EN FR
EN FR

2025‌Activity reportProject-TeamFAIRPLAY‌​‌

RNSR: 202224251U

Creation of the​​​‌ Project-Team: 2022 March 01‌

Each year, Inria research‌​‌ teams publish an Activity​​ Report presenting their work​​​‌ and results over the‌ reporting period. These reports‌​‌ follow a common structure,​​ with some optional sections​​​‌ depending on the specific‌ team. They typically begin‌​‌ by outlining the overall​​ objectives and research programme,​​​‌ including the main research‌ themes, goals, and methodological‌​‌ approaches. They also describe​​ the application domains targeted​​​‌ by the team, highlighting‌ the scientific or societal‌​‌ contexts in which their​​ work is situated.

The​​​‌ reports then present the‌ highlights of the year,‌​‌ covering major scientific achievements,​​ software developments, or teaching​​​‌ contributions. When relevant, they‌ include sections on software,‌​‌ platforms, and open data,​​ detailing the tools developed​​​‌ and how they are‌ shared. A substantial part‌​‌ is dedicated to new​​ results, where scientific contributions​​​‌ are described in detail,‌ often with subsections specifying‌​‌ participants and associated keywords.​​​‌

Finally, the Activity Report​ addresses funding, contracts, partnerships,​‌ and collaborations at various​​ levels, from industrial agreements​​​‌ to international cooperations. It​ also covers dissemination and​‌ teaching activities, such as​​ participation in scientific events,​​​‌ outreach, and supervision. The​ document concludes with a​‌ presentation of scientific production,​​ including major publications and​​​‌ those produced during the​ year.

Keywords

Computer Science​‌ and Digital Science

  • A4.8.​​ Privacy-enhancing technologies
  • A8.11. Game​​​‌ Theory
  • A9.2. Machine learning​
  • A9.9. Distributed AI, Multi-agent​‌

Other Research Topics and​​ Application Domains

  • B9.9. Ethics​​​‌
  • B9.10. Privacy

1 Team​ members, visitors, external collaborators​‌

Research Scientists

  • Patrick Loiseau​​ [Team leader,​​​‌ INRIA, Senior Researcher​, HDR]
  • Marc​‌ Abeille [CRITEO,​​ Industrial member]
  • Benjamin​​​‌ Heymann [CRITEO,​ Industrial member]
  • Simon​‌ Mauras [INRIA,​​ Researcher]
  • Hugo Richard​​​‌ [CRITEO, Industrial​ member]
  • Mariia Vladimirova​‌ [CRITEO, Industrial​​ member]
  • Maxime Vono​​​‌ [CRITEO, Industrial​ member]

Faculty Members​‌

  • Matthieu Lerasle [ENSAE​​, Professor]
  • Vianney​​​‌ Perchet [CRITEO,​ Professor, HDR]​‌
  • Cristina Rio Butucea [​​GENES, Professor]​​​‌

Post-Doctoral Fellows

  • Achraf Azize​ [ENSAE]
  • Lorenzo​‌ Croissant [ENSAE]​​
  • Simon Finster [INRIA​​​‌, Post-Doctoral Fellow,​ until Sep 2025]​‌
  • Junghun Kim [ENSAE​​]
  • Denis Sokolov [​​​‌INRIA, Post-Doctoral Fellow​, until Nov 2025​‌]
  • Bartholome Vieille [​​INRIA, Post-Doctoral Fellow​​​‌]

PhD Students

  • Louise​ Allain [INRIA,​‌ from Oct 2025]​​
  • Ahmed Ben Yahmed [​​​‌CRITEO, CIFRE]​
  • Ziyad Benomar [ENSAE​‌, until Sep 2025​​]
  • Maria Cherifa [​​​‌CRITEO, CIFRE]​
  • Hafedh El Ferchichi [​‌ENSAE]
  • Come Fiegel​​ [ENSAE]
  • Reda​​​‌ Jalal [INRIA]​
  • Mike Liu [ENSAE​‌]
  • Mathieu Molina [​​INRIA, until Sep​​​‌ 2025]
  • Giovanni Montanari​ [INRIA]
  • Corentin​‌ Pla [CRITEO,​​ CIFRE]
  • Marius Potfer​​​‌ [EDF, CIFRE​]
  • Naila Sebastián Esandi​‌ [ENSAE, from​​ Oct 2025]
  • Mélissa​​​‌ Tamine [CRITEO,​ CIFRE]
  • Matilde Tullii​‌ [ENSAE]
  • Remi​​ Verroneau [FOOTOVISION,​​​‌ CIFRE]
  • Axel Xerri​ [INRIA]
  • Minrui​‌ Xu [ENSAE]​​

Technical Staff

  • Eugenie Patard​​​‌ [INRIA, Engineer​, from Nov 2025​‌]

Interns and Apprentices​​

  • Naila Carmen Sebastian Esandi​​​‌ [INRIA, Intern​, from May 2025​‌ until Sep 2025]​​

Administrative Assistant

  • Melanie Da​​​‌ Silva [INRIA]​

External Collaborator

  • Clément Calauzenes​‌ [CRITEO]

2​​ Overall objectives

2.1 Broad​​​‌ context

One of the​ principal objectives of Machine​‌ Learning (ML) is to​​ automatically discover using past​​​‌ data some underlying structure​ behind a data generating​‌ process in order either​​ to explain past observations​​​‌ or, perhaps more importantly,​ to make predictions and/or​‌ to optimize decisions made​​ on future instances. The​​​‌ area of ML has​ exploded over the past​‌ decade and has had​​ a tremendous impact in​​​‌ many application domains such​ as computer vision or​‌ bioinformatics.

Most of the​​ current ML literature focuses​​ on the case of​​​‌ a single agent (an‌ algorithm) trying to complete‌​‌ some learning task based​​ on gathered data that​​​‌ follows an exogenous distribution‌ independent of the algorithm.‌​‌ One of the key​​ assumptions is that this​​​‌ data has sufficient “regularity”‌ for classical techniques to‌​‌ work. This classical paradigm​​ of “a single agent​​​‌ learning on nice data”,‌ however, is no longer‌​‌ adequate for many practical​​ and crucial tasks that​​​‌ imply users (who own‌ the gathered data) and/or‌​‌ other (learning) agents that​​ are also trying to​​​‌ optimize their own objectives‌ simultaneously, in a competitive‌​‌ or conflicting way. This​​ is the case, for​​​‌ instance, in most learning‌ tasks related to Internet‌​‌ applications (content recommendation/ranking, ad​​ auctions, fraud detection, etc.).​​​‌ Moreover, as such learning‌ tasks rely on users'‌​‌ personal data and as​​ their outcome affect users​​​‌ in return, it is‌ no longer sufficient to‌​‌ focus on optimizing prediction​​ performance metrics—it becomes crucial​​​‌ to consider societal and‌ ethical aspects such as‌​‌ fairness or privacy.

The​​ field of single agent​​​‌ ML builds on techniques‌ from domains such as‌​‌ statistics, optimization, or functional​​ analysis. When different agents​​​‌ are involved, a strategic‌ aspect inherent in game‌​‌ theory enters the picture.​​ Indeed, interactions—either positive or​​​‌ negative—between rational entities (firms,‌ single user at home,‌​‌ algorithms, etc.) foster individual​​ strategic behavior such as​​​‌ hiding information, misleading other‌ agents, free-riding, etc. Unfortunately,‌​‌ this selfishness degrades the​​ quality of the data​​​‌ or of the predictions,‌ prevents efficient learning and‌​‌ overall may diminish the​​ social welfare. These strategic​​​‌ aspects, together with the‌ decentralized nature of decision‌​‌ making in a multi-agent​​ environment, also make it​​​‌ harder to build algorithms‌ that meet fairness and‌​‌ privacy constraints.

The overarching​​ objective of FAIRPLAY is​​​‌ to create algorithms that‌ learn for and with‌​‌ users—and techniques to analyze​​ them—, that is​​​‌ to create procedures able‌ to perform classical learning‌​‌ tasks (prediction, decision, explanation)​​ when the data is​​​‌ generated or provided by‌ strategic agents, possibly in‌​‌ the presence of other​​ competing learning agents, while​​​‌ respecting the fairness and‌ privacy of the involved‌​‌ users. To that end,​​ we will naturally rely​​​‌ on multi-agent models where‌ the different agents may‌​‌ be either agents generating​​ or providing data, or​​​‌ agents learning in a‌ way that interacts with‌​‌ other agents; and we​​ will put a special​​​‌ focus on societal and‌ ethical aspects, in particular‌​‌ fairness and privacy. Note​​ that in FAIRPLAY, we​​​‌ focus on the technical‌ challenges inherent to formalizing‌​‌ mathematically and respecting ethical​​ properties such as non-discrimination​​​‌ or privacy, often seen‌ as constraints in the‌​‌ learning procedure. Nevertheless, throughout​​ the team's life, we​​​‌ will reflect on these‌ mathematical definitions for the‌​‌ particular applications studied, in​​ particular their philosophical roots​​​‌ and legal interpretation, through‌ interactions with HSS researchers‌​‌ and with legal specialists​​ (from Criteo).

2.1.1 Multi-agent​​​‌ systems

Any company developing‌ and implementing ML algorithms‌​‌ is in fact one​​ agent within a large​​​‌ network of users and‌ other firms. Assuming that‌​‌ the data is i.i.d.​​​‌ and can be treated​ irrespectively of the environment​‌ response—as is done in​​ the classical ML paradigm—might​​​‌ be a good first​ approximation, but should be​‌ overcome. Users, clients, suppliers,​​ and competitors are adaptive​​​‌ and change their behavior​ depending on each other's​‌ interactions. The future of​​ many ML companies—such as​​​‌ Criteo—will consist in creating​ platforms matching the demand​‌ (created by their users)​​ to the offer (proposed​​​‌ by their clients), under​ the system constraints (imposed​‌ by suppliers and competitors).​​ Each of these agents​​​‌ have different, conflicting interests​ that should be taken​‌ into account in the​​ model, which naturally becomes​​​‌ a multi-agent model.

Each​ agent in a multi-agent​‌ system may be modeled​​ as having their own​​​‌ utility function ui​ that can depend on​‌ the action of other​​ agents. Then, there are​​​‌ two main types of​ objectives: individual or collective​‌ 104. If each​​ agent is making their​​​‌ own decision, then they​ can be modeled as​‌ each optimizing their own​​ individual utility (which may​​​‌ include personal benefit as​ well as other considerations​‌ such as altruism where​​ appropriate) unilaterally and in​​​‌ a decentralized way. This​ is why a mechanism​‌ providing correct incentives to​​ agents is often necessary.​​​‌ At the other extreme,​ social welfare is the​‌ collective objective defined as​​ the cumulative sum of​​​‌ utilities of all agents.​ To optimize it, it​‌ is almost always necessary​​ to consider a centralized​​​‌ optimization or learning protocol.​ A key question in​‌ multi-agent systems is to​​ apprehend the “social cost”​​​‌ of letting agents optimize​ their own utility by​‌ choosing unilaterally their decision​​ compared to the one​​​‌ maximizing social welfare; this​ is often measured by​‌ the “price of anarchy”/“price​​ of stability” 113:​​​‌ the ratio of the​ maximum social welfare to​‌ the (worst/best) social welfare​​ when agents optimize individually.​​​‌

The natural language to​ model and study multi-agent​‌ systems is game theory​​—see below for a​​​‌ list of tools and​ techniques on which FAIRPLAY​‌ relies, game theory being​​ the first of them.​​​‌ Multi-agent systems have been​ studied in the past;​‌ but not with a​​ focus on learning systems​​​‌ where agents are either​ learning or providing data,​‌ which is our focus​​ in FAIRPLAY and leads​​​‌ to a blend of​ game theory and learning​‌ techniques. We note here​​ again that, wherever appropriate,​​​‌ we shall reflect (in​ part together with colleagues​‌ from HSS) on the​​ soundness of the utility​​​‌ framework for the considered​ applications.

2.1.2 Societal aspects​‌ and ethics

There are​​ several important ethical aspects​​​‌ that must be investigated​ in multi-agent systems involving​‌ users either as data​​ providers or as individuals​​​‌ affected by the ML​ agent decision (or both).​‌

Fairness and Discrimination

When​​ ML decisions directly affect​​​‌ humans, it is important​ to ensure that they​‌ do not violate fairness​​ principles, be they based​​​‌ on ethical or legal​ grounds. As ML made​‌ its way in many​​ areas of decision making,​​​‌ it was unfortunately repeatedly​ observed that it can​‌ lead to discrimination (regardless​​ of whether or not​​ it is intentional) based​​​‌ on gender, race, age,‌ or other sensitive attributes.‌​‌ This was observed in​​ online targeted advertisement 98​​​‌, 124, 38‌, 82, 98‌​‌, 40, but​​ also in many other​​​‌ applications such as hiring‌ 69, data-driven healthcare‌​‌ 76, or justice​​ 99. Biases also​​​‌ have the unfortunate tendency‌ to reinforce. An operating‌​‌ multi-agent learning system should​​ be able in the​​​‌ long run to get‌ rid by itself of‌​‌ inherent population biases, that​​ is, be fair amongst​​​‌ users irrespective of the‌ improperly constructed dataset.

