intuitive, model-agnostic, and distribution-free method that uses conformal risk control to provably bound unwanted content in personalized recommendations by leveraging simple binary feedback on items. We also address a limitation of traditional conformal risk control approaches, i.e., the fact that the recommender can provide a smaller set of recommended items, by leveraging implicit feedback on consumed items to expand the recommendation set while ensuring robust risk mitigation. Our experimental evaluation on data coming from a popular online video-sharing platform demonstrates that our approach ensures an effective and controllable reduction of unwanted recommendations with minimal effort. The source code is available here: https://github.com/geektoni/mitigating-harm-recsys.

Figures

Figures reproduced from arXiv: 2507.16829 by the authors.

Figure 1
Figure 1. Statistics on the user behaviour on Kuaishou data. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Validity of Property 1 in Kuaishou. (Left) Frequency of reported videos seen the second time [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Results over KuaiRand (k = 20 and β = 0). The results and standard deviation (shaded areas) are over 10 runs. guarantees would still hold in a two-stage system where the score is the probability of not reporting an item, and the ranker predicts watch time. Computational complexity. The threshold λˆ can be precomputed for any risk level α and user U with a complexity of O(Q), where Q is the size of the calibration se… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Fraction of previously seen items in the top-20 recommendations for Kuaishou ( [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Ablation analysis on β for LightGCL (k = 20). We consider only previously seen videos with a watch-time greater than W% > {0%, 50%, 100%}, or without filtering (None). The results and standard deviation (shaded areas) are computed over 10 runs. 8 Discussion and Limitat…
Figure 6
Figure 6. Figure 6: Ablation analysis on low- or high-reporting users (LightGCL, k = 20 and β = 0%). We report the results over 10 runs, where the shaded area indicates the standard deviation. content may lead to unbalanced recommendations, repeatedly exposing users to the same content. T…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 47 canonical work pages

  1. [1]

    Interactive counterfactual exploration of algorithmic harms in recommender systems

    Yongsu Ahn, Quinn K Wolter, Jonilyn Dick, Janet Dick, and Yu-Ru Lin. Interactive counterfactual exploration of algorithmic harms in recommender systems. In 2024 IEEE Visualization in Data Science (VDS), pages 35–39. IEEE, 2024

    work page 2024

  2. [2]

    Measuring and mitigating bias and harm in personalized advertising

    Muhammad Ali. Measuring and mitigating bias and harm in personalized advertising. In Proceedings of the 15th ACM Conference on Recommender Systems , RecSys ’21, page 869–872, New York, NY, USA, 2021. Association for Computing Machinery. ISBN 9781450384582. doi: 10.1145/3460231. 3473895. URL https://doi.org/10.1145/3460231.3473895

    arXiv 2021

  3. [3]

    The dynamics of repeat consumption

    Ashton Anderson, Ravi Kumar, Andrew Tomkins, and Sergei Vassilvitskii. The dynamics of repeat consumption. In Proceedings of the 23rd international conference on World wide web , pages 419–430, 2014

    work page 2014

  4. [4]

    Angelopoulos and Stephen Bates

    Anastasios N. Angelopoulos and Stephen Bates. Conformal prediction: A gentle introduction. Found. Trends Mach. Learn., 16(4):494–591, March 2023. ISSN 1935-8237. doi: 10.1561/2200000101. URL https://doi.org/10.1561/2200000101

    doi:10.1561/2200000101 2023

  5. [5]

    Learn then test: Calibrating predictive algorithms to achieve risk control

    Anastasios N Angelopoulos, Stephen Bates, Emmanuel J Cand` es, Michael I Jordan, and Lihua Lei. Learn then test: Calibrating predictive algorithms to achieve risk control. arXiv preprint arXiv:2110.01052, 2021

    arXiv 2021

  6. [6]

    Conformal risk control

    Anastasios N Angelopoulos, Stephen Bates, Adam Fisch, Lihua Lei, and Tal Schuster. Conformal risk control. ICLR, 2023

    work page 2023

  7. [7]

