Search

Large language models (LLMs) represent a promising, but controversial, tool in aiding scientific peer review. This study evaluates the usefulness of LLMs in a conference setting as a tool for vetting paper submissions against submission standards. We conduct an experiment at the 2024 Neural Information Processing Systems (NeurIPS) conference, where 234 papers were voluntarily submitted to an "LLM-based Checklist Assistant." This assistant validates whether papers adhere to the author checklist used by NeurIPS, which includes questions to ensure compliance with research and manuscript preparation standards. Evaluation of the assistant by NeurIPS paper authors suggests that the LLM-based assistant was generally helpful in verifying checklist completion. In post-usage surveys, over 70% of authors found the assistant useful, and 70% indicate that they would revise their papers or checklist responses based on its feedback. While causal attribution to the assistant is not definitive, qualitative evidence suggests that the LLM contributed to improving some submissions. Survey responses and analysis of re-submissions indicate that authors made substantive revisions to their submissions in response to specific feedback from the LLM. The experiment also highlights common issues with LLMs: inaccuracy (20/52) and excessive strictness (14/52) were the most frequent issues flagged by authors. We also conduct experiments to understand potential gaming of the system, which reveal that the assistant could be manipulated to enhance scores through fabricated justifications, highlighting potential vulnerabilities of automated review tools.

The increasing volume of papers and proposals undergoing peer review emphasizes the pressing need for greater automation to effectively manage the growing scale. In this study, we present the deployment and evaluation of machine learning and optimization techniques for assigning proposals to reviewers that was developed for the Atacama Large Millimeter/submillimeter Array (ALMA) during the Cycle 10 Call for Proposals issued in 2023. By utilizing topic modeling algorithms, we identify the proposal topics and assess reviewers' expertise based on their historical ALMA proposal submissions. We then apply an adapted version of the assignment optimization algorithm from PeerReview4All (Stelmakh et al. 2021a) to maximize the alignment between proposal topics and reviewer expertise. Our evaluation shows a significant improvement in matching reviewer expertise: the median similarity score between the proposal topic and reviewer expertise increased by 51 percentage points compared to the previous cycle, and the percentage of reviewers reporting expertise in their assigned proposals rose by 20 percentage points. Furthermore, the assignment process proved highly effective in that no proposals required reassignment due to significant mismatches, resulting in a savings of 3 to 5 days of manual effort., Comment: 19 pages, 5 figures, submitted to PASP

The number of scientific articles produced every year is growing rapidly. Providing quality control over them is crucial for scientists and, ultimately, for the public good. In modern science, this process is largely delegated to peer review -- a distributed procedure in which each submission is evaluated by several independent experts in the field. Peer review is widely used, yet it is hard, time-consuming, and prone to error. Since the artifacts involved in peer review -- manuscripts, reviews, discussions -- are largely text-based, Natural Language Processing has great potential to improve reviewing. As the emergence of large language models (LLMs) has enabled NLP assistance for many new tasks, the discussion on machine-assisted peer review is picking up the pace. Yet, where exactly is help needed, where can NLP help, and where should it stand aside? The goal of our paper is to provide a foundation for the future efforts in NLP for peer-reviewing assistance. We discuss peer review as a general process, exemplified by reviewing at AI conferences. We detail each step of the process from manuscript submission to camera-ready revision, and discuss the associated challenges and opportunities for NLP assistance, illustrated by existing work. We then turn to the big challenges in NLP for peer review as a whole, including data acquisition and licensing, operationalization and experimentation, and ethical issues. To help consolidate community efforts, we create a companion repository that aggregates key datasets pertaining to peer review. Finally, we issue a detailed call for action for the scientific community, NLP and AI researchers, policymakers, and funding bodies to help bring the research in NLP for peer review forward. We hope that our work will help set the agenda for research in machine-assisted scientific quality control in the age of AI, within the NLP community and beyond.

Peer review often involves reviewers submitting their independent reviews, followed by a discussion among reviewers of each paper. A question among policymakers is whether the reviewers of a paper should be anonymous to each other during the discussion. We shed light on this by conducting a randomized controlled trial at the UAI 2022 conference. We randomly split the reviewers and papers into two conditions--one with anonymous discussions and the other with non-anonymous discussions, and conduct an anonymous survey of all reviewers, to address the following questions: 1. Do reviewers discuss more in one of the conditions? Marginally more in anonymous (n = 2281, p = 0.051). 2. Does seniority have more influence on final decisions when non-anonymous? Yes, the decisions are closer to senior reviewers' scores in the non-anonymous condition than in anonymous (n = 484, p = 0.04). 3. Are reviewers more polite in one of the conditions? No significant difference in politeness of reviewers' text-based responses (n = 1125, p = 0.72). 4. Do reviewers' self-reported experiences differ across the two conditions? No significant difference for each of the five questions asked (n = 132 and p > 0.3). 5. Do reviewers prefer one condition over the other? Yes, there is a weak preference for anonymous discussions (n = 159 and Cohen's d= 0.25). 6. What do reviewers consider important to make policy on anonymity among reviewers? Reviewers' feeling of safety in expressing their opinions was rated most important, while polite communication among reviewers was rated least important (n = 159). 7. Have reviewers experienced dishonest behavior due to non-anonymity in discussions? Yes, roughly 7% of respondents answered affirmatively (n = 167). Overall, this experiment reveals evidence supporting an anonymous discussion setup in the peer-review process, in terms of the evaluation criteria considered., Comment: 18 pages, 4 figures, 3 tables

A major threat to the peer-review systems of computer science conferences is the existence of "collusion rings" between reviewers. In such collusion rings, reviewers who have also submitted their own papers to the conference work together to manipulate the conference's paper assignment, with the aim of being assigned to review each other's papers. The most straightforward way that colluding reviewers can manipulate the paper assignment is by indicating their interest in each other's papers through strategic paper bidding. One potential approach to solve this important problem would be to detect the colluding reviewers from their manipulated bids, after which the conference can take appropriate action. While prior work has developed effective techniques to detect other kinds of fraud, no research has yet established that detecting collusion rings is even possible. In this work, we tackle the question of whether it is feasible to detect collusion rings from the paper bidding. To answer this question, we conduct empirical analysis of two realistic conference bidding datasets, including evaluations of existing algorithms for fraud detection in other applications. We find that collusion rings can achieve considerable success at manipulating the paper assignment while remaining hidden from detection: for example, in one dataset, undetected colluders are able to achieve assignment to up to 30% of the papers authored by other colluders. In addition, when 10 colluders bid on all of each other's papers, no detection algorithm outputs a group of reviewers with more than 31% overlap with the true colluders. These results suggest that collusion cannot be effectively detected from the bidding using popular existing tools, demonstrating the need to develop more complex detection algorithms as well as those that leverage additional metadata (e.g., reviewer-paper text-similarity scores).