The‌​‌ mathematical formulation of fairness​​ has been debated in​​​‌ recent works. Although a‌ few initial works proposed‌​‌ a notion of individual​​ fairness, which mandates​​​‌ that “similar individuals” receive‌ “similar outcomes” 71,‌​‌ this notion was quickly​​ found unpractical because it​​​‌ relies on a metric‌ to define closeness that‌​‌ makes the definition somewhat​​ arbitrary. Most of the​​​‌ works then focused on‌ notions of group fairness‌​‌, which mandate equality​​ of outcome “on average”​​​‌ across different groups defined‌ by sensitive attributes (e.g.,‌​‌ race, gender, religious belief,​​ etc.). Most of the​​​‌ works on group fairness‌ focus on the classification‌​‌ problem (e.g., classifying whether​​ a job applicant is​​​‌ good or not for‌ the job) where each‌​‌ data example (X​​i,Yi​​​‌) contains a set‌ of features Xi‌​‌ and a true label​​ Yi{​​​‌0,1}‌ and the goal is‌​‌ to make a prediction​​ Y^i based​​​‌ on the features X‌i that has a‌​‌ high probability to be​​ equal to the true​​​‌ label. Assuming that there‌ is a single sensitive‌​‌ attribute si that​​ can take two values​​​‌ a or b,‌ this defines two groups:‌​‌ those for whom s​​i=a and​​​‌ those for whom s‌i=b.‌​‌ There are several different​​ concepts of group fairness​​​‌ that can be considered;‌ we shall especially focus‌​‌ on demographic parity (DP),​​ which prescribes P(​​​‌Y^i=‌1|si‌​‌=a)=​​P(Y^​​​‌i=1|‌si=b‌​‌) and equal opportunity​​ (EO) 83, which​​​‌ mandates that P(‌Y^i=‌​‌1|si​​=a,Y​​​‌i=1)‌=P(Y‌​‌^i=1​​|si=​​​‌b,Yi‌=1).‌​‌

The fair classification literature​​ proposed, for each of​​​‌ these fairness notions, ways‌ to train fair classifiers‌​‌ based on three main​​ ideas: pre-processing 132,​​​‌ in-processing 130, 131‌, 127, and‌​‌ post-processing 83. All​​ of these works, however,​​​‌ focus on idealized situations‌ where a single decision-maker‌​‌ has access to ground​​ truth data with the​​​‌ sensitive features and labels‌ in order to train‌​‌ classifiers that respect fairness​​ constraints. We use similar​​​‌ group fairness definitions and‌ extend them (in particular‌​‌ through causality), but our​​​‌ goal is to go​ further in terms of​‌ algorithms by modeling practical​​ scenarios with multiple decision-makers​​​‌ and incomplete information (in​ particular lack of ground​‌ truth on the labels).​​

Privacy vs. Incentives

ML​​​‌ algorithms, in particular in​ Internet applications, often rely​‌ on users' personal information​​ (whether it is directly​​​‌ their personal data or​ indirectly some hidden “type”​‌ – gender, ethnicity, behaviors,​​ etc.). Nevertheless, users may​​​‌ be willing to provide​ their personal information if​‌ it increases their utility.​​ This brings a number​​​‌ of key questions. First,​ how can we learn​‌ while protecting users' privacy​​ (and how should privacy​​​‌ even be defined)? Second,​ finding the right balance​‌ between those two a-priori​​ incompatible concepts is challenging;​​​‌ how much (and even​ simply how) should an​‌ agent be compensated for​​ providing useful and accurate​​​‌ data?

Differential privacy is​ the most widely used​‌ private learning framework 70​​, 72, 119​​​‌ and ensures that the​ output of an algorithm​‌ does not significantly depend​​ on a single element​​​‌ of the whole dataset.​ These privacy constraints are​‌ often too strong for​​ economic applications (as illustrated​​​‌ before, it is sometimes​ optimal to disclose some​‌ private information). f-divergence​​ privacy costs have thus​​​‌ been proposed in recent​ literature as a promising​‌ alternative 62. These​​ f-divergences, such as​​​‌ Kullback-Leibler, are also used​ by economists to measure​‌ the cost of information​​ from a Bayesian perspective,​​​‌ as in the rational​ inattention literature 123,​‌ 106, 101.​​ It was only recently​​​‌ that this approach was​ considered to measure “privacy​‌ losses” in economic mechanisms​​ 73. In this​​​‌ model, the mechanism designer​ has some prior belief​‌ on the unobserved and​​ private information. After observing​​​‌ the player's action, this​ belief is updated and​‌ the cost of information​​ corresponds to the KL​​​‌ between the prior and​ posterior distributions of this​‌ private information.

This privacy​​ concept can be refined​​​‌ up to a single​ user level, into the​‌ so-called local differential privacy.​​ Informally speaking, the algorithm​​​‌ output can also depend​ on a single user​‌ data that still must​​ be kept private. Estimation​​​‌ are actually sometimes more​ challenging under this constraint,​‌ i.e., estimation rates degrade​​ 120, 57,​​​‌ 58 but is sometimes​ more adapted to handle​‌ user-generated data 78.​​

Interestingly, we note that​​​‌ the notions of privacy​ and fairness are somewhat​‌ incompatible. This will motivate​​ Theme 2 developed in​​​‌ our research program.

2.2​ A large variety of​‌ tools and techniques

Analyzing​​ multi-agent learning systems with​​​‌ ethical constraints will require​ us to use, develop,​‌ and merge several different​​ theoretical tools and techniques.​​​‌ We describe the main​ ones here. Note that​‌ although FAIRPLAY is motivated​​ by practical use-cases and​​​‌ applications, part of the​ team's objectives is to​‌ improve those tools as​​ necessary to tackle the​​​‌ problems studied.

Game theory​ and economics

Game theory​‌ 77 is the natural​​ mathematical tool to model​​​‌ multiple interacting decision-makers (called​ players). A game is​‌ defined by a set​​ of players, a set​​ of possible actions for​​​‌ each player, and a‌ payoff function for each‌​‌ player that can depend​​ on the actions of​​​‌ all the players (that‌ is the distinguishing feature‌​‌ of a game compared​​ to an optimization problem).​​​‌ The most standard solution‌ concept is the so-called‌​‌ Nash equilibrium, which is​​ defined as a strategy​​​‌ profile (i.e., a collection‌ of possibly randomized action‌​‌ for each player) such​​ that each player is​​​‌ at best response (i.e.,‌ has the maximum payoff‌​‌ given the others' strategies).​​ It is a “static”​​​‌ (one-shot) solution concept, but‌ there also exist dynamic‌​‌ solution concepts for repeated​​ games 61, 108​​​‌.

Online and reinforcement‌ learning 54

In online‌​‌ learning (a.k.a. multi-armed bandit​​ 55, 115),​​​‌ data is gathered and‌ treated on the fly.‌​‌ For instance, consider an​​ online binary classification problem.​​​‌ Some unlabelled data X‌tRd‌​‌ is observed, and the​​ agent predicts its label​​​‌ Yt; let‌ us denote Y^‌​‌t±1​​ the prediction. The agent​​​‌ potentially observes the loss‌ 1{Yt‌​‌Y^t​​} and then receives​​​‌ another new unlabeled data‌ example Xt+‌​‌1. In that​​ specific problem, the typical​​​‌ learning objective is to‌ perform asymptotically as good‌​‌ as the best classifier​​ f* in some​​​‌ given class ,‌ i.e., such that the‌​‌ loss t=​​1T1{​​​‌YtY‌^t} is‌​‌ o(T)​​-close to maxf​​​‌t‌=1T1‌​‌{Yt≠​​f(Xt​​​‌)}; the‌ difference between those terms‌​‌ is called regret.​​ The more general model​​​‌ with an underlying state‌ of the world S‌​‌t𝒮 that​​ evolves at each step​​​‌ following some Markov Decision‌ Process (MDP, i.e., the‌​‌ transition matrix from S​​t to St​​​‌+1 depend on‌ the actions of the‌​‌ agent) and impacts the​​ loss function is called​​​‌ reinforcement learning (RL). RL‌ is an incredibly powerful‌​‌ learning technique, provided enough​​ data are available since​​​‌ learning is usually quite‌ slow. This is why‌​‌ the recent successes involve​​ settings with heavy simulations​​​‌ (like games) or well-understood‌ physical systems (like robots).‌​‌

These techniques will be​​ central to our approach​​​‌ as we aim to‌ model problems where ground‌​‌ truth data is not​​ available upfront and problems​​​‌ involving sequential decision making.‌ There have been some‌​‌ successful first results in​​ that direction. For instance,​​​‌ there are applications (e.g.,‌ cognitive radio) where several‌​‌ agents (users) aim at​​ finding a matching with​​​‌ resources (the different bandwidth).‌ They can do that‌​‌ by “probing” the resources,​​ estimating their preferences and​​​‌ trying to find some‌ stable matchings 52,‌​‌ 100.

Online algorithms​​ 50 and theoretical computer​​​‌ science

Online algorithms are‌ closely related to online‌​‌ learning with a major​​ twist. In online learning,​​​‌ the agent has “0-look‌ ahead”; for instance, in‌​‌ the online binary classification​​​‌ example, the loss at​ stage t was 1​‌{Yt≠​​Y^t}​​​‌ but Yt was​ not known in advance.​‌ The comparison class, on​​ the other hand, was​​​‌ the empirical performance of​ a given set of​‌ classifiers. In online algorithms,​​ the agents have “1-look​​​‌ ahead”; in the classification​ example, this means that​‌ Yt is known​​ before choosing Y^​​​‌t. But the​ overall objective is obviously​‌ no longer the minimisation​​ of the empirical error,​​​‌ but the minimisation of​ this error plus the​‌ total number of changes​​ (say). The comparison class​​​‌ is then larger, namely​ a subset of admissible​‌ (or the whole set)​​ sequences of prediction {​​​‌±1}T​. The typical and​‌ relevant example of online​​ problem relevant for Criteo​​​‌ that will be investigated​ is the matching problem:​‌ agents and resources arrive​​ sequentially and must be,​​​‌ if possible, paired together​ as fast as possible​‌ (and as successfully as​​ possible). Variants of these​​​‌ problems include the optimal​ stopping time question (when/how​‌ make a final decision)​​ such as prophet inequalities​​​‌ and related questions 67​,

Optimal transport 125​‌

Optimal transport is a​​ quite old problem introduced​​​‌ by Monge where an​ agent aims at moving​‌ a pile of sand​​ to fill a hole​​​‌ at the smallest possible​ price. Formally speaking, given​‌ two probability measures μ​​ and ν on some​​​‌ space 𝒳, the​ optimal transport problem consist​‌ in finding (if it​​ exists, otherwise the problem​​​‌ can be relaxed) a​ transport map T:​‌𝒳𝒳 that​​ minimizes 𝒳c​​​‌(x,T​(x))​‌dμ(x​​) for some cost​​​‌ function c:𝒳​2R,​‌ under the constraint that​​ Tμ=​​​‌ν, where T​μ is the​‌ push-forward measure of μ​​ by T. Interestingly,​​​‌ when μ and ν​ are empirical measures, i.e.,​‌ μ=1N​​n=1​​​‌δxn and​ ν=1N​‌n=1​​δyn,​​​‌ a transport map is​ nothing more than a​‌ matching between {x​​n} and {​​​‌yn} that​ minimizes the cost ∑​‌nc(x​​n,T(​​​‌xn))​.

Recently, optimal transport​‌ gained a lot of​​ interest in the ML​​​‌ community 114 thanks to​ its application to images​‌ and to new techniques​​ to compute approximate matchings​​​‌ in a tractable way​ 117. Even more​‌ unexpected applications of optimal​​ transport have been discovered:​​​‌ to protect privacy 53​, fairness 46,​‌ etc. Those connections are​​ promising, but only primitive​​​‌ for the moment. For​ instance, consider the problem​‌ of matching students to​​ schools. The unfairness level​​​‌ of a school can​ be measured as the​‌ Wasserstein distance between the​​ distribution of the students​​​‌ within that school compared​ to the overall distribution​‌ of students. Then the​​ matching algorithms could have​​ a constraint of minimizing​​​‌ the sum of (or‌ its maximum) unfairness levels;‌​‌ alternatively, we could aim​​ at designing mechanisms giving​​​‌ incentives to schools to‌ be fair in their‌​‌ allocation (or at least​​ in their list preferences),​​​‌ typically by paying a‌ higher fee if the‌​‌ unfairness level is high.​​

2.3 General objectives

The​​​‌ overarching objective of FAIRPLAY‌ of to create algorithms‌​‌ to learn for and​​ with users—and techniques to​​​‌ analyze them—, through‌ the study of multi-agent‌​‌ learning systems where the​​ agents can be cooperatively​​​‌ or competitively learning agents,‌ or agents providing or‌​‌ generating data, while​​ guaranteeing that fairness and​​​‌ privacy constraints are satisfied‌ for the involved users.‌​‌ We detail this global​​ objective into a number​​​‌ of more specific ones.‌

  • Objective 1: Developing‌​‌ fair and private mechanisms​​

     

    Our first objective is​​​‌ to incorporate ethical aspects‌ of fairness and privacy‌​‌ in mechanisms used in​​ typical problems occurring in​​​‌ Internet applications, in particular‌ auctions, matching, and recommendation.‌​‌ We will focus on​​ social welfare and consider​​​‌ realistic cases with multiple‌ agents and sequential learning‌​‌ that occur in practice​​ due to sequential decision​​​‌ making. Our objective is‌ both to construct models‌​‌ to analyze the problem,​​ to devise algorithms that​​​‌ respect the constraints at‌ stake, and to evaluate‌​‌ the different trade-offs in​​ standard notions of utility​​​‌ introduced by ethical constraints.‌

  • Objective 2: Developing‌​‌ multi-agent statistics and learning​​

     

    Data is now acquired,​​​‌ treated and/or generated by‌ a whole network of‌​‌ agents interacting with the​​ environment. There are also​​​‌ often multiple agents learning‌ either collaboratively or competitively.‌​‌ Our second objective is​​ to build a new​​​‌ set of tools to‌ perform statistics and learning‌​‌ tasks in such environments.​​ To this end, we​​​‌ aim at modeling these‌ situations as multi-agent systems‌​‌ and at studying the​​ dynamics and equilibrium of​​​‌ these complex game-theoretic situations‌ between multiple learning algorithms‌​‌ and data providers.

  • Objective​​ 3: Improving the​​​‌ theoretical state of the‌ art

     

    Research must rely‌​‌ on theoretical, proven guarantees.​​ We develop new results​​​‌ for the techniques introduced‌ before, such as prophet‌​‌ inequalities, (online) matchings, bandits​​ and RL, etc.

  • Objective​​​‌ 4: Proposing practical‌ solutions and enhancing transfer‌​‌ from research to industry​​

     

    Our last scientific objective​​​‌ is to apply and‌ implement theoretical works and‌​‌ results to practical cases.​​ This will be a​​​‌ crucial component of the‌ project as we focus‌​‌ on transfer within Criteo.​​

  • Objective 5: Scientific​​​‌ Publications

     

    We aim at‌ publishing our results in‌​‌ top-tier machine learning conferences​​ (NeurIPS, ICML, COLT, ICLR,​​​‌ etc.) and in top-tier‌ game theory journals (Games‌​‌ and Economic Behavior, Mathematics​​ of OR, etc.). We​​​‌ will also target conferences‌ at the junction of‌​‌ those fields (EC, WINE,​​ WebConf, etc.) as well​​​‌ as conferences specifically on‌ security and privacy (IEEE‌​‌ S&P, NDSS, CSS, PETS,​​ etc.) and on fairness​​​‌ (FAccT, AIES).

All the‌ five objectives are interlaced.‌​‌ For instance, fairness and​​ privacy constraints are important​​​‌ in Objective 2 whereas‌ the multi-agent aspect is‌​‌ also important in Objective​​​‌ 1. Objectives 4 and​ 5 are transversal and​‌ present in all the​​ first three objectives.