    Recommendation systems with distribution-free reliability guarantees

    Anastasios N Angelopoulos, Karl Krauth, Stephen Bates, Yixin Wang, and Michael I Jordan. Recommendation systems with distribution-free reliability guarantees. In Conformal and Probabilistic Prediction with Applications, pages 175–193. PMLR, 2023

    work page 2023

  8. [8]

    Deep learning for hate speech detection in tweets

    Pinkesh Badjatiya, Shashank Gupta, Manish Gupta, and Vasudeva Varma. Deep learning for hate speech detection in tweets. In Proceedings of the 26th international conference on World Wide Web companion, pages 759–760, 2017

    2017

Show all 66 references
  1. [9]

    Bias in search and recommender systems

    Ricardo Baeza-Yates. Bias in search and recommender systems. In Proceedings of the 14th ACM conference on recommender systems, pages 2–2, 2020

    2020

  2. [10]

    Distribution- free, risk-controlling prediction sets

    Stephen Bates, Anastasios Angelopoulos, Lihua Lei, Jitendra Malik, and Michael Jordan. Distribution- free, risk-controlling prediction sets. Journal of the ACM (JACM) , 68(6):1–34, 2021

    2021

  3. [11]

    Quantifying the invisible audience in social networks

    Michael S Bernstein, Eytan Bakshy, Moira Burke, and Brian Karrer. Quantifying the invisible audience in social networks. In Proceedings of the SIGCHI conference on human factors in computing systems, pages 21–30, 2013

    2013

  4. [12]

    Advances in bias-aware recommendation on the web

    Ludovico Boratto and Mirko Marras. Advances in bias-aware recommendation on the web. In Proceedings of the 14th ACM international conference on web search and data mining , pages 1147– 1149, 2021

    2021

  5. [13]

    Connecting user and item perspectives in popularity debiasing for collaborative recommendation

    Ludovico Boratto, Gianni Fenu, and Mirko Marras. Connecting user and item perspectives in popularity debiasing for collaborative recommendation. Information Processing & Management , 58 (1):102387, 2021

    2021

  6. [14]

    The music streaming sessions dataset

    Brian Brost, Rishabh Mehrotra, and Tristan Jehan. The music streaming sessions dataset. In The World Wide Web Conference , WWW ’19, page 2594–2600, New York, NY, USA, 2019. Association for Computing Machinery. ISBN 9781450366748. doi: 10.1145/3308558.3313641. URL https: //doi....

    2019

  7. [15]

    LightGCL: Simple yet effective graph contrastive learning for recommendation

    Xuheng Cai, Chao Huang, Lianghao Xia, and Xubin Ren. LightGCL: Simple yet effective graph contrastive learning for recommendation. In The Eleventh International Conference on Learning Representations, 2023. URL https://openreview.net/forum?id=FKXVK9dyMM. 14

    2023

  8. [16]

    Trustworthy and profit: A new value-based neighbor selection method in recommender systems under shilling attacks

    Yuanfeng Cai and Dan Zhu. Trustworthy and profit: A new value-based neighbor selection method in recommender systems under shilling attacks. Decision Support Systems , 124:113112, 2019

    2019

  9. [17]

    Margarida Campos, Ant´ onio Farinhas, Chrysoula Zerva, M´ ario A. T. Figueiredo, and Andr´ e F. T. Martins. Conformal prediction for natural language processing: A survey. Transactions of the Association for Computational Linguistics , 12:1497–1516, 11 2024. ISSN 2307-387X. do...

    2024 doi

  10. [18]

    Harm mitigation in recommender systems under user preference dynamics

    Jerry Chee, Shankar Kalyanaraman, Sindhu Kiranmai Ernala, Udi Weinsberg, Sarah Dean, and Stratis Ioannidis. Harm mitigation in recommender systems under user preference dynamics. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages 25...

    2024

  11. [19]

    Sigformer: Sign-aware graph transformer for recommendation

    Sirui Chen, Jiawei Chen, Sheng Zhou, Bohao Wang, Shen Han, Chanfei Su, Yuqing Yuan, and Can Wang. Sigformer: Sign-aware graph transformer for recommendation. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , SI...