Is it possible to reliably evaluate the quality of peer reviews? We study this question driven by two primary motivations -- incentivizing high-quality reviewing using assessed quality of reviews and measuring changes to review quality in experiments. We conduct a large scale study at the NeurIPS 2022 conference, a top-tier conference in machine learning, in which we invited (meta)-reviewers and authors to evaluate reviews given to submitted papers. First, we conduct a RCT to examine bias due to the length of reviews. We generate elongated versions of reviews by adding substantial amounts of non-informative content. Participants in the control group evaluate the original reviews, whereas participants in the experimental group evaluate the artificially lengthened versions. We find that lengthened reviews are scored (statistically significantly) higher quality than the original reviews. In analysis of observational data we find that authors are positively biased towards reviews recommending acceptance of their own papers, even after controlling for confounders of review length, quality, and different numbers of papers per author. We also measure disagreement rates between multiple evaluations of the same review of 28%-32%, which is comparable to that of paper reviewers at NeurIPS. Further, we assess the amount of miscalibration of evaluators of reviews using a linear model of quality scores and find that it is similar to estimates of miscalibration of paper reviewers at NeurIPS. Finally, we estimate the amount of variability in subjective opinions around how to map individual criteria to overall scores of review quality and find that it is roughly the same as that in the review of papers. Our results suggest that the various problems that exist in reviews of papers -- inconsistency, bias towards irrelevant factors, miscalibration, subjectivity -- also arise in reviewing of reviews.

Peer review frequently follows a process where reviewers first provide initial reviews, authors respond to these reviews, then reviewers update their reviews based on the authors' response. There is mixed evidence regarding whether this process is useful, including frequent anecdotal complaints that reviewers insufficiently update their scores. In this study, we aim to investigate whether reviewers anchor to their original scores when updating their reviews, which serves as a potential explanation for the lack of updates in reviewer scores. We design a novel randomized controlled trial to test if reviewers exhibit anchoring. In the experimental condition, participants initially see a flawed version of a paper that is later corrected, while in the control condition, participants only see the correct version. We take various measures to ensure that in the absence of anchoring, reviewers in the experimental group should revise their scores to be identically distributed to the scores from the control group. Furthermore, we construct the reviewed paper to maximize the difference between the flawed and corrected versions, and employ deception to hide the true experiment purpose. Our randomized controlled trial consists of 108 researchers as participants. First, we find that our intervention was successful at creating a difference in perceived paper quality between the flawed and corrected versions: Using a permutation test with the Mann-Whitney U statistic, we find that the experimental group's initial scores are lower than the control group's scores in both the Evaluation category (Vargha-Delaney A=0.64, p=0.0096) and Overall score (A=0.59, p=0.058). Next, we test for anchoring by comparing the experimental group's revised scores with the control group's scores. We find no significant evidence of anchoring in either the Overall (A=0.50, p=0.61) or Evaluation category (A=0.49, p=0.61)., Comment: 14 pages (19 including references and appendix), 2 figures

Given the rapid ascent of large language models (LLMs), we study the question: (How) can large language models help in reviewing of scientific papers or proposals? We first conduct some pilot studies where we find that (i) GPT-4 outperforms other LLMs (Bard, Vicuna, Koala, Alpaca, LLaMa, Dolly, OpenAssistant, StableLM), and (ii) prompting with a specific question (e.g., to identify errors) outperforms prompting to simply write a review. With these insights, we study the use of LLMs (specifically, GPT-4) for three tasks: 1. Identifying errors: We construct 13 short computer science papers each with a deliberately inserted error, and ask the LLM to check for the correctness of these papers. We observe that the LLM finds errors in 7 of them, spanning both mathematical and conceptual errors. 2. Verifying checklists: We task the LLM to verify 16 closed-ended checklist questions in the respective sections of 15 NeurIPS 2022 papers. We find that across 119 {checklist question, paper} pairs, the LLM had an 86.6% accuracy. 3. Choosing the "better" paper: We generate 10 pairs of abstracts, deliberately designing each pair in such a way that one abstract was clearly superior than the other. The LLM, however, struggled to discern these relatively straightforward distinctions accurately, committing errors in its evaluations for 6 out of the 10 pairs. Based on these experiments, we think that LLMs have a promising use as reviewing assistants for specific reviewing tasks, but not (yet) for complete evaluations of papers or proposals.

Peer review assignment algorithms aim to match research papers to suitable expert reviewers, working to maximize the quality of the resulting reviews. A key challenge in designing effective assignment policies is evaluating how changes to the assignment algorithm map to changes in review quality. In this work, we leverage recently proposed policies that introduce randomness in peer-review assignment--in order to mitigate fraud--as a valuable opportunity to evaluate counterfactual assignment policies. Specifically, we exploit how such randomized assignments provide a positive probability of observing the reviews of many assignment policies of interest. To address challenges in applying standard off-policy evaluation methods, such as violations of positivity, we introduce novel methods for partial identification based on monotonicity and Lipschitz smoothness assumptions for the mapping between reviewer-paper covariates and outcomes. We apply our methods to peer-review data from two computer science venues: the TPDP'21 workshop (95 papers and 35 reviewers) and the AAAI'22 conference (8,450 papers and 3,145 reviewers). We consider estimates of (i) the effect on review quality when changing weights in the assignment algorithm, e.g., weighting reviewers' bids vs. textual similarity (between the review's past papers and the submission), and (ii) the "cost of randomization", capturing the difference in expected quality between the perturbed and unperturbed optimal match. We find that placing higher weight on text similarity results in higher review quality and that introducing randomization in the reviewer-paper assignment only marginally reduces the review quality. Our methods for partial identification may be of independent interest, while our off-policy approach can likely find use evaluating a broad class of algorithmic matching systems.