3​​​‌ Research program

To reach​ the objectives laid out​‌ above, we organize the​​ research in three themes.​​​‌ The first one focuses​ on developing fair mechanisms.​‌ The second one considers​​ private mechanisms, and in​​​‌ particular considers the challenge​ of reconciling fairness and​‌ privacy—which are often conflicting​​ notions. The last theme,​​​‌ somewhat transverse to the​ first two, consists in​‌ leveraging/incorporating structure in all​​ those problems in order​​​‌ to speed up learning.​ Of course, all themes​‌ share common points on​​ both the problems/applications considered​​​‌ and the methods and​ tools used to tackle​‌ them; hence there are​​ cross-fertilization between the different​​​‌ themes.

3.1 Theme 1:​ Developing fair mechanisms for​‌ auctions and matching problems​​

3.1.1 Fairness in auction-based​​​‌ systems

Online ads platforms​ are nowadays used to​‌ advertise not just products,​​ but also opportunities such​​​‌ as jobs, houses, or​ financial services. This makes​‌ it crucial for such​​ platforms to respect fairness​​​‌ criteria (be it only​ for legal reasons), as​‌ an unfair ad system​​ would deprive a part​​​‌ of the population of​ some potentially interesting opportunities.​‌ Despite this pressing need,​​ there is currently no​​​‌ technical solution in place​ to provably prevent discriminations.​‌ One of the main​​ challenge is that ad​​​‌ impression decisions are the​ outcome of an auction​‌ mechanism that involves bidding​​ decisions of multiple self-interested​​​‌ agents controlling only a​ small part of the​‌ process, while group fairness​​ notions are defined on​​​‌ the outcome of a​ large number of impressions.​‌ We propose to investigate​​ two mechanisms to guarantee​​​‌ fairness in such a​ complex auction-based system (note​‌ that we focus on​​ online ad auctions but​​​‌ the work has broader​ applicability).

  • Advertiser-centric (or bidder-centric)​‌ fairness

    We first focus​​ on advertiser-centric fairness, i.e.,​​​‌ the advertiser of a​ third-party needs to make​‌ sure that the reached​​ audience is fair independently​​​‌ of the ad auction​ platform. A key difficulty​‌ is that the advertiser​​ does not control the​​​‌ final decision for each​ ad impression, which depends​‌ on the bids of​​ other advertisers competing in​​​‌ the same auction and​ on the platform's mechanism.​‌ Hence, it is necessary​​ that the advertiser keeps​​​‌ track of the auctions​ won for each of​‌ the groups and dynamically​​ adjusts its bids in​​​‌ order to maintain the​ required balance.

    A first​‌ difficulty is to model​​ the behavior of other​​​‌ advertisers. We can first​ use a mean-field games​‌ approach similar to 86​​ that approximates the other​​​‌ bidders by an (unknown)​ distribution and checks equilibrium​‌ consistency; this makes sense​​ if there are many​​​‌ bidders. We can also​ leverage refined mean-field approximations​‌ 80 to provide better​​ approximations for smaller numbers​​​‌ of advertisers. Then a​ second difficulty is to​‌ find an optimal bidding​​ policy that enforces the​​​‌ fairness constraint. We can​ investigate two approaches. One​‌ is based on an​​ MDP (Markov Decision Process)​​​‌ that encodes the current​ fairness level and imposes​‌ a hard constraint. The​​ second is based on​​ modeling the problem as​​​‌ a contextual bandit problem.‌ We note that in‌​‌ addition to fairness constraints,​​ privacy constraints may complicate​​​‌ the optimal solution finding.‌

  • Platform-centric (or auction-centric) fairness‌​‌

    We also consider the​​ problem from the platform's​​​‌ perspective, i.e., we assume‌ that it is the‌​‌ platform's responsibility to enforce​​ the fairness constraint. We​​​‌ also focus here on‌ demographic parity. To make‌​‌ the solution practical, we​​ do not consider modification​​​‌ of the auction mechanism,‌ instead we consider a‌​‌ given mechanism and let​​ the platform adapt dynamically​​​‌ the bids of each‌ advertiser to achieve the‌​‌ fairness guarantee. This approach​​ would be similar to​​​‌ the pacing multipliers used‌ by some platforms 66‌​‌, 65, but​​ using different multipliers for​​​‌ the different groups (i.e.,‌ different values of the‌​‌ sensitive attribute).

    Following recent​​ theoretical work on auction​​​‌ fairness 59, 85‌, 63 (which assumes‌​‌ that the targeted population​​ of all ads is​​​‌ known in advance along‌ with all their characteristics),‌​‌ we can formulate fairness​​ as a constraint in​​​‌ an optimization problem for‌ each advertiser. We study‌​‌ fairness in this static​​ auction problem in which​​​‌ the auction mechanism is‌ fixed (e.g., to second‌​‌ price). We then move​​ to the online setting​​​‌ in which users (but‌ also advertisers) are dynamic‌​‌ and in which decisions​​ must be taken online,​​​‌ which we approach through‌ dynamic adjustment of pacing‌​‌ multipliers.

3.1.2 Fairness in​​ matching and selection problems​​​‌

In this second part,‌ we study fairness in‌​‌ selection and matching problems​​ such as hiring or​​​‌ college admission. The selection‌ problem corresponds to any‌​‌ situation in which one​​ needs to select (i.e.,​​​‌ assign a binary label‌ to) data examples or‌​‌ individuals but with a​​ constraint on the number​​​‌ of maximum number of‌ positive labels. There are‌​‌ many applications of selection​​ problems such as police​​​‌ checks, loan approvals, or‌ medical screening. The matching‌​‌ problem can be seen​​ as the more general​​​‌ variant with multiple selectors.‌ Again, a particular focus‌​‌ is put here on​​ cases involving repeated selection/matching​​​‌ problems and multiple decision‌ makers.

  • Fair repeated multistage‌​‌ selection

    In our work​​ 74, we identified​​​‌ that a key source‌ of discrimination in (static)‌​‌ selection problems is differential​​ variance, i.e., the​​​‌ fact that one has‌ quality estimates that have‌​‌ different variances for different​​ groups. In practice, however,​​​‌ the selection problem is‌ often ran repeatedly (e.g.,‌​‌ at each hiring campaign)​​ and with partial (and​​​‌ increasing) information to exploit‌ for making decisions.

    Here,‌​‌ we consider the repeated​​ multistage selection problem, where​​​‌ at each round a‌ multistage selection problem is‌​‌ solved. A key aspect​​ is that, at the​​​‌ end of a round,‌ one learns extra information‌​‌ about the candidates that​​ were selected—hence one can​​​‌ refine (i.e., decrease the‌ variance of) the quality‌​‌ estimate for the groups​​ in which more candidates​​​‌ were selected. We will‌ first rethink fairness constraints‌​‌ in this type of​​ repeated decision making problems.​​​‌ Then we will both‌ study the discrimination that‌​‌ come out of natural​​​‌ (e.g., greedy) procedure as​ well as design (near)​‌ optimal ones for the​​ constraints at stake. We​​​‌ also investigate how the​ constraints affect the selection​‌ utility.

  • Multiple decision-makers
    Next,​​ we investigate cases with​​​‌ multiple decision-makers. We propose​ two cases in particular.​‌ The first one is​​ the simple two-stage selection​​​‌ problem but where the​ decision-maker doing the first-stage​‌ selection is different from​​ the decision-maker doing the​​​‌ second-stage selection. This is​ a typical case for​‌ instance for recruiting agencies​​ that propose sublists of​​​‌ candidates to different firms​ that wish to hire.​‌ The second case is​​ when multiple-decision makers are​​​‌ trying to make a​ selection simultaneously—a typical example​‌ of this being the​​ college admission problem (or​​​‌ faculty recruitment). We intend​ to model it as​‌ a game between the​​ different colleges and to​​​‌ study both static solutions​ as well as dynamic​‌ solutions with sequential learning,​​ again modeling it as​​​‌ a bandit problem and​ looking for regret-minimizing algorithms​‌ under fairness constraints. A​​ number of important questions​​​‌ arise here: if each​ college makes its selection​‌ independently and strategically (but​​ based on quality estimates​​​‌ with variances that differ​ amongst groups), how does​‌ it affect the “global​​ fairness” metrics (meaning in​​​‌ aggregate across the different​ colleges) and the “local​‌ fairness” metrics (meaning for​​ an individual college)? What​​​‌ changes if there is​ a central admission system​‌ (such as Parcoursup)? And​​ in this later case,​​​‌ how to handle fairness​ on the side of​‌ colleges (i.e., treat each​​ college fairly in some​​​‌ sense)?
  • Fair matching with​ incentives in two-sided platforms​‌

    We will study specifically​​ the case of a​​​‌ platform matching demand on​ one side to offer​‌ on the other side,​​ with fairness constraints on​​​‌ each side. This is​ the case for instance​‌ in online job markets​​ (or crowdsourcing). This is​​​‌ similar to the previous​ case but, in addition,​‌ here there is an​​ extra incentives problem: companies​​​‌ need to give the​ right incentives to job​‌ applicants to accept the​​ jobs; while the platform​​​‌ doing the match needs​ to ensure fairness on​‌ both sides (job applicants​​ and companies). This gives​​​‌ rise to a complicated​ interplay between learning and​‌ incentives that we will​​ tackle again in the​​​‌ repeated setting.

    We finally​ mention that, in many​‌ of these matching problems,​​ there is an important​​​‌ time component: each agent​ needs to be matched​‌ “as soon as possible”,​​ yielding a trade-off between​​​‌ the delay to be​ matched and the quality​‌ of the match. There​​ is also a problem​​​‌ of participation incentives; that​ is how the matching​‌ algorithm used affect the​​ behavior of the participants​​​‌ in the matching “market”​ (participation or not, information​‌ revelation, etc.). In the​​ long-term, we will incorporate​​​‌ these aspects in the​ above models.

Throughout the​‌ work in this theme,​​ we will also consider​​​‌ a question transverse and​ present in all the​‌ models above: how can​​ we handle multidimensional fairness​​​‌, that is, where​ there are multiple sensitive​‌ attributes and consequently an​​ exponential number of sub-groups​​ defined by all intersections;​​​‌ this combinatorial is challenging‌ and, for the moment,‌​‌ still exploratory.

3.2 Theme​​ 2: Reconciling, and enforcing​​​‌ privacy with fairness

In‌ the previous theme, we‌​‌ implicitly assumed that we​​ know the users' group,​​​‌ i.e., their sensitive attributes‌ such as gender, age,‌​‌ or ethnicity. In practice,​​ one of the key​​​‌ question when implementing fairness‌ mechanisms is how to‌​‌ measure/control fairness metrics without​​ having access to these​​​‌ protected attributes. This question‌ relates to the link‌​‌ between privacy and fairness​​ and the trade-off between​​​‌ them (as fairness requires‌ data and privacy tends‌​‌ to protect it) 118​​, 68.

A​​​‌ first option to solve‌ this problem would be‌​‌ (when it is possible)​​ to build proxies 81​​​‌, 121 for protected‌ attributes using available information‌​‌ (e.g., websites visited or​​ products bought) and to​​​‌ measure or control for‌ fairness using those in‌​‌ place of the protected​​ attributes. As the accuracy​​​‌ of these proxies cannot‌ be assessed, however, they‌​‌ cannot be used for​​ any type of “public​​​‌ certification”—that is, for a‌ company to show reasonable‌​‌ fairness guarantees to clients​​ (e.g., as a commercial​​​‌ argument), or (even less)‌ to regulators. Moreover, in‌​‌ many cases, the entity​​ responsible for fairness should​​​‌ not be accessing sensitive‌ information, even through proxies,‌​‌ for privacy reasons.

In​​ FAIRPLAY, we investigate a​​​‌ different means of certifying‌ fairness of decisions without‌​‌ having access to sensitive​​ attributes, by partnering with​​​‌ a trusted third-party that‌ collects protected attributes (that‌​‌ could for instance be​​ a regulator or a​​​‌ public entity, such as‌ Nielsen, say). We distinguish‌​‌ two cases:

  1. If the​​ third-party and the company​​​‌ share a common identifier‌ of users, then computing‌​‌ the fairness metric without​​ leaking information to each​​​‌ other will boil down‌ to a problem of‌​‌ secure multi-party computation (SMC).​​ In such a case,​​​‌ there could be a‌ need to be able‌​‌ to learn, which opens​​ the topic of learning​​​‌ and privacy under SMC.‌ This scenario, however, is‌​‌ likely not the most​​ realistic one as having​​​‌ a common identifier requires‌ a deep technical integration.‌​‌
  2. If the third-party and​​ the company do not​​​‌ share a common identifier‌ of users, but there‌​‌ are common features that​​ they both observe 89​​​‌, then it is‌ possible only to partially‌​‌ identify the joint distribution.​​ With additional structural assumptions​​​‌ on the distribution, however,‌ it could be identified‌​‌ accurately enough to estimate​​ biases and fairness metrics.​​​‌ This becomes a distribution‌ identification problem and brings‌​‌ a number of questions​​ such as: how to​​​‌ do the distribution identification?‌ how to optimally query‌​‌ data from the third​​ party to train fair​​​‌ algorithms with high utility?‌ etc. An important point‌​‌ to keep in mind​​ in such a study​​​‌ is that it is‌ likely that the third‌​‌ party user-base is different​​ from that of the​​​‌ company. It will therefore‌ be key to handle‌​‌ the covariate shift from​​ one distribution to the​​​‌ other while estimating biases.‌

This distribution identification problem‌​‌ will be important in​​​‌ the context of privacy,​ even independently of fairness​‌ issues. Indeed, in the​​ near future, most learning​​​‌ will happen in a​ privacy-preserving setting (for instance,​‌ because of the Chrome​​ privacy sandbox). This will​​​‌ require new learning schemes​ (different from e.g., Empirical​‌ Risk Minimization) as samples​​ from the usual joint​​​‌ distribution (X,​Y) of samples/labels​‌ will no longer be​​ observed. Only aggregated data—e.g.,​​​‌ (empirical) marginals of the​ form E[Y​‌|X2=​​4,X7​​​‌=``lemonde.fr'']—will​ be observed, with a​‌ limited budget of requests.​​ This also brings questions​​​‌ such as how to​ mix it with ERM​‌ on some parts of​​ the traffic, what is​​​‌ the (certainly adaptive or​ active) optimal strategy to​‌ query the marginals, etc.​​ This problem will be​​​‌ further complicated by the​ fact that privacy (for​‌ instance through the variety​​ of consents) will be​​​‌ heterogeneous: all features are​ not available all the​‌ time. This is therefore​​ strongly related to learning​​​‌ with missing features and​ imputation 88.