    2024

  12. [20]

    The effect of people recom- menders on echo chambers and polarization

    Federico Cinus, Marco Minici, Corrado Monti, and Francesco Bonchi. The effect of people recom- menders on echo chambers and polarization. In Proceedings of the International AAAI Conference on Web and Social Media , volume 16, pages 90–101, 2022

    2022

  13. [21]

    Recode: Modeling repeat consumption with neural ode

    Sunhao Dai, Changle Qu, Sirui Chen, Xiao Zhang, and Jun Xu. Recode: Modeling repeat consumption with neural ode. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , SIGIR ’24, page 2599–2603, New York, NY, USA, 2...

    2024

  14. [22]

    Towards human-ai complementarity with prediction sets

    Giovanni De Toni, Nastaran Okati, Suhas Thejaswi, Eleni Straitouri, and Manuel Gomez-Rodriguez. Towards human-ai complementarity with prediction sets. In Proceedings of the 38th International Conference on Neural Information Processing Systems , NIPS ’24, Red Hook, NY, USA, 20...

    2025

  15. [23]

    Multimedia recom- mender systems: Algorithms and challenges

    Yashar Deldjoo, Markus Schedl, Bal´ azs Hidasi, Yinwei Wei, and Xiangnan He. Multimedia recom- mender systems: Algorithms and challenges. In Recommender systems handbook, pages 973–1014. Springer, 2021

    2021

  16. [24]

    Facebook content moderators are suffering from ptsd symptoms

    Elizabeth Dwoskin. Facebook content moderators are suffering from ptsd symptoms. now they’re suing. The Washington Post , 2020. URL https://www.washingtonpost.com/technology/2020/ 05/12/facebook-content-moderator-ptsd/

    2020

  17. [25]

    Towards responsible media recommendation

    Mehdi Elahi, Dietmar Jannach, Lars Skjærven, Erik Knudsen, Helle Sjøvaag, Kristian Tolonen, Øyvind Holmstad, Igor Pipkin, Eivind Throndsen, Agnes Stenbom, et al. Towards responsible media recommendation. AI and Ethics , pages 1–12, 2022

    2022

  18. [26]

    ” i always assumed that i wasn’t really that close to [her]” reasoning about invisible algorithms in news feeds

    Motahhare Eslami, Aimee Rickman, Kristen Vaccaro, Amirhossein Aleyasen, Andy Vuong, Karrie Karahalios, Kevin Hamilton, and Christian Sandvig. ” i always assumed that i wasn’t really that close to [her]” reasoning about invisible algorithms in news feeds. In Proceedings of the ...

    2015

  19. [27]

    First i” like” it, then i hide it: Folk theories of social feeds

    Motahhare Eslami, Karrie Karahalios, Christian Sandvig, Kristen Vaccaro, Aimee Rickman, Kevin Hamilton, and Alex Kirlik. First i” like” it, then i hide it: Folk theories of social feeds. In Proceedings of the 2016 cHI conference on human factors in computing systems , pages 23...

    2016

  20. [28]

    Recommender systems and misinformation: The problem or the solution? In OHARS@RecSys, pages 40–50, 2020

    Miriam Fern´ andez and Alejandro Bellog ´ ın. Recommender systems and misinformation: The problem or the solution? In OHARS@RecSys, pages 40–50, 2020. URL https://ceur-ws.org/Vol-2758/ OHARS-paper3.pdf. 15

    2020

  21. [29]

    A survey of graph neural networks for recommender systems: Challenges, methods, and directions

    Chen Gao, Yu Zheng, Nian Li, Yinfeng Li, Yingrong Qin, Jinghua Piao, Yuhan Quan, Jianxin Chang, Depeng Jin, Xiangnan He, and Yong Li. A survey of graph neural networks for recommender systems: Challenges, methods, and directions. ACM Trans. Recomm. Syst., 1(1), March 2023. doi...