Many peer-review venues are either using or looking to use algorithms to assign submissions to reviewers. The crux of such automated approaches is the notion of the "similarity score"--a numerical estimate of the expertise of a reviewer in reviewing a paper--and many algorithms have been proposed to compute these scores. However, these algorithms have not been subjected to a principled comparison, making it difficult for stakeholders to choose the algorithm in an evidence-based manner. The key challenge in comparing existing algorithms and developing better algorithms is the lack of the publicly available gold-standard data that would be needed to perform reproducible research. We address this challenge by collecting a novel dataset of similarity scores that we release to the research community. Our dataset consists of 477 self-reported expertise scores provided by 58 researchers who evaluated their expertise in reviewing papers they have read previously. We use this data to compare several popular algorithms employed in computer science conferences and come up with recommendations for stakeholders. Our main findings are as follows. First, all algorithms make a non-trivial amount of error. For the task of ordering two papers in terms of their relevance for a reviewer, the error rates range from 12%-30% in easy cases to 36%-43% in hard cases, highlighting the vital need for more research on the similarity-computation problem. Second, most existing algorithms are designed to work with titles and abstracts of papers, and in this regime the Specter+MFR algorithm performs best. Third, to improve performance, it may be important to develop modern deep-learning based algorithms that can make use of the full texts of papers: the classical TD-IDF algorithm enhanced with full texts of papers is on par with the deep-learning based Specter+MFR that cannot make use of this information.

Many practical applications, ranging from paper-reviewer assignment in peer review to job-applicant matching for hiring, require human decision makers to identify relevant matches by combining their expertise with predictions from machine learning models. In many such model-assisted document matching tasks, the decision makers have stressed the need for assistive information about the model outputs (or the data) to facilitate their decisions. In this paper, we devise a proxy matching task that allows us to evaluate which kinds of assistive information improve decision makers' performance (in terms of accuracy and time). Through a crowdsourced (N=271 participants) study, we find that providing black-box model explanations reduces users' accuracy on the matching task, contrary to the commonly-held belief that they can be helpful by allowing better understanding of the model. On the other hand, custom methods that are designed to closely attend to some task-specific desiderata are found to be effective in improving user performance. Surprisingly, we also find that the users' perceived utility of assistive information is misaligned with their objective utility (measured through their task performance).

There is widespread debate on whether to anonymize author identities in peer review. The key argument for anonymization is to mitigate bias, whereas arguments against anonymization posit various uses of author identities in the review process. The Innovations in Theoretical Computer Science (ITCS) 2023 conference adopted a middle ground by initially anonymizing the author identities from reviewers, revealing them after the reviewer had submitted their initial reviews, and allowing the reviewer to change their review subsequently. We present an analysis of the reviews pertaining to the identification and use of author identities. Our key findings are: (I) A majority of reviewers self-report not knowing and being unable to guess the authors' identities for the papers they were reviewing. (II) After the initial submission of reviews, 7.1% of reviews changed their overall merit score and 3.8% changed their self-reported reviewer expertise. (III) There is a very weak and statistically insignificant correlation of the rank of authors' affiliations with the change in overall merit; there is a weak but statistically significant correlation with respect to change in reviewer expertise. We also conducted an anonymous survey to obtain opinions from reviewers and authors. The main findings from the 200 survey responses are: (i) A vast majority of participants favor anonymizing author identities in some form. (ii) The "middle-ground" initiative of ITCS 2023 was appreciated. (iii) Detecting conflicts of interest is a challenge that needs to be addressed if author identities are anonymized. Overall, these findings support anonymization of author identities in some form (e.g., as was done in ITCS 2023), as long as there is a robust and efficient way to check conflicts of interest.

There are a number of forums where people participate under pseudonyms. One example is peer review, where the identity of reviewers for any paper is confidential. When participating in these forums, people frequently engage in "batching": executing multiple related tasks (e.g., commenting on multiple papers) at nearly the same time. Our empirical analysis shows that batching is common in two applications we consider $\unicode{x2013}$ peer review and Wikipedia edits. In this paper, we identify and address the risk of deanonymization arising from linking batched tasks. To protect against linkage attacks, we take the approach of adding delay to the posting time of batched tasks. We first show that under some natural assumptions, no delay mechanism can provide a meaningful differential privacy guarantee. We therefore propose a "one-sided" formulation of differential privacy for protecting against linkage attacks. We design a mechanism that adds zero-inflated uniform delay to events and show it can preserve privacy. We prove that this noise distribution is in fact optimal in minimizing expected delay among mechanisms adding independent noise to each event, thereby establishing the Pareto frontier of the trade-off between the expected delay for batched and unbatched events. Finally, we conduct a series of experiments on Wikipedia and Bitcoin data that corroborate the practical utility of our algorithm in obfuscating batching without introducing onerous delay to a system.

How do author perceptions match up to the outcomes of the peer-review process and perceptions of others? In a top-tier computer science conference (NeurIPS 2021) with more than 23,000 submitting authors and 9,000 submitted papers, we survey the authors on three questions: (i) their predicted probability of acceptance for each of their papers, (ii) their perceived ranking of their own papers based on scientific contribution, and (iii) the change in their perception about their own papers after seeing the reviews. The salient results are: (1) Authors have roughly a three-fold overestimate of the acceptance probability of their papers: The median prediction is 70% for an approximately 25% acceptance rate. (2) Female authors exhibit a marginally higher (statistically significant) miscalibration than male authors; predictions of authors invited to serve as meta-reviewers or reviewers are similarly calibrated, but better than authors who were not invited to review. (3) Authors' relative ranking of scientific contribution of two submissions they made generally agree (93%) with their predicted acceptance probabilities, but there is a notable 7% responses where authors think their better paper will face a worse outcome. (4) The author-provided rankings disagreed with the peer-review decisions about a third of the time; when co-authors ranked their jointly authored papers, co-authors disagreed at a similar rate -- about a third of the time. (5) At least 30% of respondents of both accepted and rejected papers said that their perception of their own paper improved after the review process. The stakeholders in peer review should take these findings into account in setting their expectations from peer review.

Many applications such as hiring and university admissions involve evaluation and selection of applicants. These tasks are fundamentally difficult, and require combining evidence from multiple different aspects (what we term "attributes"). In these applications, the number of applicants is often large, and a common practice is to assign the task to multiple evaluators in a distributed fashion. Specifically, in the often-used holistic allocation, each evaluator is assigned a subset of the applicants, and is asked to assess all relevant information for their assigned applicants. However, such an evaluation process is subject to issues such as miscalibration (evaluators see only a small fraction of the applicants and may not get a good sense of relative quality), and discrimination (evaluators are influenced by irrelevant information about the applicants). We identify that such attribute-based evaluation allows alternative allocation schemes. Specifically, we consider assigning each evaluator more applicants but fewer attributes per applicant, termed segmented allocation. We compare segmented allocation to holistic allocation on several dimensions via theoretical and experimental methods. We establish various tradeoffs between these two approaches, and identify conditions under which one approach results in more accurate evaluation than the other.