In​‌ relation to the above​​ problems, a key question​​​‌ is to determine what​ is the most appropriate​‌ definition of fairness to​​ consider. Recall that it​​​‌ is well-known that usual​ fairness metrics are not​‌ compatible 93. Moreover,​​ in online advertising, fairness​​​‌ can be measured at​ multiple different levels: at​‌ the level of bids,​​ at the level of​​​‌ audience reached, at the​ level of clicking users,​‌ etc. Fairness at the​​ level of bids does​​​‌ not imply fairness of​ the audience reached (see​‌ Theme 1); yet external​​ auditors would measure which​​​‌ ad is displayed—as was​ done for some ad​‌ platforms 122—hence in​​ terms of public image,​​​‌ that would be the​ appropriate level to define​‌ fairness. Intimately, the above​​ problem relates to the​​​‌ question of measuring what​ is the relevant audience​‌ of an ad, which​​ would define the label​​​‌ if one were to​ use the EO fairness​‌ notion. This label is​​ typically not available. We​​​‌ can explore three ways​ to overcome this issue.​‌ The first is to​​ find a sequential way​​​‌ to learn the label​ through users clicking on​‌ ads. The second and​​ third options are to​​​‌ focus in a first​ step on individual fairness,​‌ or on counterfactual fairness​​ 96, which has​​​‌ many possible different level​ of assumptions and was​‌ popularized in 2020 97​​. The notion of​​​‌ counterfactual is key in​ causality 116. A​‌ model is said counterfactually​​ fair if its prediction​​​‌ does not change (too​ much) by intervening on​‌ the sensitive attribute. Several​​ works already propose ways​​​‌ of designing models that​ are counterfactually fair 92​‌, 129, 128​​. This seems to​​​‌ be quite an interesting,​ but challenging direction to​‌ follow.

Finally, an alternative​​ direction would be to​​​‌ purse modeling the trade-off​ between privacy and fairness.​‌ For instance, in some​​ game theoretic models, users​​​‌ can choose the quantity​ of data that they​‌ reveal 79, 53​​, so that the​​ objective functions integrate different​​​‌ levels of fairness/privacy. Then‌ those models model should‌​‌ be studied both in​​ terms of equilibrium and​​​‌ in the online setup,‌ with the objective of‌​‌ identifying how the strategic​​ privacy considerations affect the​​​‌ fairness-utility tradeoff.

3.3 Theme‌ 3: Exploiting structure in‌​‌ online algorithms and learning​​ problems

Our last research​​​‌ direction is somewhat transverse,‌ with possible application to‌​‌ improving algorithms in the​​ first three themes. We​​​‌ explore how the underlying‌ structure can be exploited,‌​‌ in the online and​​ learning problems considered before,​​​‌ to improve performance. Note‌ that, in all these‌​‌ problems, we will incorporate​​ the fairness and privacy​​​‌ aspects even if they‌ are somewhat transverse to‌​‌ the structure considered.1​​ The following sections are​​​‌ illustrating examples on how‌ hidden structure can be‌​‌ leveraged in specific examples.​​

3.3.1 Leveraging structure in​​​‌ online matching

Finding large‌ matchings in graphs is‌​‌ a longstanding problem with​​ a rich history and​​​‌ many practical and theoretical‌ applications 107, 84‌​‌, 44, 43​​. Recall that given​​​‌ a graph G=‌(𝒱,ℰ‌​‌)—where 𝒱 is​​ a set of vertices​​​‌ and is a‌ set of edges—, a‌​‌ matching ℰ​​ is a subset of​​​‌ edges such that each‌ vertex belongs to at‌​‌ most one edge e​​. In​​​‌ that context, a perfect‌ matching is a‌​‌ matching where each vertex​​ v𝒱 is​​​‌ associated to an edge‌ e,‌​‌ and a maximum matching​​ is a matching of​​​‌ maximum size (one can‌ also consider weights on‌​‌ edges). Here, we study​​ an online setting, which​​​‌ is more adequate in‌ applications such as Internet‌​‌ advertising where ad impressions​​ must be assigned to​​​‌ available ad slots 107‌, 56. Consider‌​‌ a bipartite graph, where​​ 𝒱=U∪​​​‌V is the union‌ of two disjoints sets.‌​‌ Nodes uU​​ are known beforehand, whereas​​​‌ nodes vV‌ are discovered one at‌​‌ a time, along with​​ the edges they belong​​​‌ to, and must be‌ either immediately matched to‌​‌ an available (i.e., unmatched​​ yet) vertex u∈​​​‌U or discarded. Online‌ bipartite matching is relevant‌​‌ in two-sided markets besides​​ ad allocations such as​​​‌ assigning tasks to workers‌ 84.

A natural‌​‌ measure for the quality​​ of an online matching​​​‌ algorithm is the “competitive‌ ratio” (CR): the ratio‌​‌ between the size of​​ the created matching to​​​‌ the size of the‌ optimal one 107.‌​‌ The seminal work 91​​ introduced an optimal algorithm​​​‌ for the adversarial case‌ 47, that guarantees‌​‌ a CR of 1​​-1e;​​​‌ but focusing on a‌ pessimistic worst-case. In practice,‌​‌ some relevant knowledge (either​​ given a priori or​​​‌ learned during the process)‌ on the underlying structure‌​‌ of the problem can​​ be leveraged. The focus​​​‌ then shifted to models‌ taking into account some‌​‌ type of stochasticity in​​ the arrival model, mostly​​​‌ for the i.i.d. model‌ where arriving vertices v‌​‌V are drawn​​​‌ from a fixed distribution​ 𝒟87, 45​‌, 75, 90​​, 102, 103​​​‌. The classical approach​ consists in optimizing the​‌ CR over the distribution​​ 𝒟. Even in​​​‌ this seemingly optimistic framework,​ however, it is now​‌ known that there is​​ no hope for a​​​‌ CR of more than​ 0.823 103. Moreover,​‌ this generally leads to​​ very large linear programs​​​‌ (LP).

A more recent​ approach restricts the distribution​‌ 𝒟 over which the​​ problem is optimized to​​​‌ classes of graphs with​ an underlying stochastic structure.​‌ The benefit of this​​ approach is two-fold: it​​​‌ gives hope for higher​ competitive ratios, and for​‌ simpler algorithms. Experiments also​​ proved that complex algorithms​​​‌ optimized on 𝒟 fared​ no better than simple​‌ greedy heuristics on “real-life”​​ graphs 51. A​​​‌ few results along these​ lines show that is​‌ a promising path. For​​ instance, 56 studied the​​​‌ problem on graphs where​ each vertex has degree​‌ at least d and​​ found a competitive ratio​​​‌ of 1-(​1-1/​‌d)d.​​ On d-regular graphs, 64​​​‌ designed a 1-​O(logd​‌/d) competitive​​ algorithm. 105 showed that​​​‌ greedy algorithms were highly​ efficient on Erdös-Renyi random​‌ graphs, with a competitive​​ ratio of 0.837 in​​​‌ the worst case. 42​ showed that in a​‌ specific market with two​​ types of matching agents,​​​‌ the behavior of the​ matching algorithm varies with​‌ the homogeneity of the​​ market. Our goal here​​​‌ is to go beyond​ the independence assumption underlying​‌ all these works.

  • Introducing​​ correlation and inhomogeneity

    We​​​‌ will start by deriving​ and studying optimal online​‌ matching strategies on widely​​ studied classes of graphs​​​‌ that present simple inhomongeneity​ or correlation structures (which​‌ are often present in​​ applications). The stochastic block​​​‌ model 39 is often​ used to generate graphs​‌ with an underlying community​​ structure. It presents a​​​‌ simple correlation structure: two​ vertices in the same​‌ community are more likely​​ to have a common​​​‌ neighbors than two vertices​ in different communities. Another​‌ focus point will be​​ a generalized version of​​​‌ the Erdös-Renyi model, where​ the inplace vertices u​‌𝒰 are divided​​ into sets si​​​‌, where u∈​si generates an​‌ edge with probability p​​i=ci​​​‌n. These two​ settings should give us​‌ a better understanding of​​ how heterogeneity and correlation​​​‌ affect the matching performance.​

    Deriving the competitive ratio​‌ implies to study the​​ asymptotic size of maximum​​​‌ matchings in random graphs.​ Two methods are usually​‌ used. The first and​​ constructive one is the​​​‌ study of the Karp-Sipser​ algorithm on the graph​‌ 48. The second​​ one involves the rich​​​‌ theory of graph weak​ local convergence 49.​‌ A straightforward application of​​ the methods, however, requires​​​‌ the graph to have​ independence properties; adapting them​‌ to graphs with a​​ correlation structure will require​​​‌ new ideas.

  • Configuration models​ and random geometric graphs​‌
    A configuration model is​​ described as follows (in​​ the bi-partite case). Each​​​‌ vertex u𝒰‌ has a number of‌​‌ half-edges drawn for the​​ same distribution π𝒰​​​‌ and each vertex v‌𝒱 has a‌​‌ number of half-edges drawn​​ from πV (with​​​‌ the assumption that the‌ expected total numbers of‌​‌ half edges from 𝒰​​ and 𝒱 are the​​​‌ same). Then a vertex‌ v𝒱 that‌​‌ arrives in the sequential​​ fashion has its half-edges​​​‌ “completed” by a (still‌ free) half-edge of 𝒰‌​‌. This is a​​ standard way of creating​​​‌ random graphs with (almost)‌ fixed distribution of degrees.‌​‌ Here the question would​​ simply be the competitive​​​‌ ratio of some greedy‌ algorithm, whether the distributions‌​‌ π𝒰 and π​​𝒱 are known beforehand​​​‌ or learned on the‌ fly. An interesting variant‌​‌ of this problem would​​ be to assume the​​​‌ existence of a (hidden‌ or not) geometric graph.‌​‌ Each u𝒰​​ is drawn i.i.d in​​​‌ d (say a‌ Gaussian centered at 0)‌​‌ and similarly for v​​𝒱. Then​​​‌ there is an edge‌ between u and v‌​‌ with a probability depending​​ on the distance between​​​‌ them. Here again, interesting‌ variants can be explored‌​‌ depending on whether the​​ distribution is known or​​​‌ not, and whether the‌ locations of u and/or‌​‌ v are observed or​​ not.
  • Learning while matching​​​‌
    In practical applications, the‌ full stochastic structure of‌​‌ the graphs may not​​ be known beforehand. This​​​‌ begs the question: what‌ will happen to the‌​‌ performance of the algorithms​​ if the graph parameters​​​‌ are learned while matching?‌ In the generalized Erdös-Renyi‌​‌ graph, this will correspond​​ to learning the probability​​​‌ of generating edges. For‌ the stochastic block model,‌​‌ the matching algorithm will​​ have to perform online​​​‌ community detection.

3.3.2 Exploiting‌ side-information in sequential learning‌​‌

We end with an​​ open direction that may​​​‌ be relevant to many‌ of the problems considered‌​‌ above: how to use​​ side-information to speed-up learning.​​​‌ In many sequential learning‌ problems where one receives‌​‌ feedback for each action​​ taken, it is actually​​​‌ possible to deduce, for‌ free, extra information from‌​‌ the known structure of​​ the problem. However, how​​​‌ to incorporate that information‌ in the learning process‌​‌ is often unclear. We​​ describe it through two​​​‌ examples.

  • One-sided feedback in‌ auctions
    In online ad‌​‌ auctions, the advertisers' strategy​​ is to bid in​​​‌ a compact set of‌ possible bids. After placing‌​‌ a bid, the advertiser​​ learns whether they won​​​‌ the auction or not;‌ but even if they‌​‌ do not observe the​​ bids of other advertisers,​​​‌ they can deduce for‌ free some extra information:‌​‌ if they win they​​ learn that they would​​​‌ have won with any‌ higher bid and if‌​‌ they loose they learn​​ that they would have​​​‌ lost with any lower‌ bid 126, 60‌​‌. We will investigate​​ how to incorporate this​​​‌ extra information in RL‌ procedures devised in Theme‌​‌ 1. One option is​​ by leveraging the Kaplan-Meier​​​‌ estimator.
  • Side-information in dynamic‌ resource allocation problems and‌​‌ matching
    Generalizing the idea​​​‌ above, one can observe​ side-information in many other​‌ problems 41. Typically,​​ in resource allocation problems​​​‌ (e.g., how to allocate​ a budget of ad​‌ impressions), one can leverage​​ a monotony property: one​​​‌ would not have gained​ more by allocating less.​‌ Similarly, in matching with​​ unknown weights, it is​​​‌ often possible upon doing​ a particular match to​‌ learn the weight of​​ other potential pairs.

4​​​‌ Application domains

4.1 Typical​ problems and use-cases

In​‌ FAIRPLAY, we focus mainly​​ on problems involving learning​​​‌ that relate to Internet​ applications. We will tackle​‌ generic problems (in particular​​ auctions and matching) with​​​‌ different applications in mind,​ in particular some applications​‌ in the context of​​ Criteo's business but also​​​‌ others. A crucial property​ of those applications is​‌ the aforementioned ethical concerns,​​ namely fairness and/or privacy.​​​‌ The team was shaped​ and motivated by several​‌ such use-cases, from more​​ practical (with possible short​​​‌ or middle term applications​ in particular in Criteo​‌ products) to more theoretical​​ and exploratory ones. We​​​‌ describe first here the​ main types of generic​‌ problems and use-cases considered​​ in this context.

Auctions​​​‌ 95

There are many​ different types of auctions​‌ that an agent can​​ use to sell one​​​‌ or several items in​ her possession to n​‌ potential buyers. This is​​ the typical way in​​​‌ which spots to place​ ads are sold to​‌ potential advertisers. In case​​ of a single item,​​​‌ the seller ask buyers​ to bid bi​‌[0,​​1] on the​​​‌ item and the winner​ of the item is​‌ designating via an “allocation​​ rule” that maps bids​​​‌ b[0​,1]n​‌ to a winner in​​ {0,...​​​‌,n} (0​ refers to the no​‌ winner case). Then the​​ payment rule p:​​​‌[0,1​]n[​‌0,1]​​n indicates the amount​​​‌ of money that each​ bidder must pay to​‌ the seller. Auctions are​​ specific cases of a​​​‌ broader family of “mechanisms”.​ Knowing the allocation and​‌ payment rules, bidders have​​ incentives to bid strategically.​​​‌ Different auctions (or rules)​ end up with different​‌ revenue to the seller,​​ who can choose the​​​‌ optimal rules. This is​ rather standard in economics,​‌ but these interactions become​​ way more intricate when​​​‌ repeated over time (as​ in the online ad​‌ market 111), when​​ several items are sold​​​‌ at the same time​ (for instance in bundles),​‌ when the buyers have​​ partial information about the​​​‌ actual value of the​ item 126 and/or reciprocally​‌ when the seller does​​ not know the value​​​‌ distributions of the buyer.​ In that case, she​‌ might be tempted to​​ try to learn them​​​‌ from the previous bids​ in order to design​‌ the optimal mechanism. Knowing​​ this, the bidders have​​​‌ incentives to long term​ strategic behaviors, ending up​‌ in a quite complicated​​ game between learning algorithms​​​‌ 112. This setting​ of interacting algorithms is​‌ actually of interest by​​ itself, irrespectively of ad​​ auctions. It is noteworthy​​​‌ also that traditional auction‌ mechanisms do not guarantee‌​‌ any fairness notion and​​ that the literature on​​​‌ fixing that (for applications‌ where it matters) is‌​‌ only nascent 59,​​ 110, 63,​​​‌ 85.