    2023 doi

  22. [30]

    Kuairand: An unbiased sequential recommendation dataset with randomly exposed videos

    Chongming Gao, Shijun Li, Yuan Zhang, Jiawei Chen, Biao Li, Wenqiang Lei, Peng Jiang, and Xiangnan He. Kuairand: An unbiased sequential recommendation dataset with randomly exposed videos. In Proceedings of the 31st ACM International Conference on Information & Knowledge Manag...

    2022

  23. [31]

    Beyond accuracy: evaluating rec- ommender systems by coverage and serendipity

    Mouzhi Ge, Carla Delgado-Battenfeld, and Dietmar Jannach. Beyond accuracy: evaluating rec- ommender systems by coverage and serendipity. In Proceedings of the fourth ACM conference on Recommender systems, pages 257–260, 2010

    2010

  24. [32]

    A survey on trustworthy recommender systems

    Yingqiang Ge, Shuchang Liu, Zuohui Fu, Juntao Tan, Zelong Li, Shuyuan Xu, Yunqi Li, Yikun Xian, and Yongfeng Zhang. A survey on trustworthy recommender systems. ACM Transactions on Recommender Systems, 3(2):1–68, 2024

    2024

  25. [33]

    Balancing consumer and business value of recommender systems: A simulation-based analysis

    Nada Ghanem, Stephan Leitner, and Dietmar Jannach. Balancing consumer and business value of recommender systems: A simulation-based analysis. Electronic Commerce Research and Applications, 55:101195, 2022

    2022

  26. [34]

    Research directions in recommender systems for health and well-being: A preface to the special issue

    Hanna Hauptmann, Alan Said, and Christoph Trattner. Research directions in recommender systems for health and well-being: A preface to the special issue. User Modeling and User-Adapted Interaction, 32(5):781–786, 2022

    2022

  27. [35]

    Neural collaborative filtering

    Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. Neural collaborative filtering. In Proceedings of the 26th international conference on world wide web , pages 173–182, 2017

    2017

  28. [36]

    User tampering in reinforcement learning recommender systems

    Atoosa Kasirzadeh and Charles Evans. User tampering in reinforcement learning recommender systems. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society , pages 58–69, 2023

    2023

  29. [37]

    Beinke, Gerrit Y

    Thorsten Krause, Alina Deriyeva, Jan H. Beinke, Gerrit Y. Bartels, and Oliver Thomas. Mitigating exposure bias in recommender systems—a comparative analysis of discrete choice models. ACM Trans. Recomm. Syst., 3(2), November 2024. doi: 10.1145/3641291. URL https://doi.org/10. ...

    2024 doi

  30. [38]

    Graph transformer for recommendation

    Chaoliu Li, Lianghao Xia, Xubin Ren, Yaowen Ye, Yong Xu, and Chao Huang. Graph transformer for recommendation. In Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval , pages 1680–1689, 2023

    2023

  31. [39]

    Structured conformal inference for matrix completion with applications to group recommender systems

    Ziyi Liang, Tianmin Xie, Xin Tong, and Matteo Sesia. Structured conformal inference for matrix completion with applications to group recommender systems. arXiv preprint arXiv:2404.17561 , 2024

    2024 arXiv

  32. [40]

    How to train your youtube recommender to avoid unwanted videos

    Alexander Liu, Siqi Wu, and Paul Resnick. How to train your youtube recommender to avoid unwanted videos. In Proceedings of the International AAAI Conference on Web and Social Media , volume 18, pages 930–942, 2024

    2024

  33. [41]

    Understanding web browsing behaviors through weibull analysis of dwell time

    Chao Liu, Ryen W White, and Susan Dumais. Understanding web browsing behaviors through weibull analysis of dwell time. In Proceedings of the 33rd international ACM SIGIR conference on Research and development in information retrieval , pages 379–386, 2010

    2010

  34. [42]

    Machine learning based recommender system for e-commerce

    Manal Loukili, Fay¸ cal Messaoudi, and Mohammed El Ghazi. Machine learning based recommender system for e-commerce. IAES International Journal of Artificial Intelligence , 12(4):1803–1811, 2023