Many conferences rely on paper bidding as a key component of their reviewer assignment procedure. These bids are then taken into account when assigning reviewers to help ensure that each reviewer is assigned to suitable papers. However, despite the benefits of using bids, reliance on paper bidding can allow malicious reviewers to manipulate the paper assignment for unethical purposes (e.g., getting assigned to a friend's paper). Several different approaches to preventing this manipulation have been proposed and deployed. In this paper, we enumerate certain desirable properties that algorithms for addressing bid manipulation should satisfy. We then offer a high-level analysis of various approaches along with directions for future investigation.

In conference peer review, reviewers are often asked to provide "bids" on each submitted paper that express their interest in reviewing that paper. A paper assignment algorithm then uses these bids (along with other data) to compute a high-quality assignment of reviewers to papers. However, this process has been exploited by malicious reviewers who strategically bid in order to unethically manipulate the paper assignment, crucially undermining the peer review process. For example, these reviewers may aim to get assigned to a friend's paper as part of a quid-pro-quo deal. A critical impediment towards creating and evaluating methods to mitigate this issue is the lack of any publicly-available data on malicious paper bidding. In this work, we collect and publicly release a novel dataset to fill this gap, collected from a mock conference activity where participants were instructed to bid either honestly or maliciously. We further provide a descriptive analysis of the bidding behavior, including our categorization of different strategies employed by participants. Finally, we evaluate the ability of each strategy to manipulate the assignment, and also evaluate the performance of some simple algorithms meant to detect malicious bidding. The performance of these detection algorithms can be taken as a baseline for future research on detecting malicious bidding.

In peer review, reviewers are usually asked to provide scores for the papers. The scores are then used by Area Chairs or Program Chairs in various ways in the decision-making process. The scores are usually elicited in a quantized form to accommodate the limited cognitive ability of humans to describe their opinions in numerical values. It has been found that the quantized scores suffer from a large number of ties, thereby leading to a significant loss of information. To mitigate this issue, conferences have started to ask reviewers to additionally provide a ranking of the papers they have reviewed. There are however two key challenges. First, there is no standard procedure for using this ranking information and Area Chairs may use it in different ways (including simply ignoring them), thereby leading to arbitrariness in the peer-review process. Second, there are no suitable interfaces for judicious use of this data nor methods to incorporate it in existing workflows, thereby leading to inefficiencies. We take a principled approach to integrate the ranking information into the scores. The output of our method is an updated score pertaining to each review that also incorporates the rankings. Our approach addresses the two aforementioned challenges by: (i) ensuring that rankings are incorporated into the updates scores in the same manner for all papers, thereby mitigating arbitrariness, and (ii) allowing to seamlessly use existing interfaces and workflows designed for scores. We empirically evaluate our method on synthetic datasets as well as on peer reviews from the ICLR 2017 conference, and find that it reduces the error by approximately 30% as compared to the best performing baseline on the ICLR 2017 data., Comment: 14 main pages, 7 appendix pages

Citations play an important role in researchers' careers as a key factor in evaluation of scientific impact. Many anecdotes advice authors to exploit this fact and cite prospective reviewers to try obtaining a more positive evaluation for their submission. In this work, we investigate if such a citation bias actually exists: Does the citation of a reviewer's own work in a submission cause them to be positively biased towards the submission? In conjunction with the review process of two flagship conferences in machine learning and algorithmic economics, we execute an observational study to test for citation bias in peer review. In our analysis, we carefully account for various confounding factors such as paper quality and reviewer expertise, and apply different modeling techniques to alleviate concerns regarding the model mismatch. Overall, our analysis involves 1,314 papers and 1,717 reviewers and detects citation bias in both venues we consider. In terms of the effect size, by citing a reviewer's work, a submission has a non-trivial chance of getting a higher score from the reviewer: an expected increase in the score is approximately 0.23 on a 5-point Likert item. For reference, a one-point increase of a score by a single reviewer improves the position of a submission by 11% on average., Comment: 19 pages, 3 figures

Double-blind conferences have engaged in debates over whether to allow authors to post their papers online on arXiv or elsewhere during the review process. Independently, some authors of research papers face the dilemma of whether to put their papers on arXiv due to its pros and cons. We conduct a study to substantiate this debate and dilemma via quantitative measurements. Specifically, we conducted surveys of reviewers in two top-tier double-blind computer science conferences -- ICML 2021 (5361 submissions and 4699 reviewers) and EC 2021 (498 submissions and 190 reviewers). Our two main findings are as follows. First, more than a third of the reviewers self-report searching online for a paper they are assigned to review. Second, outside the review process, we find that preprints from better-ranked affiliations see a weakly higher visibility, with a correlation of 0.06 in ICML and 0.05 in EC. In particular, papers associated with the top-10-ranked affiliations had a visibility of approximately 11% in ICML and 22% in EC, whereas the remaining papers had a visibility of 7% and 18% respectively., Comment: 17 pages, 3 figures

Reviewers in peer review are often miscalibrated: they may be strict, lenient, extreme, moderate, etc. A number of algorithms have previously been proposed to calibrate reviews. Such attempts of calibration can however leak sensitive information about which reviewer reviewed which paper. In this paper, we identify this problem of calibration with privacy, and provide a foundational building block to address it. Specifically, we present a theoretical study of this problem under a simplified-yet-challenging model involving two reviewers, two papers, and an MAP-computing adversary. Our main results establish the Pareto frontier of the tradeoff between privacy (preventing the adversary from inferring reviewer identity) and utility (accepting better papers), and design explicit computationally-efficient algorithms that we prove are Pareto optimal., Comment: 31 pages, 6 figures

Strategic behavior is a fundamental problem in a variety of real-world applications that require some form of peer assessment, such as peer grading of homeworks, grant proposal review, conference peer review of scientific papers, and peer assessment of employees in organizations. Since an individual's own work is in competition with the submissions they are evaluating, they may provide dishonest evaluations to increase the relative standing of their own submission. This issue is typically addressed by partitioning the individuals and assigning them to evaluate the work of only those from different subsets. Although this method ensures strategyproofness, each submission may require a different type of expertise for effective evaluation. In this paper, we focus on finding an assignment of evaluators to submissions that maximizes assigned evaluators' expertise subject to the constraint of strategyproofness. We analyze the price of strategyproofness: that is, the amount of compromise on the assigned evaluators' expertise required in order to get strategyproofness. We establish several polynomial-time algorithms for strategyproof assignment along with assignment-quality guarantees. Finally, we evaluate the methods on a dataset from conference peer review.