Matching 94‌, 109

A matching‌​‌ is nothing more than​​ a bi-partite graph between​​​‌ some agents (patients, doctors,‌ students) and some resources‌​‌ (respectively, organs, hospital, schools).​​ The objective is to​​​‌ figure out what is‌ the “optimal” matching for‌​‌ a given criterion. Interestingly,​​ there are two different—and​​​‌ mostly unrelated yet—concepts of‌ “good matching”. The first‌​‌ one is called “stable”​​ in the sense that​​​‌ each agent expresses preferences‌ over resources (and vice-versa)‌​‌ and be such that​​ no couple (agent-resource) that​​​‌ are un-paired would prefer‌ to be paired together‌​‌ than with their current​​ paired resource/agent. In the​​​‌ other concept of matching,‌ agents and resources are‌​‌ described by some features​​ (say vectors in R​​​‌d, denoted by‌ an for agents‌​‌ and rm for​​ resources) and pairing a​​​‌n to rm‌ incurs a cost of‌​‌ c(an​​,rm)​​​‌, for some a‌ given function c:‌​‌(Rd)​​2[0​​​‌,1].‌ The objective is then‌​‌ to minimize the total​​ cost of the matching​​​‌ nc(‌an,r‌​‌σ(n)​​), where σ​​​‌(n) is‌ the resource allocated to‌​‌ agent n.

Matching​​ is used is many​​​‌ different applications such as‌ university admission (e.g., in‌​‌ Parcoursup). Notice that strategic​​ interactions arise in matching​​​‌ if agents or resources‌ can disclose their preferences/features‌​‌ to each other. Learning​​ is also present as​​​‌ soon as not everything‌ is known, e.g., the‌​‌ preferences or costs. Many​​ applications of matching (again,​​​‌ such as college admission)‌ are typical examples where‌​‌ fairness and privacy are​​ of utmost importance. Finally,​​​‌ matching is also at‌ the basis of several‌​‌ Internet applications and Criteo​​ products, for instance to​​​‌ solve the problem of‌ matching a given ad‌​‌ budget to fixed ad​​ slots.

Ethical notions in​​​‌ those use-cases

In both‌ problems, individual users are‌​‌ involved and there is​​ a clear need to​​​‌ consider fairness and privacy.‌ However, the precise motivation‌​‌ and instantiation of these​​ notions depends on the​​​‌ specific use-case. In fact,‌ it is often part‌​‌ of the research question​​ to decide which are​​​‌ the most relevant fairness‌ and privacy notions, as‌​‌ mentioned in Section 2.1​​. We will throughout​​​‌ the team's life put‌ an important focus on‌​‌ this question, as well​​ as on the question​​​‌ of the impact of‌ the chosen notion on‌​‌ performance.

4.2 Application areas​​

In FAIRPLAY, we consider​​​‌ both applications to Criteo‌ use-cases (online advertisement) and‌​‌ other applications (with other​​ appropriate partners).

4.2.1 Online​​​‌ advertisement

Online advertising offers‌ an application area for‌​‌ all of the research​​ themes of FAIRPLAY; which​​​‌ we investigate primarily with‌ Criteo.

First, online advertising‌​‌ is a typical application​​​‌ of online auctions and​ we consider applications of​‌ the work on auctions​​ to Criteo use-cases, in​​​‌ particular the work on​ advertiser-centric fairness where the​‌ advertiser is Criteo. From​​ a practical point of​​​‌ view, privacy will have​ to be enforced in​‌ such applications. For instance,​​ when information is provided​​​‌ to advertisers to define​ audiences or to visualize​‌ the performance of their​​ campaigns (insights) there is​​​‌ a possibility of leaking​ sensitive information on users.​‌ In particular, excellent proxies​​ on protected attributes should​​​‌ probably not be leaked​ to advertisers, or transformed​‌ before (e.g., with the​​ differential privacy techniques). This​​​‌ is therefore also an​ application of the fairness-vs-privacy​‌ research thread.

Note that,​​ even before considering those​​​‌ questions, the first very​ important theoretical question is​‌ to determine what is​​ the more appropriate definition​​​‌ of fairness (as there​ are, as mentioned above,​‌ many different variations) in​​ those applications. We recall​​​‌ that it is well-known​ that usual fairness metrics​‌ are not compatible 93​​. Moreover, in online​​​‌ advertising, fairness can be​ measured in term of​‌ bidding and recommendation or​​ in term of what​​​‌ ads are actually displayed.​ Being fair on bidding​‌ does not lead to​​ fairness in ads displaying​​​‌ 110, mainly because​ of the other advertising​‌ actors. While fairness in​​ bidding and/or recommendation seem​​​‌ the most important because​ they only rely on​‌ our models, external auditors​​ can easily get data​​​‌ on which ads we​ display.

We will also​‌ investigate applications of fair​​ matching techniques to online​​​‌ advertsing and to Criteo​ matching products—namely retargeting (personalized​‌ ads displayed on a​​ website) and retail media​​​‌ (sponsored products on a​ merchant website). Indeed, one​‌ of Criteo major products,​​ retail media can be​​​‌ cast as an online​ matching problem. On a​‌ given e-commerce website (say,​​ target), several advertisers—currently brands—are​​​‌ running campaigns so that​ their products are “sponsored”​‌ or “boosted”, i.e., they​​ appear higher on the​​​‌ list of results of​ a given query. The​‌ budgets (from a Criteo​​ perspective) must be cleared​​​‌ (daily, monthly or annually).​ This constraint is easy​‌ thanks to the high​​ traffic, but the main​​​‌ issue is that, without​ control/pacing/matching in times, the​‌ budget is depleted after​​ only a few hours​​​‌ on a relatively low​ quality traffic (i.e., users​‌ that generate few conversions​​ hence a small ROI​​​‌ for the advertisers). The​ question is therefore whether​‌ an individual user should​​ be matched or not​​​‌ to boosted/sponsored products at​ a given time so​‌ that the ROI of​​ the advertisers is maximized,​​​‌ the campaign budget is​ depleted and the retailer​‌ does not suffer too​​ much from this active​​​‌ corruption of its organic​ results. Those are three​‌ different and concurrent objectives​​ (for respectively the advertisers,​​​‌ Criteo and the retailers)​ that must be somehow​‌ conciliated. This problem (and​​ more generally this type​​​‌ of problems) offers a​ rich application area to​‌ the FAIRPLAY research program.​​ Indeed, it is crucial​​​‌ to ensure that fairness​ and privacy are respected.​‌ On the other hand,​​ users, clicks, conversion arrival​​ are not “worst case”.​​​‌ They rather follow some‌ complicated—but certainly learnable—process; which‌​‌ allows applying our results​​ on exploiting structure.

4.2.2​​​‌ “Physical matching”

We investigate‌ a number of other‌​‌ applications of matching: assignment​​ of daycare slots to​​​‌ kids, mutation of professors‌ to different academies, assignment‌​‌ of kidneys to patients,​​ assignment of job applicants​​​‌ to jobs. In all‌ these applications, there are‌​‌ crucial constraints of fairness​​ that complicate the matching.​​​‌ We leverage existing partnership‌ with the CNAF, the‌​‌ French Ministry of Education​​ and the Agence de​​​‌ la biomédecine in Paris‌ for the first three‌​‌ applications; for the last​​ we will consolidate a​​​‌ nascent partnership with Pole‌ Emploi and LinkedIn.

5‌​‌ Highlights of the year​​

5.1 Awards

Mariia Vladimirova​​​‌ was named “Rising Star‌ in AI Ethics” issued‌​‌ by Women in AI​​ Ethics organization (March 2025).​​​‌

Vianney Perchet and Patrick‌ Loiseau were named Hi!‌​‌ Paris fellows in 2025.​​ They will lead a​​​‌ synergy fellowship on “Design,‌ Incentivization, Optimization and (Reinforcement-‌​‌ )Learning of Multi-Layered Market”​​ together with two economists:​​​‌ Yuki Tamura and Pablo‌ Winant.

Marc Abeille and‌​‌ Lorenzo Croissant (with co-authors)​​ received an best paper​​​‌ award at ALT’25.

6‌ New results

6.1 Auctions‌​‌ and mechanism design

Participants:​​ Benjamin Heymann, Vianney​​​‌ Perchet, Maxime Vono‌.

In 17,‌​‌ we compare uniform price​​ and discriminatory multi-unit auctions​​​‌ through the lens of‌ regret minimization. By framing‌​‌ the strategic participation of​​ bidders as an online​​​‌ learning problem, we establish‌ new theoretical performance bounds‌​‌ for these fundamental pricing​​ rules.

We also deeply​​​‌ investigate settings with explicit‌ constraints. In 2,‌​‌ we study competitive and​​ revenue-optimal pricing when buyers​​​‌ have budget limits, and‌ in 12, we‌​‌ tackle the problem of​​ selling multiple complements with​​​‌ packaging costs, designing mechanisms‌ that account for the‌​‌ overhead of combinatorial bundling.​​ In 30, we​​​‌ explore side-by-side first-price auctions‌ in the presence of‌​‌ imperfect bidders, extending the​​ understanding of bidder behavior​​​‌ in decentralized environments. Finally,‌ to support applied research‌​‌ in this space, we​​ introduce CriteoPrivateAds in 31​​​‌, a real-world bidding‌ dataset specifically designed to‌​‌ help the community build​​ and test private advertising​​​‌ systems.

6.2 Matching markets‌

Participants: Matthieu Lerasle,‌​‌ Patrick Loiseau, Simon​​ Mauras, Vianney Perchet​​​‌.

In 5,‌ we provide a direct‌​‌ mathematical proof of the​​ short-side advantage in random​​​‌ matching markets, formally demonstrating‌ why the side with‌​‌ fewer participants in unbalanced​​ bipartite markets inherently secures​​​‌ vastly superior matches.

We‌ expand the scope of‌​‌ classical matching by introducing​​ realistic structural constraints. In​​​‌ 1, we explore‌ the correlation of rankings‌​‌ in matching markets, analyzing​​ how interdependent participant preferences​​​‌ impact stability and efficiency.‌ In 26, we‌​‌ investigate two-sided matching under​​ resource-regional caps, and in​​​‌ 25, we focus‌ on the theoretical properties‌​‌ of optimal unimodular matchings.​​ Applying these matching frameworks​​​‌ to real-world policy, we‌ analyze college admissions in‌​‌ France in 32,​​ evaluating the structural impacts​​​‌ of affirmative action, overlapping‌ reserves, and housing quotas.‌​‌

6.3 Bandits, reinforcement learning,​​​‌ and games

Participants: Benjamin​ Heymann, Vianney Perchet​‌, Hugo Richard.​​

We significantly advanced the​​​‌ theory of learning in​ games and multi-agent systems.​‌ In 20, accepted​​ at AAMAS 2025, we​​​‌ study learning in games​ with progressive hiding, proposing​‌ algorithms that allow agents​​ to strategically conceal information​​​‌ while maintaining efficiency. In​ 11, we solve​‌ the “harder path” of​​ achieving last-iterate convergence for​​​‌ uncoupled learning in zero-sum​ games with bandit feedback,​‌ proving that independent learners​​ can reach equilibrium even​​​‌ with limited feedback.

Our​ work on bandit frameworks​‌ spans numerous complex environments.​​ In 4, we​​​‌ propose robust algorithms for​ adversarial bandits facing arbitrary​‌ strategies. For combinatorial and​​ infinite action spaces, we​​​‌ developed oracle-efficient combinatorial semi-bandits​ 14, established methods​‌ for tracking significant shifts​​ in infinite-armed bandits 18​​​‌, and extended linear​ bandits beyond inner product​‌ spaces utilizing bandit optimal​​ transport 23. We​​​‌ also study bandits under​ strict constraints, such as​‌ multi-armed bandits with minimum​​ aggregated revenue constraints 21​​​‌ and offline contextual bandits​ with counterfactual sample identification​‌ 27.

We further​​ investigate fundamental RL limits​​​‌ and the intersection of​ learning with privacy. In​‌ 6, we formalize​​ the hardness of reinforcement​​​‌ learning with transition look-ahead.​ In privacy, we establish​‌ the optimal best arm​​ identification under differential privacy​​​‌ 13 and provide optimal​ regret bounds for general​‌ bandits under differential privacy​​ constraints 7.

6.4​​​‌ Online algorithms and prophet​ inequalities

Participants: Matthieu Lerasle​‌, Patrick Loiseau,​​ Simon Mauras, Vianney​​​‌ Perchet.

In 15​, we revisit prophet​‌ inequalities and demonstrate that​​ competing with the top​​​‌ items is surprisingly​ easy, establishing vastly improved​‌ competitive ratios when the​​ benchmark is relaxed from​​​‌ the single absolute best​ item. In 10,​‌ we investigate online combinatorial​​ allocation with interdependent values,​​​‌ bringing together online secretary-style​ selection with complex valuation​‌ structures to provide the​​ first known approximations for​​​‌ this sequential mechanism design​ problem.

We also heavily​‌ focused on the tradeoffs​​ inherent in learning-augmented algorithms.​​​‌ In 9, we​ systematically analyze how online​‌ algorithms can leverage potentially​​ imperfect machine-learned predictions while​​​‌ maintaining robust worst-case guarantees.​ We extend this in​‌ 8, exploring Pareto-optimality,​​ smoothness, and stochasticity in​​​‌ learning-augmented One-Max-Search problems. Finally,​ in 24, we​‌ establish optimal regret under​​ local search trees for​​​‌ contextual ranking and matching.​

6.5 Data valuation, privacy,​‌ and causal inference

Participants:​​ Cristina Butucea, Benjamin​​​‌ Heymann, Patrick Loiseau​, Maxime Vono,​‌ Hugo Richard.

We​​ extensively researched data valuation​​​‌ methodologies to ensure fair​ credit assignment in machine​‌ learning. In 34,​​ we provide an efficient​​​‌ Shapley value approximation via​ language model arithmetic for​‌ LLM fine-tuning data valuation.​​ We further analyze the​​​‌ impact of the utility​ function itself in semivalue-based​‌ data valuation pipelines in​​ 33.