    2023

  35. [43]

    Youtube regrets: A crowdsourced investigation into youtube’s recommendation algorithm

    Jesse McCrosky and Brandi Geurkink. Youtube regrets: A crowdsourced investigation into youtube’s recommendation algorithm. Mozilla Foundation. Retrieved November , 15:2021, 2021. 16

    2021

  36. [44]

    Controlling Fairness and Bias in Dynamic Learning-to-Rank

    Marco Morik, Ashudeep Singh, Jessica Hong, and Thorsten Joachims. Controlling Fairness and Bias in Dynamic Learning-to-Rank. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval , SIGIR ’20, pages 429–438. Associati...

    2020

  37. [45]

    User personality and user satisfaction with recommender systems

    Tien T Nguyen, F Maxwell Harper, Loren Terveen, and Joseph A Konstan. User personality and user satisfaction with recommender systems. Information systems frontiers , 20:1173–1189, 2018

    2018

  38. [46]

    Pairwise preference regression for cold-start recommendation

    Seung-Taek Park and Wei Chu. Pairwise preference regression for cold-start recommendation. In Proceedings of the third ACM conference on Recommender systems , pages 21–28, 2009

    2009

  39. [47]

    Triggered essential reviewing: the effect of technology affordances on service experience evaluations

    Gabriele Piccoli. Triggered essential reviewing: the effect of technology affordances on service experience evaluations. European Journal of Information Systems , 25(6):477–492, 2016. doi: 10. 1057/s41303-016-0019-9. URL https://doi.org/10.1057/s41303-016-0019-9

    2016 doi

  40. [48]

    Repeatnet: A repeat aware neural recommendation machine for session-based recommendation

    Pengjie Ren, Zhumin Chen, Jing Li, Zhaochun Ren, Jun Ma, and Maarten de Rijke. Repeatnet: A repeat aware neural recommendation machine for session-based recommendation. Proceedings of the AAAI Conference on Artificial Intelligence , 33(01):4806–4813, Jul. 2019. doi: 10.1609/aa...

    2019 doi

  41. [49]

    How social media influencers affect be- havioural intentions towards recommended brands: the role of emotional attachment and information value

    Raquel S´ anchez-Fern´ andez and David Jim´ enez-Castillo. How social media influencers affect be- havioural intentions towards recommended brands: the role of emotional attachment and information value. Journal of Marketing Management , 37(11-12):1123–1147, 2021

    2021

  42. [50]

    The effectiveness of moderating harmful online content

    Philipp J Schneider and Marian-Andrei Rizoiu. The effectiveness of moderating harmful online content. Proceedings of the National Academy of Sciences , 120(34):e2307360120, 2023

    2023

  43. [51]

    Siren: Sign-aware rec- ommendation using graph neural networks

    Changwon Seo, Kyeong-Joong Jeong, Sungsu Lim, and Won-Yong Shin. Siren: Sign-aware rec- ommendation using graph neural networks. IEEE Transactions on Neural Networks and Learning Systems, 35(4):4729–4743, 2022

    2022

  44. [52]

    A tutorial on conformal prediction

    Glenn Shafer and Vladimir Vovk. A tutorial on conformal prediction. Journal of Machine Learning Research, 9(3), 2008

    2008

  45. [53]

    Smith, Lucia Jayne, and Robin Burke

    Jessie J. Smith, Lucia Jayne, and Robin Burke. Recommender systems and algorithmic hate. In Proceedings of the 16th ACM Conference on Recommender Systems , RecSys ’22, page 592–597, New York, NY, USA, 2022. Association for Computing Machinery. ISBN 9781450392785. doi: 10.1145/...