Many scientific conferences employ a two-phase paper review process, where some papers are assigned additional reviewers after the initial reviews are submitted. Many conferences also design and run experiments on their paper review process, where some papers are assigned reviewers who provide reviews under an experimental condition. In this paper, we consider the question: how should reviewers be divided between phases or conditions in order to maximize total assignment similarity? We make several contributions towards answering this question. First, we prove that when the set of papers requiring additional review is unknown, a simplified variant of this problem is NP-hard. Second, we empirically show that across several datasets pertaining to real conference data, dividing reviewers between phases/conditions uniformly at random allows an assignment that is nearly as good as the oracle optimal assignment. This uniformly random choice is practical for both the two-phase and conference experiment design settings. Third, we provide explanations of this phenomenon by providing theoretical bounds on the suboptimality of this random strategy under certain natural conditions. From these easily-interpretable conditions, we provide actionable insights to conference program chairs about whether a random reviewer split is suitable for their conference.

It is common to evaluate a set of items by soliciting people to rate them. For example, universities ask students to rate the teaching quality of their instructors, and conference organizers ask authors of submissions to evaluate the quality of the reviews. However, in these applications, students often give a higher rating to a course if they receive higher grades in a course, and authors often give a higher rating to the reviews if their papers are accepted to the conference. In this work, we call these external factors the "outcome" experienced by people, and consider the problem of mitigating these outcome-induced biases in the given ratings when some information about the outcome is available. We formulate the information about the outcome as a known partial ordering on the bias. We propose a debiasing method by solving a regularized optimization problem under this ordering constraint, and also provide a carefully designed cross-validation method that adaptively chooses the appropriate amount of regularization. We provide theoretical guarantees on the performance of our algorithm, as well as experimental evaluations.

Peer review is the backbone of academia and humans constitute a cornerstone of this process, being responsible for reviewing papers and making the final acceptance/rejection decisions. Given that human decision making is known to be susceptible to various cognitive biases, it is important to understand which (if any) biases are present in the peer-review process and design the pipeline such that the impact of these biases is minimized. In this work, we focus on the dynamics of between-reviewers discussions and investigate the presence of herding behaviour therein. In that, we aim to understand whether reviewers and more senior decision makers get disproportionately influenced by the first argument presented in the discussion when (in case of reviewers) they form an independent opinion about the paper before discussing it with others. Specifically, in conjunction with the review process of ICML 2020 -- a large, top tier machine learning conference -- we design and execute a randomized controlled trial with the goal of testing for the conditional causal effect of the discussion initiator's opinion on the outcome of a paper.

Conference peer review constitutes a human-computation process whose importance cannot be overstated: not only it identifies the best submissions for acceptance, but, ultimately, it impacts the future of the whole research area by promoting some ideas and restraining others. A surge in the number of submissions received by leading AI conferences has challenged the sustainability of the review process by increasing the burden on the pool of qualified reviewers which is growing at a much slower rate. In this work, we consider the problem of reviewer recruiting with a focus on the scarcity of qualified reviewers in large conferences. Specifically, we design a procedure for (i) recruiting reviewers from the population not typically covered by major conferences and (ii) guiding them through the reviewing pipeline. In conjunction with ICML 2020 -- a large, top-tier machine learning conference -- we recruit a small set of reviewers through our procedure and compare their performance with the general population of ICML reviewers. Our experiment reveals that a combination of the recruiting and guiding mechanisms allows for a principled enhancement of the reviewer pool and results in reviews of superior quality compared to the conventional pool of reviews as evaluated by senior members of the program committee (meta-reviewers).

Modern machine learning and computer science conferences are experiencing a surge in the number of submissions that challenges the quality of peer review as the number of competent reviewers is growing at a much slower rate. To curb this trend and reduce the burden on reviewers, several conferences have started encouraging or even requiring authors to declare the previous submission history of their papers. Such initiatives have been met with skepticism among authors, who raise the concern about a potential bias in reviewers' recommendations induced by this information. In this work, we investigate whether reviewers exhibit a bias caused by the knowledge that the submission under review was previously rejected at a similar venue, focusing on a population of novice reviewers who constitute a large fraction of the reviewer pool in leading machine learning and computer science conferences. We design and conduct a randomized controlled trial closely replicating the relevant components of the peer-review pipeline with $133$ reviewers (master's, junior PhD students, and recent graduates of top US universities) writing reviews for $19$ papers. The analysis reveals that reviewers indeed become negatively biased when they receive a signal about paper being a resubmission, giving almost 1 point lower overall score on a 10-point Likert item ($\Delta = -0.78, \ 95\% \ \text{CI} = [-1.30, -0.24]$) than reviewers who do not receive such a signal. Looking at specific criteria scores (originality, quality, clarity and significance), we observe that novice reviewers tend to underrate quality the most.

Quantifying systematic disparities in numerical quantities such as employment rates and wages between population subgroups provides compelling evidence for the existence of societal biases. However, biases in the text written for members of different subgroups (such as in recommendation letters for male and non-male candidates), though widely reported anecdotally, remain challenging to quantify. In this work, we introduce a novel framework to quantify bias in text caused by the visibility of subgroup membership indicators. We develop a nonparametric estimation and inference procedure to estimate this bias. We then formalize an identification strategy to causally link the estimated bias to the visibility of subgroup membership indicators, provided observations from time periods both before and after an identity-hiding policy change. We identify an application wherein "ground truth" bias can be inferred to evaluate our framework, instead of relying on synthetic or secondary data. Specifically, we apply our framework to quantify biases in the text of peer reviews from a reputed machine learning conference before and after the conference adopted a double-blind reviewing policy. We show evidence of biases in the review ratings that serves as "ground truth", and show that our proposed framework accurately detects these biases from the review text without having access to the review ratings.

We consider the issue of strategic behaviour in various peer-assessment tasks, including peer grading of exams or homeworks and peer review in hiring or promotions. When a peer-assessment task is competitive (e.g., when students are graded on a curve), agents may be incentivized to misreport evaluations in order to improve their own final standing. Our focus is on designing methods for detection of such manipulations. Specifically, we consider a setting in which agents evaluate a subset of their peers and output rankings that are later aggregated to form a final ordering. In this paper, we investigate a statistical framework for this problem and design a principled test for detecting strategic behaviour. We prove that our test has strong false alarm guarantees and evaluate its detection ability in practical settings. For this, we design and execute an experiment that elicits strategic behaviour from subjects and release a dataset of patterns of strategic behaviour that may be of independent interest. We then use the collected data to conduct a series of real and semi-synthetic evaluations that demonstrate a strong detection power of our test.