In the​​​‌ realm of privacy and​ fairness, we present a​‌ framework for distribution-aware mean​​ estimation under user-level local​​​‌ differential privacy in 16​, improving the statistical​‌ utility of private aggregations.​​ We also highlight critical​​ gaps in current AI​​​‌ ethics frameworks, showing in‌ 37 that fairness in‌​‌ Generative AI remains understudied​​ and undervalued.

For causal​​​‌ inference and counterfactual reasoning,‌ we introduce CUVET in‌​‌ 19, a scalable​​ partitioning approach for continuous​​​‌ treatment assignment. We also‌ developed methods for non-linear‌​‌ counterfactual aggregate optimization 29​​ and counterfactual simulations for​​​‌ large-scale systems with burnout‌ variables 28.

Finally,‌​‌ we apply our statistical​​ frameworks to specialized data​​​‌ modalities: solving off-the-grid learning‌ of mixtures from a‌​‌ continuous dictionary 22,​​ developing supervised contamination detection​​​‌ for flow cytometry applications‌ 3, and utilizing‌​‌ feature-augmented Hawkes processes 36​​ alongside explainable expected goal​​​‌ metrics 35 for advanced‌ event-level attribution in sports‌​‌ analytics.

7 Partnerships and​​ cooperations

7.1 National initiatives​​​‌

Foundry (PEPR IA)

Participants:‌ Patrick Loiseau.

  • Title:‌​‌
    Foundry: Foundation of robustness​​ and reliability in AI​​​‌
  • Partner Institution(s):
    • Inria
    • CNRS‌
    • Université Paris Dauphine
    • Institut‌​‌ Mines Telecom
    • ENS de​​ Lyon
  • Date/Duration:
    2023-2027 (4​​​‌ years)
  • Additionnal info/keywords:
    PEPR‌ IA projet cible, 245k‌​‌ euros. Fairness, matching, auctions.​​
FairPlay (ANR JCJC)

Participants:​​​‌ Patrick Loiseau.

  • Title:‌
    FairPlay: Fair algorithms via‌​‌ game theory and sequential​​ learning
  • Partner Institution(s):
    • Inria​​​‌
  • Date/Duration:
    2021-2025 (4 years)‌
  • Additionnal info/keywords:
    ANR JCJC‌​‌ project, 245k euros. Fairness,​​ matching, auctions.
DOOM (ANR)​​​‌

Participants: Vianney Perchet.‌

  • Title:
    DOOM: Design of‌​‌ Optimal Online Matching Markets​​
  • Partner Institution(s):
    • Crest, Genes​​​‌
  • Date/Duration:
    2024-2028 (4 years)‌
  • Additionnal info/keywords:
    ANR project,‌​‌ 456k euros. online markets,​​ learning, matching.

7.2 Regional​​​‌ initiatives

DOLLY (Hi! Paris)‌

Participants: Vianney Perchet,‌​‌ Patrick Loiseau.

  • Title:​​
    Design, Incentivization, Optimization and​​​‌ (Reinforcement-)Learning of Multi-Layered Market‌
  • Partner Institution(s):
    • Crest, Genes‌​‌
    • Inria
  • Date/Duration:
    2025-2029 (4​​ years)
  • Additionnal info/keywords:
    Hi!​​​‌ Paris internal synergy fellowship,‌ 720k euros. online markets,‌​‌ learning, matching.

8 Dissemination​​

8.1 Promoting scientific activities​​​‌

8.1.1 Scientific events: organisation‌

Member of the organizing‌​‌ committees

Participants: Vianney Perchet​​.

  • Title:
    From matchings​​​‌ to markets. A tale‌ of Mathematics, Economics and‌​‌ Computer Science.
  • Partner Institution(s):​​
    • Crest, Genes
  • Date/Duration:
    April​​​‌ 2025
  • Location:
    Cargese

Participants:‌ Vianney Perchet.

  • Title:‌​‌
    Member of the scientific​​ advisory committee of the​​​‌ Hi! Paris summer school‌
  • Partner Institution(s):
  • Date/Duration:
    July‌​‌ 2025
  • Location:
    Ecole Polytechnique​​

Participants: Mariia Vladimirova.​​​‌

  • Title:
    Organizer of the‌ Trustworthy AI symposium labeled‌​‌ as AI Action Summit​​ event
  • Date/Duration:
    January 2025​​​‌
  • Location:
    Criteo HQ, PAris‌

8.1.2 Scientific events: selection‌​‌

Chair of conference program​​ committees

Participants: Patrick Loiseau​​​‌.

  • Title:
    Program chair‌ of the Hi! Paris‌​‌ summer school
  • Date/Duration:
    July​​ 2025
  • Location:
    Ecole Polytechnique​​​‌
Member of the conference‌ program committees
  • Patrick Loiseau:‌​‌
    ICML, NeurIPS
  • Vianney Perchet:​​
    NeurIPS, ICLR, ICML, COLT,​​​‌ ALT, EC
  • Simon Mauras:‌
    WINE, EC
  • Hugo Richard:‌​‌
    NeurIPS, ICML, AISTATS
  • Clément​​ Calauzènes:
    ICML, NeurIPS
  • Benjamin​​​‌ Heymann
    AISTATS
  • Maxime Vono‌
    NeurIPS

8.1.3 Invited talks‌​‌

  • Vianney Perchet: Speaker at​​ the French Scientific Summit​​​‌ on Artificial Intelligence (organized‌ by the government at‌​‌ Ecole Polytechnique in Feb​​ 2025).
  • Patrick Loiseau: Keynote​​​‌ speaker at Netgcoop 2025‌
  • Mariia Vladimirova: Keynote talks‌​‌ at Women in Programmatic​​ Network (9 April 2025,​​​‌ Paris, France) and European‌ Women in Tech (26‌​‌ June 2025, Amsterdam, Netherlands).​​​‌ Invited talks for Women​ at Privacy on AI​‌ essentials (30 May 2024)​​ and the University of​​​‌ Manchester (19 February 2025,​ Manchester, Great Britain)
  • Cristina​‌ Butucea: Invited talks (2025)​​ Tsinghua University Beijing, Southeast​​​‌ University Nanjing, China; Symposium​ on Mathematical Foundations of​‌ Trustworthy Machine Learning, Ascona,​​ Switzerland;

8.1.4 Leadership within​​​‌ the scientific community

Vianney​ Perchet is president of​‌ the Scientific Committee, Program​​ Gaspard Monge for Optimisation,​​​‌ Operations Research and Data-Science,​ Paris-Saclay (2021-2025). Several members​‌ of the team are​​ regularly called as experts​​​‌ for their scientific expertise​ in various selection processes:​‌

  • selection of research grants:​​ Mariia Vladimirova was expert​​​‌ for ANR AAPG projects​ (2024), Patrick Loiseau was​‌ expert for the ANR​​ and the Royal Society​​​‌ (2024), Vianney Perchet, Cristina​ Butucea and Patrick Loiseau​‌ were members of ANR​​ evaluation panels
  • recruitment committess​​​‌ (many in France, some​ in Germany), habilitation juries,​‌ and tenure committees (Tufts,​​ Hambourg, Harvard)

8.2 Teaching​​​‌ - Supervision - Juries​ - Educational and pedagogical​‌ outreach

8.2.1 Supervision

  • Patrick​​ Loiseau
    : PhD students:​​​‌ Mathieu Molina, Reda Jalal,​ Minrui Xu, Marie Generali,​‌ Melissa Tamine, Louise Allain;​​ postdocs: Simon Finster, Denis​​​‌ Sokolov
  • Vianney Perchet
    :​ PhD students: Come Fiegel,​‌ Maria Cherifa, Mathieu Molina,​​ Ziyad Benomar, Mike Liu,​​​‌ Hafedh El Ferchichi, Matilde​ Tullii, Giovanni Montanari, Naila​‌ Sebastian Esandi. postdocs: Bartholomé​​ Vieille, Achraf Azize, Junghun​​​‌ Kim
  • Matthieu Lerasle
    PhD​ Students: Clara Carlier, Hugo​‌ Chardon, Hafedh El Ferchichi.​​
  • Cristina Butucea
    PhD students:​​​‌ Nayel Bettache, Henning Stein,​ Antoine Schoonaert
  • Marc Abeille​‌
    : PhD student: Ahmed​​ Ben Yahmed
  • Clément Calauzènes​​​‌
    : PhD student: Morgane​ Goibert, Maria Cherifa
  • Benjamin​‌ Heymann
    : PhD student:​​ Mélissa Tamine
  • Maxime Vono​​​‌
    : PhD student: Mélissa​ Tamine
  • Simon Mauras
    :​‌ PhD student: Minrui Xu,​​ Reda Jalal.

8.2.2 Teaching​​​‌

  • ENSAE:
     
    • Algorithm design and​ analysis
      3A lectures
    • Programming​‌ projet
      1A
    • Advanced Optimization​​
      Third year, lectures
    • Theoretical​​​‌ Foundations of Machine Learning​
      Second year, lectures
    • Stopping​‌ time and online algorithms​​
      Third year, lectures
    • Statistics​​​‌
      (ML) 1st and second​ year
    • Nonparametric Statistics
      3rd​‌ year, M2
    • Mathematical Foundations​​ of Probabilities
      1st year​​​‌
  • Ecole Polytechnique:
     
    • INF421: design​ and analysis of algorithms​‌
      (Patrick Loiseau). Second-year level,​​ PCs.
    • INF581: Advanced Machine​​​‌ Learning and Autonomous Agents​
      (Patrick Loiseau). Third-year/M1 level,​‌ lectures and labs.
    • MAP433:​​ Statistics
      (ML). First-year cycle​​​‌ polytechnicien, PCs.
    • MAP576: Learning​ Theory
      (ML). Second-year cycle​‌ polytechnicien, Lecture.
  • Université Paris-Saclay:​​
    • High Dimensional Probability
      (ML).​​​‌ Master 2
    • Stopping Time​ and Random Algorithm
      (ML).​‌ Master 2
  • PSL:
    • Introduction​​ to machine learning
      (Hugo​​​‌ Richard). L3 level, Lectures​ and labs.
  • Master IASD:​‌
    • Recommender Systems
      (Clément Calauzènes).​​ Master 2, Lectures.

8.3​​​‌ Popularization

8.3.1 Specific official​ responsibilities in science outreach​‌ structures

In March 2025,​​ Simon Mauras joined Inria​​​‌ Saclay’s scientific outreach team​ as the new scientific​‌ referent of the research​​ center. In this capacity,​​​‌ Simon participates in the​ animation of external events​‌ (CaféIA, Fête de la​​ Science) and school outreach​​​‌ activities (school visits with​ the “Chiche!” program).

8.3.2​‌ Productions (articles, videos, podcasts,​​ serious games, ...)

Beyond​​​‌ actions in scientific mediation,​ our team is strongly​‌ committed to the scientific​​ dissemination of our research,​​ and interacts in a​​​‌ regular basis with the‌ communication team to produce‌​‌ resources aimed at the​​ general public:

8.3.3 Others‌​‌ science outreach relevant activities​​

Simon Mauras and Patrick​​​‌ Loiseau also participated in‌ the Inria Saclay program‌​‌ to provide internships for​​ students in “3eme” (middle​​​‌ school) and “2nde” (high‌ school). They created activities‌​‌ based on matching problems​​ motivated by Parcoursup to​​​‌ show these young students‌ what science looks like‌​‌ and what impact it​​ can have in practice.​​​‌

9 Scientific production

9.1‌ Publications of the year‌​‌

International journals

International peer-reviewed​​​‌ conferences

  • 7 inproceedingsA.‌Achraf Azize, Y.‌​‌Yulian Wu, J.​​Junya Honda, F.​​​‌Francesco Orabona, S.‌Shinji Ito and D.‌​‌Debabrota Basu. Optimal​​ Regret of Bandits under​​​‌ Differential Privacy.NeurIPS‌ 2025 - 39th Annual‌​‌ Conference on Neural Information​​ Processing SystemsSan Diego​​​‌ (USA), United States2025‌HALback to text‌​‌
  • 8 inproceedingsZ.Ziyad​​ Benomar, L.Lorenzo​​​‌ Croissant, V.Vianney‌ Perchet and S.Spyros‌​‌ Angelopoulos. Pareto-Optimality, Smoothness,​​ and Stochasticity in Learning-Augmented​​​‌ One-Max-Search.ICML 2025‌ - 42nd International Conference‌​‌ on Machine LearningVancouver,​​ CanadaJuly 2025HAL​​​‌back to text
  • 9‌ inproceedingsZ.Ziyad Benomar‌​‌ and V.Vianney Perchet​​. On Tradeoffs in​​​‌ Learning-Augmented Algorithms.AISTATS‌ 2025 - The 28th‌​‌ International Conference on Artificial​​ Intelligence and StatisticsPhuket,​​​‌ ThailandJanuary 2025HAL‌back to text
  • 10‌​‌ inproceedingsM.Michal Feldman​​​‌, S.Simon Mauras​, D.Divyarthi Mohan​‌ and R.Rebecca Reiffenhäuser​​. Online Combinatorial Allocation​​​‌ with Interdependent Values.​EC 2025 - 26th​‌ ACM Conference on Economics​​ and ComputationStanford, United​​​‌ StatesACMJuly 2025​, 189-205HALDOI​‌back to text
  • 11​​ inproceedingsC.Côme Fiegel​​​‌, P.Pierre Menard​, T.Tadashi Kozuno​‌, M.Michal Valko​​ and V.Vianney Perchet​​​‌. The Harder Path:​ Last Iterate Convergence for​‌ Uncoupled Learning in Zero-Sum​​ Games with Bandit Feedback​​​‌.Proceedings of Machine​ Learning ResearchICML 2025​‌ - 42nd International Conference​​ on Machine Learning267​​​‌Vancouver, Canada2025HAL​back to text
  • 12​‌ inproceedingsS.Simon Finster​​. Selling Multiple Complements​​​‌ with Packaging Costs.​EC 2025 - 26th​‌ ACM Conference on Economics​​ and ComputationStanford (CA),​​​‌ United StatesACMJuly​ 2025, 250-250HAL​‌DOIback to text​​
  • 13 inproceedingsM.Marc​​​‌ Jourdan and A.Achraf​ Azize. Optimal Best​‌ Arm Identification under Differential​​ Privacy.NeurIPS 2025​​​‌ - 39thh Annual Conference​ on Neural Information Processing​‌ SystemsSan Diego, United​​ StatesOctober 2025HAL​​​‌back to text
  • 14​ inproceedingsJ.-H.Jung-Hun Kim​‌, M.Milan Vojnović​​ and M.-H.Min-Hwan Oh​​​‌. Oracle-Efficient Combinatorial Semi-Bandits​.The Thirty-Ninth Annual​‌ Conference on Neural Information​​ Processing SystemsSan Diego,​​​‌ United StatesDecember 2025​HALback to text​‌
  • 15 inproceedingsM.Mathieu​​ Molina, N.Nicolas​​​‌ Gast, P.Patrick​ Loiseau and V.Vianney​‌ Perchet. Prophet Inequalities:​​ Competing with the Top​​​‌ Items is Easy.​SODA 2025 - ACM-SIAM​‌ Symposium on Discrete Algorithms​​New Orleans, United States​​​‌ACM2025, 1-45​HALback to text​‌
  • 16 inproceedingsC.Corentin​​ Pla, M.Maxime​​​‌ Vono and H.Hugo​ Richard. Distribution-Aware Mean​‌ Estimation under User-level Local​​ Differential Privacy.Proceedings​​​‌ of Machine Learning Research​AISTATS 2025 - 28th​‌ International Conference on Artificial​​ Intelligence and Statistics258​​​‌Phuket, ThailandMay 2025​HALback to text​‌
  • 17 inproceedingsM.Marius​​ Potfer and V.Vianney​​​‌ Perchet. Comparing Uniform​ Price and Discriminatory Multi-Unit​‌ Auctions through Regret Minimization​​.NeurIPS 2025 -​​​‌ 39th Annual Conference on​ Neural Information Processing Systems​‌San Diego (California), United​​ StatesDecember 2025HAL​​​‌back to text
  • 18​ inproceedingsJ.Joe Suk​‌ and J.-H.Jung-Hun Kim​​. Tracking Most Significant​​​‌ Shifts in Infinite-Armed Bandits​.ICML 2025 -​‌ 42nd International Conference on​​ Machine LearningVancouver, Canada​​​‌July 2025HALback​ to text