    2022

  46. [54]

    Designing decision support systems using counter- factual prediction sets

    Eleni Straitouri and Manuel Gomez Rodriguez. Designing decision support systems using counter- factual prediction sets. In Proceedings of the 41st International Conference on Machine Learning , ICML’24. JMLR.org, 2024

    2024

  47. [55]

    Improving expert predictions with conformal prediction

    Eleni Straitouri, Lequn Wang, Nastaran Okati, and Manuel Gomez Rodriguez. Improving expert predictions with conformal prediction. In International Conference on Machine Learning , pages 32633–32653. PMLR, 2023

    2023

  48. [56]

    An audit of misinformation filter bubbles on youtube: Bubble bursting and recent behavior changes

    Matus Tomlein, Branislav Pecher, Jakub Simko, Ivan Srba, Robert Moro, Elena Stefancova, Michal Kompan, Andrea Hrckova, Juraj Podrouzek, and Maria Bielikova. An audit of misinformation filter bubbles on youtube: Bubble bursting and recent behavior changes. In Proceedings of the...

    2021

  49. [57]

    Ohars: Second workshop on online misinformation- and harm-aware recommender systems

    Antonela Tommasel, Daniela Godoy, and Arkaitz Zubiaga. Ohars: Second workshop on online misinformation- and harm-aware recommender systems. In Proceedings of the 15th ACM Conference on Recommender Systems , RecSys ’21, page 789–791, New York, NY, USA, 2021. Association for Com...

    2021

  50. [58]

    I want to break free! recommending friends from outside the echo chamber

    Antonela Tommasel, Juan Manuel Rodriguez, and Daniela Godoy. I want to break free! recommending friends from outside the echo chamber. In Proceedings of the 15th ACM Conference on Recommender Systems, pages 23–33, 2021

    2021

  51. [59]

    Modeling item-specific temporal dynamics of repeat consumption for recommender systems

    Chenyang Wang, Min Zhang, Weizhi Ma, Yiqun Liu, and Shaoping Ma. Modeling item-specific temporal dynamics of repeat consumption for recommender systems. In The world wide web conference, pages 1977–1987, 2019

    1977

  52. [60]

    Uncertainty quantification for fairness in two-stage recommender systems

    Lequn Wang and Thorsten Joachims. Uncertainty quantification for fairness in two-stage recommender systems. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining, WSDM ’23, page 940–948, New York, NY, USA, 2023. Association for Computing Ma...

    2023

  53. [61]

    Two-stage conformal risk control with application to ranked retrieval

    Yunpeng Xu, Wenge Guo, and Zhi Wei. Two-stage conformal risk control with application to ranked retrieval. arXiv preprint arXiv:2404.17769 , 2024

    2024 arXiv

  54. [62]

    Prefrec: Recommender systems with human preferences for reinforcing long-term user engagement

    Wanqi Xue, Qingpeng Cai, Zhenghai Xue, Shuo Sun, Shuchang Liu, Dong Zheng, Peng Jiang, Kun Gai, and Bo An. Prefrec: Recommender systems with human preferences for reinforcing long-term user engagement. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and...

    2023

  55. [63]

    A unified graph transformer for overcoming isolations in multi-modal recommendation

    Zixuan Yi and Iadh Ounis. A unified graph transformer for overcoming isolations in multi-modal recommendation. In Proceedings of the 18th ACM Conference on Recommender Systems , RecSys ’24, page 518–527, New York, NY, USA, 2024. Association for Computing Machinery. ISBN 979840...

    2024

  56. [64]

    Silence is also evidence: interpreting dwell time for recommendation from psychological perspective

    Peifeng Yin, Ping Luo, Wang-Chien Lee, and Min Wang. Silence is also evidence: interpreting dwell time for recommendation from psychological perspective. In Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining , pages 989–997, 2013

    2013

  57. [65]

    Folk theories of algorithms: Understanding digital irritation

    Brita Ytre-Arne and Hallvard Moe. Folk theories of algorithms: Understanding digital irritation. Media, Culture & Society , 43(5):807–824, 2021. doi: 10.1177/0163443720972314. URL https: //doi.org/10.1177/0163443720972314

    2021 doi

  58. [66]

    Fairness in ranking, part i: Score-based ranking

    Meike Zehlike, Ke Yang, and Julia Stoyanovich. Fairness in ranking, part i: Score-based ranking. ACM Computing Surveys , 55(6):1–36, 2022. 18

    2022

Pith tools

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