A major impediment to research on improving peer review is the unavailability of peer-review data, since any release of such data must grapple with the sensitivity of the peer review data in terms of protecting identities of reviewers from authors. We posit the need to develop techniques to release peer-review data in a privacy-preserving manner. Identifying this problem, in this paper we propose a framework for privacy-preserving release of certain conference peer-review data -- distributions of ratings, miscalibration, and subjectivity -- with an emphasis on the accuracy (or utility) of the released data. The crux of the framework lies in recognizing that a part of the data pertaining to the reviews is already available in public, and we use this information to post-process the data released by any privacy mechanism in a manner that improves the accuracy (utility) of the data while retaining the privacy guarantees. Our framework works with any privacy-preserving mechanism that operates via releasing perturbed data. We present several positive and negative theoretical results, including a polynomial-time algorithm for improving on the privacy-utility tradeoff.

We consider three important challenges in conference peer review: (i) reviewers maliciously attempting to get assigned to certain papers to provide positive reviews, possibly as part of quid-pro-quo arrangements with the authors; (ii) "torpedo reviewing," where reviewers deliberately attempt to get assigned to certain papers that they dislike in order to reject them; (iii) reviewer de-anonymization on release of the similarities and the reviewer-assignment code. On the conceptual front, we identify connections between these three problems and present a framework that brings all these challenges under a common umbrella. We then present a (randomized) algorithm for reviewer assignment that can optimally solve the reviewer-assignment problem under any given constraints on the probability of assignment for any reviewer-paper pair. We further consider the problem of restricting the joint probability that certain suspect pairs of reviewers are assigned to certain papers, and show that this problem is NP-hard for arbitrary constraints on these joint probabilities but efficiently solvable for a practical special case. Finally, we experimentally evaluate our algorithms on datasets from past conferences, where we observe that they can limit the chance that any malicious reviewer gets assigned to their desired paper to 50% while producing assignments with over 90% of the total optimal similarity. Our algorithms still achieve this similarity while also preventing reviewers with close associations from being assigned to the same paper.

A number of applications involve sequential arrival of users, and require showing each user an ordering of items. A prime example (which forms the focus of this paper) is the bidding process in conference peer review where reviewers enter the system sequentially, each reviewer needs to be shown the list of submitted papers, and the reviewer then "bids" to review some papers. The order of the papers shown has a significant impact on the bids due to primacy effects. In deciding on the ordering of papers to show, there are two competing goals: (i) obtaining sufficiently many bids for each paper, and (ii) satisfying reviewers by showing them relevant items. In this paper, we begin by developing a framework to study this problem in a principled manner. We present an algorithm called SUPER*, inspired by the A* algorithm, for this goal. Theoretically, we show a local optimality guarantee of our algorithm and prove that popular baselines are considerably suboptimal. Moreover, under a community model for the similarities, we prove that SUPER* is near-optimal whereas the popular baselines are considerably suboptimal. In experiments on real data from ICLR 2018 and synthetic data, we find that SUPER* considerably outperforms baselines deployed in existing systems, consistently reducing the number of papers with fewer than requisite bids by 50-75% or more, and is also robust to various real world complexities.

A number of applications require two-sample testing on ranked preference data. For instance, in crowdsourcing, there is a long-standing question of whether pairwise comparison data provided by people is distributed similar to ratings-converted-to-comparisons. Other examples include sports data analysis and peer grading. In this paper, we design two-sample tests for pairwise comparison data and ranking data. For our two-sample test for pairwise comparison data, we establish an upper bound on the sample complexity required to correctly distinguish between the distributions of the two sets of samples. Our test requires essentially no assumptions on the distributions. We then prove complementary lower bounds showing that our results are tight (in the minimax sense) up to constant factors. We investigate the role of modeling assumptions by proving lower bounds for a range of pairwise comparison models (WST, MST,SST, parameter-based such as BTL and Thurstone). We also provide testing algorithms and associated sample complexity bounds for the problem of two-sample testing with partial (or total) ranking data.Furthermore, we empirically evaluate our results via extensive simulations as well as two real-world datasets consisting of pairwise comparisons. By applying our two-sample test on real-world pairwise comparison data, we conclude that ratings and rankings provided by people are indeed distributed differently. On the other hand, our test recognizes no significant difference in the relative performance of European football teams across two seasons. Finally, we apply our two-sample test on a real-world partial and total ranking dataset and find a statistically significant difference in Sushi preferences across demographic divisions based on gender, age and region of residence., Comment: 40 pages, 4 figures

We consider the issue of biases in scholarly research, specifically, in peer review. There is a long standing debate on whether exposing author identities to reviewers induces biases against certain groups, and our focus is on designing tests to detect the presence of such biases. Our starting point is a remarkable recent work by Tomkins, Zhang and Heavlin which conducted a controlled, large-scale experiment to investigate existence of biases in the peer reviewing of the WSDM conference. We present two sets of results in this paper. The first set of results is negative, and pertains to the statistical tests and the experimental setup used in the work of Tomkins et al. We show that the test employed therein does not guarantee control over false alarm probability and under correlations between relevant variables coupled with any of the following conditions, with high probability, can declare a presence of bias when it is in fact absent: (a) measurement error, (b) model mismatch, (c) reviewer calibration. Moreover, we show that the setup of their experiment may itself inflate false alarm probability if (d) bidding is performed in non-blind manner or (e) popular reviewer assignment procedure is employed. Our second set of results is positive and is built around a novel approach to testing for biases that we propose. We present a general framework for testing for biases in (single vs. double blind) peer review. We then design hypothesis tests that under minimal assumptions guarantee control over false alarm probability and non-trivial power even under conditions (a)--(c) as well as propose an alternative experimental setup which mitigates issues (d) and (e). Finally, we show that no statistical test can improve over the non-parametric tests we consider in terms of the assumptions required to control for the false alarm probability.

A number of applications (e.g., AI bot tournaments, sports, peer grading, crowdsourcing) use pairwise comparison data and the Bradley-Terry-Luce (BTL) model to evaluate a given collection of items (e.g., bots, teams, students, search results). Past work has shown that under the BTL model, the widely-used maximum-likelihood estimator (MLE) is minimax-optimal in estimating the item parameters, in terms of the mean squared error. However, another important desideratum for designing estimators is fairness. In this work, we consider fairness modeled by the notion of bias in statistics. We show that the MLE incurs a suboptimal rate in terms of bias. We then propose a simple modification to the MLE, which "stretches" the bounding box of the maximum-likelihood optimizer by a small constant factor from the underlying ground truth domain. We show that this simple modification leads to an improved rate in bias, while maintaining minimax-optimality in the mean squared error. In this manner, our proposed class of estimators provably improves fairness represented by bias without loss in accuracy.