Conferences without​‌ proceedings

Reports &​‌ preprints

9.2 Cited​‌ publications

  • 38 unpublished24​​ CFR § 100.75 -​​​‌ Discriminatory advertisements, statements and​ notices.., https://www.law.cornell.edu/cfr/text/24/100.75back​‌ to text
  • 39 article​​E.Emmanuel Abbe.​​​‌ Community detection and stochastic​ block models: recent developments​‌.The Journal of​​ Machine Learning Research18​​​‌12017, 6446--6531​back to text
  • 40​‌ inproceedingsM.Muhammad Ali​​, P.Piotr Sapiezynski​​​‌, M.Miranda Bogen​, A.Aleksandra Korolova​‌, A.Alan Mislove​​ and A.Aaron Rieke​​​‌. Discrimination through optimization:​ How Facebook's ad delivery​‌ can lead to skewed​​ outcomes.CSCW2019​​​‌back to text
  • 41​ inproceedingsN.Noga Alon​‌, N.Nicolo Cesa-Bianchi​​, O.Ofer Dekel​​​‌ and T.Tomer Koren​. Online learning with​‌ feedback graphs: Beyond bandits​​.Annual Conference on​​​‌ Learning Theory40Microtome​ Publishing2015back to​‌ text
  • 42 articleI.​​Itai Ashlagi, M.​​​‌Maximilien Burq, P.​Patrick Jaillet and V.​‌Vahideh Manshadi. On​​ matching and thickness in​​​‌ heterogeneous dynamic markets.​Operations Research674​‌2019, 927--949back​​ to text
  • 43 misc​​​‌I.Itai Ashlagi,​ P.Patrick Jaillet and​‌ V. H.Vahideh H.​​ Manshadi. Kidney Exchange​​​‌ in Dynamic Sparse Heterogenous​ Pools.2013back​‌ to text
  • 44 inproceedings​​P.Pranjal Awasthi and​​​‌ T.Tuomas Sandholm.​ Online stochastic optimization in​‌ the large: Application to​​ kidney exchange.Twenty-First​​​‌ International Joint Conference on​ Artificial Intelligence2009back​‌ to text
  • 45 inproceedings​​B.Bahman Bahmani and​​​‌ M.Michael Kapralov.​ Improved bounds for online​‌ stochastic matching.European​​ Symposium on AlgorithmsSpringer​​​‌2010, 170--181back​ to text
  • 46 inproceedings​‌E.E. del Barrio​​, F.F. Gamboa​​​‌, P.P. Gordaliza​ and J.-M.J.-M. Loubes​‌. Obtaining fairness using​​ optimal transport theory.​​​‌ arXiv:1806.031952018, 1--25​back to text
  • 47​‌ articleB.Benjamin Birnbaum​​ and C.Claire Mathieu​​​‌. On-line bipartite matching​ made simple.Acm​‌ Sigact News391​​2008, 80--87back​​​‌ to text
  • 48 article​B.Bela Bollobas and​‌ G.Graham Brightwell.​​ The width of random​​​‌ graph orders.The​ Mathematical Scientist2001​‌ 1995back to text​​
  • 49 miscC.Charles​​​‌ Bordenave, M.Marc​ Lelarge and J.Justin​‌ Salez. Matchings on​​ infinite graphs.2011​​​‌back to text
  • 50​ bookA.A. Borodin​‌ and R.R. El-Yaniv​​. Online Computation and​​​‌ Competitive Analysis.Cambridge​ University Presss1998back​‌ to text
  • 51 misc​​A.Allan Borodin,​​​‌ C.Christodoulos Karavasilis and​ D.Denis Pankratov.​‌ An Experimental Study of​​ Algorithms for Online Bipartite​​​‌ Matching.2018back​ to text
  • 52 inproceedings​‌E.E. Boursier and​​ V.V. Perchet.​​​‌ SIC-MMAB: Synchronisation Involves Communication​ in Multiplayer Multi-Armed Bandits​‌.arXiv:1809.081512018,​​ 1--31back to text​​
  • 53 inproceedingsE.Etienne​​​‌ Boursier and V.Vianney‌ Perchet. Utility/Privacy Trade-off‌​‌ through the lens of​​ Optimal Transport.Proceedings​​​‌ of the Twenty Third‌ International Conference on Artificial‌​‌ Intelligence and Statistics108​​Proceedings of Machine Learning​​​‌ ResearchOnlinePMLRAugust‌ 2020, 591--601back‌​‌ to textback to​​ text
  • 54 articleS.​​​‌S. Bubeck and N.‌N. Cesa-Bianchi. Regret‌​‌ Analysis of Stochastic and​​ Nonstochastic Multi-armed Bandit Problems​​​‌.Machine Learning5‌12012, 1--122‌​‌back to text
  • 55​​ articleS.S. Bubeck​​​‌, V.V. Perchet‌ and P.P. Rigollet‌​‌. Bounded regret in​​ stochastic multi-armed bandits.​​​‌Journal of Machine Learning‌ Research: Workshop and Conference‌​‌ Proceedings (COLT)302013​​, 122--134back to​​​‌ text
  • 56 inproceedingsN.‌Niv Buchbinder, K.‌​‌Kamal Jain and J.​​ (.Joseph (Seffi) Naor​​​‌. Online Primal-Dual Algorithms‌ for Maximizing Ad-Auctions Revenue‌​‌.Berlin, Heidelberg2007​​, 253--264back to​​​‌ textback to text‌
  • 57 articleC.Cristina‌​‌ Butucea, A.Amandine​​ Dubois, M.Martin​​​‌ Kroll, A.Adrien‌ Saumard and others.‌​‌ Local differential privacy: Elbow​​ effect in optimal density​​​‌ estimation and adaptation over‌ Besov ellipsoids.Bernoulli‌​‌2632020,​​ 1727--1764back to text​​​‌
  • 58 articleC.Cristina‌ Butucea, A.Angelika‌​‌ Rohde and L.Lukas​​ Steinberger. Interactive versus​​​‌ non-interactive locally, differentially private‌ estimation: Two elbows for‌​‌ the quadratic functional.​​Annals of Statsto​​​‌ appear2023back to‌ text
  • 59 inproceedingsL.‌​‌ E.L Elisa Celis​​, A.Anay Mehrotra​​​‌ and N. K.Nisheeth‌ K Vishnoi. Toward‌​‌ Controlling Discrimination in Online​​ Ad Auctions.ICML​​​‌2019back to text‌back to text
  • 60‌​‌ inproceedingsN.N. Cesa-Bianchi​​, C.C. Gentile​​​‌ and Y.Y. Mansour‌. Regret minimization for‌​‌ reserve prices in second-price​​ auctions.Proceedings of​​​‌ SODA 20132013back‌ to text
  • 61 book‌​‌N.Nicolò Cesa-Bianchi and​​ G.Gabor Lugosi.​​​‌ Prediction, Learning, and Games‌.Cambridge University Press‌​‌2006back to text​​
  • 62 inproceedingsK.K.​​​‌ Chaudhuri, J.J.‌ Imola and A.A.‌​‌ Machanavajjhala. Capacity bounded​​ differential privacy.Advances​​​‌ in Neural Information Processing‌ Systems2019, 3469--3478‌​‌back to text
  • 63​​ miscS.Shuchi Chawla​​​‌ and M.Meena Jagadeesan‌. Fairness in ad‌​‌ auctions through inverse proportionality​​.arXiv:2003.139662020back​​​‌ to textback to‌ text
  • 64 inproceedingsI.‌​‌ R.Ilan Reuven Cohen​​ and D.David Wajc​​​‌. Randomized online matching‌ in regular graphs.‌​‌Proceedings of the Twenty-Ninth​​ Annual ACM-SIAM Symposium on​​​‌ Discrete AlgorithmsSIAM2018‌, 960--979back to‌​‌ text
  • 65 inproceedingsV.​​Vincent Conitzer, C.​​​‌Christian Kroer, D.‌Debmalya Panigrahi, O.‌​‌Okke Schrijvers, E.​​Eric Sodomka, N.​​​‌ E.Nicolas E Stier-Moses‌ and C.Chris Wilkens‌​‌. Pacing equilibrium in​​ first-price auction markets.​​​‌EC2019back to‌ text
  • 66 inproceedingsV.‌​‌Vincent Conitzer, C.​​Christian Kroer, E.​​​‌Eric Sodomka and N.‌ E.Nicolas E Stier-Moses‌​‌. Multiplicative pacing equilibria​​​‌ in auction markets.​WINE2018back to​‌ text
  • 67 inproceedingsJ.​​José Correa, P.​​​‌Paul Dütting, F.​Felix Fischer and K.​‌Kevin Schewior. Prophet​​ inequalities for iid random​​​‌ variables from an unknown​ distribution.Proceedings of​‌ the 2019 ACM Conference​​ on Economics and Computation​​​‌2019, 3--17back​ to text
  • 68 inproceedings​‌R.Rachel Cummings,​​ V.Varun Gupta,​​​‌ D.Dhamma Kimpara and​ J.Jamie Morgenstern.​‌ On the Compatibility of​​ Privacy and Fairness.​​​‌FairUMAP2019back to​ text
  • 69 miscJ.​‌Jeffrey Dastin. Amazon​​ scraps secret AI recruiting​​​‌ tool that showed bias​ against women.Reuters,​‌ https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G2018back to​​ text
  • 70 articleC.​​​‌C. Dwork. Differential​ privacy.Encyclopedia of​‌ Cryptography and Security2011​​, 338--340back to​​​‌ text
  • 71 inproceedingsC.​Cynthia Dwork, M.​‌Moritz Hardt, T.​​Toniann Pitassi, O.​​​‌Omer Reingold and R.​Richard Zemel. Fairness​‌ through awareness.ITCS​​2012back to text​​​‌
  • 72 inproceedingsC.Cynthia​ Dwork, F.Frank​‌ McSherry, K.Kobbi​​ Nissim and A.Adam​​​‌ Smith. Calibrating noise​ to sensitivity in private​‌ data analysis.Theory​​ of cryptography conferenceSpringer​​​‌2006, 265--284back​ to text
  • 73 techreport​‌R.R. Eilat,​​ K.K. Eliaz and​​​‌ X.X. Mu.​ Optimal Privacy-Constrained Mechanisms.​‌C.E.P.R. Discussion Papers2019​​back to text
  • 74​​​‌ inproceedingsV.Vitalii Emelianov​, N.Nicolas Gast​‌, K. P.Krishna​​ P. Gummadi and P.​​​‌Patrick Loiseau. On​ the Effect of Positive​‌ Discrimination on Multistage Selection​​ Problems in the Presence​​​‌ of Implicit Variance.​EC2020back to​‌ text
  • 75 miscJ.​​Jon Feldman, A.​​​‌Aranyak Mehta, V.​Vahab Mirrokni and S.​‌S. Muthukrishnan. Online​​ Stochastic Matching: Beating 1-1/e​​​‌.2009back to​ text
  • 76 techreportK.​‌Kadija Ferryman and M.​​Mikaela Pitcan. Fairness​​​‌ in Precision Medicine.​Data & Society2018​‌back to text
  • 77​​ bookD.D. Fudenberg​​​‌ and J.J. Tirole​. Game Theory.​‌MIT press1991back​​ to text
  • 78 article​​​‌E.Evrard Garcelon,​ V.Vianney Perchet,​‌ C.Ciara Pike-Burke and​​ M.Matteo Pirotta.​​​‌ Local Differentially Private Regret​ Minimization in Reinforcement Learning​‌.arXiv preprint arXiv:2010.07778​​2020back to text​​​‌
  • 79 articleN.Nicolas​ Gast, S.Stratis​‌ Ioannidis, P.Patrick​​ Loiseau and B.Benjamin​​​‌ Roussillon. Linear Regression​ from Strategic Data Sources​‌.ACM Transactions on​​ Economics and Computation8​​​‌2May 2020,​ 10:1--10:24back to text​‌
  • 80 inproceedingsN.Nicolas​​ Gast and B.Benny​​​‌ Van Houdt. A​ Refined Mean Field Approximation​‌.SIGMETRICS2017back​​ to text
  • 81 article​​​‌M.Maya Gupta,​ A.Andrew Cotter,​‌ M. M.Mahdi Milani​​ Fard and S.Serena​​​‌ Wang. Proxy fairness​.arXiv preprint arXiv:1806.11212​‌2018back to text​​
  • 82 inproceedingsA.Anikó​​​‌ Hannák, C.Claudia​ Wagner, D.David​‌ Garcia, A.Alan​​ Mislove, M.Markus​​ Strohmaier and C.Christo​​​‌ Wilson. Bias in‌ online freelance marketplaces: Evidence‌​‌ from taskrabbit and fiverr​​.CSCW2017back​​​‌ to text
  • 83 inproceedings‌M.Moritz Hardt,‌​‌ E.Eric Price and​​ N.Nathan Srebro.​​​‌ Equality of Opportunity in‌ Supervised Learning.NIPS‌​‌2016back to text​​back to text
  • 84​​​‌ inproceedingsC.-J.Chien-Ju Ho‌ and J. W.Jennifer‌​‌ Wortman Vaughan. Online​​ task assignment in crowdsourcing​​​‌ markets.Twenty-sixth AAAI‌ conference on artificial intelligence‌​‌2012back to text​​back to text
  • 85​​​‌ inproceedingsC.Christina Ilvento‌, M.Meena Jagadeesan‌​‌ and S.Shuchi Chawla​​. Multi-Category Fairness in​​​‌ Sponsored Search Auctions.‌FAT*2020back to‌​‌ textback to text​​
  • 86 articleK.Krishnamurthy​​​‌ Iyer, R.Ramesh‌ Johari and M.Mukund‌​‌ Sundararajan. Mean field​​ equilibria of dynamic auctions​​​‌ with learning.Management‌ Science60122014‌​‌, 2949--2970back to​​ text