Vehicular mobile crowd sensing is a fast-emerging paradigm to collect data about the environment by mounting sensors on vehicles such as taxis. An important problem in vehicular crowd sensing is to design payment mechanisms to incentivize drivers (agents) to collect data, with the overall goal of obtaining the maximum amount of data (across multiple vehicles) for a given budget. Past works on this problem consider a setting where each agent operates in isolation---an assumption which is frequently violated in practice. In this paper, we design an incentive mechanism to incentivize agents who can engage in arbitrary collusions. We then show that in a "homogeneous" setting, our mechanism is optimal, and can do as well as any mechanism which knows the agents' preferences a priori. Moreover, if the agents are non-colluding, then our mechanism automatically does as well as any other non-colluding mechanism. We also show that our proposed mechanism has strong (and asymptotically optimal) guarantees for a more general "heterogeneous" setting. Experiments based on synthesized data and real-world data reveal gains of over 30\% attained by our mechanism compared to past literature.

It is common to see a handful of reviewers reject a highly novel paper, because they view, say, extensive experiments as far more important than novelty, whereas the community as a whole would have embraced the paper. More generally, the disparate mapping of criteria scores to final recommendations by different reviewers is a major source of inconsistency in peer review. In this paper we present a framework inspired by empirical risk minimization (ERM) for learning the community's aggregate mapping. The key challenge that arises is the specification of a loss function for ERM. We consider the class of $L(p,q)$ loss functions, which is a matrix-extension of the standard class of $L_p$ losses on vectors; here the choice of the loss function amounts to choosing the hyperparameters $p, q \in [1,\infty]$. To deal with the absence of ground truth in our problem, we instead draw on computational social choice to identify desirable values of the hyperparameters $p$ and $q$. Specifically, we characterize $p=q=1$ as the only choice of these hyperparameters that satisfies three natural axiomatic properties. Finally, we implement and apply our approach to reviews from IJCAI 2017.

We consider the problem of automated assignment of papers to reviewers in conference peer review, with a focus on fairness and statistical accuracy. Our fairness objective is to maximize the review quality of the most disadvantaged paper, in contrast to the commonly used objective of maximizing the total quality over all papers. We design an assignment algorithm based on an incremental max-flow procedure that we prove is near-optimally fair. Our statistical accuracy objective is to ensure correct recovery of the papers that should be accepted. We provide a sharp minimax analysis of the accuracy of the peer-review process for a popular objective-score model as well as for a novel subjective-score model that we propose in the paper. Our analysis proves that our proposed assignment algorithm also leads to a near-optimal statistical accuracy. Finally, we design a novel experiment that allows for an objective comparison of various assignment algorithms, and overcomes the inherent difficulty posed by the absence of a ground truth in experiments on peer-review. The results of this experiment as well as of other experiments on synthetic and real data corroborate the theoretical guarantees of our algorithm.

We consider peer review in a conference setting where there is typically an overlap between the set of reviewers and the set of authors. This overlap can incentivize strategic reviews to influence the final ranking of one's own papers. In this work, we address this problem through the lens of social choice, and present a theoretical framework for strategyproof and efficient peer review. We first present and analyze an algorithm for reviewer-assignment and aggregation that guarantees strategyproofness and a natural efficiency property called unanimity, when the authorship graph satisfies a simple property. Our algorithm is based on the so-called partitioning method, and can be thought as a generalization of this method to conference peer review settings. We then empirically show that the requisite property on the authorship graph is indeed satisfied in the submission data from the ICLR conference, and further demonstrate a simple trick to make the partitioning method more practically appealing for conference peer review. Finally, we complement our positive results with negative theoretical results where we prove that under various ways of strengthening the requirements, it is impossible for any algorithm to be strategyproof and efficient.

Cardinal scores (numeric ratings) collected from people are well known to suffer from miscalibrations. A popular approach to address this issue is to assume simplistic models of miscalibration (such as linear biases) to de-bias the scores. This approach, however, often fares poorly because people's miscalibrations are typically far more complex and not well understood. In the absence of simplifying assumptions on the miscalibration, it is widely believed by the crowdsourcing community that the only useful information in the cardinal scores is the induced ranking. In this paper, inspired by the framework of Stein's shrinkage, empirical Bayes, and the classic two-envelope problem, we contest this widespread belief. Specifically, we consider cardinal scores with arbitrary (or even adversarially chosen) miscalibrations which are only required to be consistent with the induced ranking. We design estimators which despite making no assumptions on the miscalibration, strictly and uniformly outperform all possible estimators that rely on only the ranking. Our estimators are flexible in that they can be used as a plug-in for a variety of applications, and we provide a proof-of-concept for A/B testing and ranking. Our results thus provide novel insights in the eternal debate between cardinal and ordinal data.

Neural Information Processing Systems (NIPS) is a top-tier annual conference in machine learning. The 2016 edition of the conference comprised more than 2,400 paper submissions, 3,000 reviewers, and 8,000 attendees. This represents a growth of nearly 40% in terms of submissions, 96% in terms of reviewers, and over 100% in terms of attendees as compared to the previous year. The massive scale as well as rapid growth of the conference calls for a thorough quality assessment of the peer-review process and novel means of improvement. In this paper, we analyze several aspects of the data collected during the review process, including an experiment investigating the efficacy of collecting ordinal rankings from reviewers. Our goal is to check the soundness of the review process, and provide insights that may be useful in the design of the review process of subsequent conferences.

We consider the problem of noisy matrix completion, in which the goal is to reconstruct a structured matrix whose entries are partially observed in noise. Standard approaches to this underdetermined inverse problem are based on assuming that the underlying matrix has low rank, or is well-approximated by a low rank matrix. In this paper, we propose a richer model based on what we term the "permutation-rank" of a matrix. We first describe how the classical non-negative rank model enforces restrictions that may be undesirable in practice, and how and these restrictions can be avoided by using the richer permutation-rank model. Second, we establish the minimax rates of estimation under the new permutation-based model, and prove that surprisingly, the minimax rates are equivalent up to logarithmic factors to those for estimation under the typical low rank model. Third, we analyze a computationally efficient singular-value-thresholding algorithm, known to be optimal for the low-rank setting, and show that it also simultaneously yields a consistent estimator for the low-permutation rank setting. Finally, we present various structural results characterizing the uniqueness of the permutation-rank decomposition, and characterizing convex approximations of the permutation-rank polytope.