  • 87 articleP.​​​‌Patrick Jaillet and X.‌Xin Lu. Online‌​‌ stochastic matching: New algorithms​​ with better bounds.​​​‌Mathematics of Operations Research‌3932014,‌​‌ 624--646back to text​​
  • 88 articleJ.Julie​​​‌ Josse, N.Nicolas‌ Prost, E.Erwan‌​‌ Scornet and G.Gaël​​ Varoquaux. On the​​​‌ consistency of supervised learning‌ with missing values.‌​‌arXiv preprint arXiv:1902.069312019​​back to text
  • 89​​​‌ articleN.Nathan Kallus‌, X.Xiaojie Mao‌​‌ and A.Angela Zhou​​. Assessing algorithmic fairness​​​‌ with unobserved protected class‌ using data combination.‌​‌arXiv preprint arXiv:1906.002852019​​back to text
  • 90​​​‌ inproceedingsC.Chinmay Karande‌, A.Aranyak Mehta‌​‌ and P.Pushkar Tripathi​​. Online bipartite matching​​​‌ with unknown distributions.‌Proceedings of the forty-third‌​‌ annual ACM symposium on​​ Theory of computing2011​​​‌, 587--596back to‌ text
  • 91 inproceedingsR.‌​‌ M.Richard M Karp​​, U. V.Umesh​​​‌ V Vazirani and V.‌ V.Vijay V Vazirani‌​‌. An optimal algorithm​​ for on-line bipartite matching​​​‌.Proceedings of the‌ twenty-second annual ACM symposium‌​‌ on Theory of computing​​1990, 352--358back​​​‌ to text
  • 92 inproceedings‌N.Niki Kilbertus,‌​‌ M. R.Mateo Rojas​​ Carulla, G.Giambattista​​​‌ Parascandolo, M.Moritz‌ Hardt, D.Dominik‌​‌ Janzing and B.Bernhard​​ Schölkopf. Avoiding discrimination​​​‌ through causal reasoning.‌Advances in Neural Information‌​‌ Processing Systems2017,​​ 656--666back to text​​​‌
  • 93 articleJ. S.‌Joon Sik Kim,‌​‌ J.Jiahao Chen and​​ A.Ameet Talwalkar.​​​‌ Model-Agnostic Characterization of Fairness‌ Trade-offs.arXiv preprint‌​‌ arXiv:2004.034242020back to​​ textback to text​​​‌
  • 94 bookD.D.E.‌ Knuth. Stable Marriage‌​‌ and Its Relation to​​ Other Combinatorial Problems: An​​​‌ Introduction to the Mathematical‌ Analysis of Algorithms.‌​‌English translation, (CRM Proceedings​​ and Lecture Notes), American​​​‌ Mathematical Society1996back‌ to text
  • 95 book‌​‌V.V. Krishna.​​ Auction Theory.Elsevier​​​‌2009back to text‌
  • 96 inproceedingsM. J.‌​‌Matt J Kusner,​​ J.Joshua Loftus,​​​‌ C.Chris Russell and‌ R.Ricardo Silva.‌​‌ Counterfactual fairness.Advances​​​‌ in neural information processing​ systems2017, 4066--4076​‌back to text
  • 97​​ miscM. J.Matt​​​‌ J Kusner and J.​ R.Joshua R Loftus​‌. The long road​​ to fairer algorithms.​​​‌2020back to text​
  • 98 articleA.Anja​‌ Lambrecht and C.Catherine​​ Tucker. Algorithmic Bias?​​​‌ An Empirical Study of​ Apparent Gender-Based Discrimination in​‌ the Display of STEM​​ Career Ads.Management​​​‌ Science2019back to​ textback to text​‌
  • 99 miscJ.Jeff​​ Larson, S.Surya​​​‌ Mattu, L.Lauren​ Kirchner and J.Julia​‌ Angwin. How We​​ Analyzed the COMPAS Recidivism​​​‌ Algorithm.ProPublica, https://www.propublica.org/article/how-we-analyzed-the-compas-recidivism-algorithm​2016back to text​‌
  • 100 inproceedingsL.L.​​ Liu, H.H.​​​‌ Mania and M. I.​M. I.| Jordan.​‌ Competing Bandits in Matching​​ Markets.arXiv:1906.053632019​​​‌, 1--15back to​ text
  • 101 articleB.​‌B. Maćkowiak and M.​​M. Wiederholt. Business​​​‌ cycle dynamics under rational​ inattention.The Review​‌ of Economic Studies82​​42015, 1502--1532​​​‌back to text
  • 102​ inproceedingsM.Mohammad Mahdian​‌ and Q.Qiqi Yan​​. Online bipartite matching​​​‌ with random arrivals: an​ approach based on strongly​‌ factor-revealing lps.Proceedings​​ of the forty-third annual​​​‌ ACM symposium on Theory​ of computing2011,​‌ 597--606back to text​​
  • 103 articleV. H.​​​‌Vahideh H Manshadi,​ S. O.Shayan Oveis​‌ Gharan and A.Amin​​ Saberi. Online stochastic​​​‌ matching: Online actions based​ on offline statistics.​‌Mathematics of Operations Research​​3742012,​​​‌ 559--573back to text​back to text
  • 104​‌ bookA.A. Mas-Colell​​, M.M. Whinston​​​‌ and J.J. Green​. Microeconomic Theory.​‌Oxford University Press1995​​back to text
  • 105​​​‌ articleA.Andrew Mastin​ and P.Patrick Jaillet​‌. Greedy online bipartite​​ matching on random graphs​​​‌.arXiv preprint arXiv:1307.2536​2013back to text​‌
  • 106 articleF.F.​​ Matėjka and A.A.​​​‌ McKay. Rational inattention​ to discrete choices: A​‌ new foundation for the​​ multinomial logit model.​​​‌American Economic Review105​12015, 272--98​‌back to text
  • 107​​ articleA.Aranyak Mehta​​​‌. Online Matching and​ Ad Allocation.Found.​‌ Trends Theor. Comput. Sci.​​84October 2013​​​‌, 265–368URL: https://doi.org/10.1561/0400000057​DOIback to text​‌back to textback​​ to text
  • 108 book​​​‌J.-F.Jean-Francois Mertens,​ S.Sylvain Sorin and​‌ S.Shmuel Zamir.​​ Repeated Games.Econometric​​​‌ Society MonographsCambridge University​ Press2015back to​‌ text
  • 109 articleJ.​​J. Munkres. Algorithms​​​‌ for the Assignment and​ Transportation Problems.Journal​‌ of the Society for​​ Industrial and Applied Mathematics​​​‌511957,​ 32--38back to text​‌
  • 110 inproceedingsM.Milad​​ Nasr and M. C.​​​‌Michael Carl Tschantz.​ Bidding Strategies with Gender​‌ Nondiscrimination Constraints for Online​​ Ad Auctions.FAT*​​​‌2020back to text​back to text
  • 111​‌ inproceedingsT.T. Nedelec​​, M.M. Abeille​​​‌, C.C. Calauzenes​, B.B. Heymann​‌, V.V. Perchet​​ and N.N. El​​ Karoui. The bidder’s​​​‌ standpoint : a simple‌ way to improve bidding‌​‌ strategies in revenue-maximizing auctions​​.Workshop on Learning​​​‌ in the Presence of‌ Strategic Behavior (EC 2019)‌​‌2019back to text​​
  • 112 inproceedingsT.T.​​​‌ Nedelec, N.N.‌ El Karoui and V.‌​‌V. Perchet. Learning​​ to bid in revenue-maximizing​​​‌ auctions.Proceedings of‌ the 36th International Conference‌​‌ on Machine Learning (ICML​​ 2019)2019, 4781--4789​​​‌back to text
  • 113‌ bookN.N. Nisan‌​‌, T.T. Roughgarden​​, E.E. Tardos​​​‌ and V.V. Vazirani‌. Algorithmic Game Theory‌​‌.New York, NY,​​ USACambridge University Press​​​‌2007back to text‌
  • 114 articleV.V.‌​‌ Perchet and M.M.​​ Quincampoix. A differential​​​‌ game on Wasserstein space.‌ Application to weak approachability‌​‌ with partial monitoring.​​Journal of Dynamics and​​​‌ Games62019,‌ 65--85back to text‌​‌
  • 115 articleV.V.​​ Perchet and P.P.​​​‌ Rigollet. The multi-armed‌ bandit problem with covariates‌​‌.Annals of Statistics​​412013, 693--721​​​‌back to text
  • 116‌ bookJ.Jonas Peters‌​‌, D.Dominik Janzing​​ and B.Bernhard Schölkopf​​​‌. Elements of causal‌ inference.The MIT‌​‌ Press2017back to​​ text
  • 117 bookG.​​​‌G. Peyré and M.‌M. Cuturi. Computational‌​‌ Optimal Transport.ArXiv:1803.00567​​2018back to text​​​‌
  • 118 inproceedingsB.Bashir‌ Rastegarpanah, M.Mark‌​‌ Crovella and K.Krishna​​ Gummadi. Fair Inputs​​​‌ and Fair Outputs: The‌ Incompatibility of Fairness in‌​‌ Privacy and Accuracy.​​FairUMAP2020back to​​​‌ text
  • 119 inproceedingsJ.‌J. Reed and B.‌​‌B.C. Pierce. Distance​​ makes the types grow​​​‌ stronger: a calculus for‌ differential privacy.ACM‌​‌ Sigplan Notices452010​​, 157--168back to​​​‌ text
  • 120 articleA.‌Angelika Rohde, L.‌​‌Lukas Steinberger and others​​. Geometrizing rates of​​​‌ convergence under local differential‌ privacy constraints.Annals‌​‌ of Statistics485​​2020, 2646--2670back​​​‌ to text
  • 121 article‌A.Alexey Romanov,‌​‌ M.Maria De-Arteaga,​​ H.Hanna Wallach,​​​‌ J.Jennifer Chayes,‌ C.Christian Borgs,‌​‌ A.Alexandra Chouldechova,​​ S.Sahin Geyik,​​​‌ K.Krishnaram Kenthapadi,‌ A.Anna Rumshisky and‌​‌ A. T.Adam Tauman​​ Kalai. What's in​​​‌ a Name? Reducing Bias‌ in Bios without Access‌​‌ to Protected Attributes.​​arXiv preprint arXiv:1904.052332019​​​‌back to text
  • 122‌ articleP.Piotr Sapiezynski‌​‌, A.Avijit Gosh​​, L.Levi Kaplan​​​‌, A.Alan Mislove‌ and A.Aaron Rieke‌​‌. Algorithms that" Don't​​ See Color": Comparing Biases​​​‌ in Lookalike and Special‌ Ad Audiences.arXiv‌​‌ preprint arXiv:1912.075792019back​​ to text
  • 123 article​​​‌C.C.A. Sims.‌ Implications of rational inattention‌​‌.Journal of monetary​​ Economics5032003​​​‌, 665--690back to‌ text
  • 124 inproceedingsT.‌​‌Till Speicher, M.​​Muhammad Ali, G.​​​‌Giridhari Venkatadri, F.‌ N.Filipe Nunes Ribeiro‌​‌, G.George Arvanitakis​​, F.Fabr\'icio Benevenuto​​​‌, K. P.Krishna‌ P. Gummadi, P.‌​‌Patrick Loiseau and A.​​​‌Alan Mislove. On​ the Potential for Discrimination​‌ in Online Targeted Advertising​​.ACM FAT*2018​​​‌back to text
  • 125​ bookC.C. Villani​‌. Topics in optimal​​ transportation.58Graduate​​​‌ studies in Mathematics, AMS​2003back to text​‌
  • 126 inproceedingsJ.J.​​ Weed, V.V.​​​‌ Perchet and P.P.​ Rigollet. Online learning​‌ in repeated auctions.​​Proceedings of COLT 2016​​​‌2016back to text​back to text
  • 127​‌ inproceedingsB.Blake Woodworth​​, S.Suriya Gunasekar​​​‌, M. I.Mesrob​ I. Ohannessian and N.​‌Nathan Srebro. Learning​​ Non-Discriminatory Predictors.COLT​​​‌2017back to text​
  • 128 inproceedingsY.Yongkai​‌ Wu, L.Lu​​ Zhang and X.Xintao​​​‌ Wu. Counterfactual Fairness:​ Unidentification, Bound and Algorithm.​‌.IJCAI2019,​​ 1438--1444back to text​​​‌
  • 129 inproceedingsY.Yongkai​ Wu, L.Lu​‌ Zhang, X.Xintao​​ Wu and H.Hanghang​​​‌ Tong. Pc-fairness: A​ unified framework for measuring​‌ causality-based fairness.Advances​​ in Neural Information Processing​​​‌ Systems2019, 3404--3414​back to text
  • 130​‌ inproceedingsM. B.Muhammad​​ Bilal Zafar, I.​​​‌Isabel Valera, M.​Manuel Gomez Rodriguez and​‌ K. P.Krishna P.​​ Gummadi. Fairness Beyond​​​‌ Disparate Treatment & Disparate​ Impact: Learning Classification Without​‌ Disparate Mistreatment.WWW​​2017back to text​​​‌
  • 131 inproceedingsM. B.​Muhammad Bilal Zafar,​‌ I.Isabel Valera,​​ M.Manuel Gomez Rodriguez​​​‌, K. P.Krishna​ P. Gummadi and A.​‌Adrian Weller. From​​ Parity to Preference-based Notions​​​‌ of Fairness in Classification​.NIPS2017back​‌ to text
  • 132 inproceedings​​R.Rich Zemel,​​​‌ Y.Yu Wu,​ K.Kevin Swersky,​‌ T.Toni Pitassi and​​ C.Cynthia Dwork.​​​‌ Learning Fair Representations.​ICML2013back to​‌ text
  1. 1One may​​ worry that the structure​​​‌ added adds biases. This​ is typically fine because​‌ we control for fairness​​ on the output directly,​​​‌ but this may lead​ to a degradation of​‌ the benefit of adding​​ structure.