The task of aggregating and denoising crowd-labeled data has gained increased significance with the advent of crowdsourcing platforms and massive datasets. We propose a permutation-based model for crowd labeled data that is a significant generalization of the classical Dawid-Skene model, and introduce a new error metric by which to compare different estimators. We derive global minimax rates for the permutation-based model that are sharp up to logarithmic factors, and match the minimax lower bounds derived under the simpler Dawid-Skene model. We then design two computationally-efficient estimators: the WAN estimator for the setting where the ordering of workers in terms of their abilities is approximately known, and the OBI-WAN estimator where that is not known. For each of these estimators, we provide non-asymptotic bounds on their performance. We conduct synthetic simulations and experiments on real-world crowdsourcing data, and the experimental results corroborate our theoretical findings., Comment: in IEEE Transactions on Information Theory (online), 2020

We consider sequential or active ranking of a set of n items based on noisy pairwise comparisons. Items are ranked according to the probability that a given item beats a randomly chosen item, and ranking refers to partitioning the items into sets of pre-specified sizes according to their scores. This notion of ranking includes as special cases the identification of the top-k items and the total ordering of the items. We first analyze a sequential ranking algorithm that counts the number of comparisons won, and uses these counts to decide whether to stop, or to compare another pair of items, chosen based on confidence intervals specified by the data collected up to that point. We prove that this algorithm succeeds in recovering the ranking using a number of comparisons that is optimal up to logarithmic factors. This guarantee does not require any structural properties of the underlying pairwise probability matrix, unlike a significant body of past work on pairwise ranking based on parametric models such as the Thurstone or Bradley-Terry-Luce models. It has been a long-standing open question as to whether or not imposing these parametric assumptions allows for improved ranking algorithms. For stochastic comparison models, in which the pairwise probabilities are bounded away from zero, our second contribution is to resolve this issue by proving a lower bound for parametric models. This shows, perhaps surprisingly, that these popular parametric modeling choices offer at most logarithmic gains for stochastic comparisons., Comment: improved log factor in main result; added discussion on comparison probabilities close to zero; added numerical results

We study methods for aggregating pairwise comparison data in order to estimate outcome probabilities for future comparisons among a collection of n items. Working within a flexible framework that imposes only a form of strong stochastic transitivity (SST), we introduce an adaptivity index defined by the indifference sets of the pairwise comparison probabilities. In addition to measuring the usual worst-case risk of an estimator, this adaptivity index also captures the extent to which the estimator adapts to instance-specific difficulty relative to an oracle estimator. We prove three main results that involve this adaptivity index and different algorithms. First, we propose a three-step estimator termed Count-Randomize-Least squares (CRL), and show that it has adaptivity index upper bounded as $\sqrt{n}$ up to logarithmic factors. We then show that that conditional on the hardness of planted clique, no computationally efficient estimator can achieve an adaptivity index smaller than $\sqrt{n}$. Second, we show that a regularized least squares estimator can achieve a poly-logarithmic adaptivity index, thereby demonstrating a $\sqrt{n}$-gap between optimal and computationally achievable adaptivity. Finally, we prove that the standard least squares estimator, which is known to be optimally adaptive in several closely related problems, fails to adapt in the context of estimating pairwise probabilities.

We consider a problem of prediction based on opinions elicited from heterogeneous rational agents with private information. Making an accurate prediction with a minimal cost requires a joint design of the incentive mechanism and the prediction algorithm. Such a problem lies at the nexus of statistical learning theory and game theory, and arises in many domains such as consumer surveys and mobile crowdsourcing. In order to elicit heterogeneous agents' private information and incentivize agents with different capabilities to act in the principal's best interest, we design an optimal joint incentive mechanism and prediction algorithm called COPE (COst and Prediction Elicitation), the analysis of which offers several valuable engineering insights. First, when the costs incurred by the agents are linear in the exerted effort, COPE corresponds to a "crowd contending" mechanism, where the principal only employs the agent with the highest capability. Second, when the costs are quadratic, COPE corresponds to a "crowd-sourcing" mechanism that employs multiple agents with different capabilities at the same time. Numerical simulations show that COPE improves the principal's profit and the network profit significantly (larger than 30% in our simulations), comparing to those mechanisms that assume all agents have equal capabilities.

We consider data in the form of pairwise comparisons of n items, with the goal of precisely identifying the top k items for some value of k < n, or alternatively, recovering a ranking of all the items. We analyze the Copeland counting algorithm that ranks the items in order of the number of pairwise comparisons won, and show it has three attractive features: (a) its computational efficiency leads to speed-ups of several orders of magnitude in computation time as compared to prior work; (b) it is robust in that theoretical guarantees impose no conditions on the underlying matrix of pairwise-comparison probabilities, in contrast to some prior work that applies only to the BTL parametric model; and (c) it is an optimal method up to constant factors, meaning that it achieves the information-theoretic limits for recovering the top k-subset. We extend our results to obtain sharp guarantees for approximate recovery under the Hamming distortion metric, and more generally, to any arbitrary error requirement that satisfies a simple and natural monotonicity condition., Comment: Changes in version 2: In addition to recovery in the exact and Hamming metrics, v2 analyzes a general, abstract recovery criterion based on a notion of "allowed sets"

There are various parametric models for analyzing pairwise comparison data, including the Bradley-Terry-Luce (BTL) and Thurstone models, but their reliance on strong parametric assumptions is limiting. In this work, we study a flexible model for pairwise comparisons, under which the probabilities of outcomes are required only to satisfy a natural form of stochastic transitivity. This class includes parametric models including the BTL and Thurstone models as special cases, but is considerably more general. We provide various examples of models in this broader stochastically transitive class for which classical parametric models provide poor fits. Despite this greater flexibility, we show that the matrix of probabilities can be estimated at the same rate as in standard parametric models. On the other hand, unlike in the BTL and Thurstone models, computing the minimax-optimal estimator in the stochastically transitive model is non-trivial, and we explore various computationally tractable alternatives. We show that a simple singular value thresholding algorithm is statistically consistent but does not achieve the minimax rate. We then propose and study algorithms that achieve the minimax rate over interesting sub-classes of the full stochastically transitive class. We complement our theoretical results with thorough numerical simulations.