The International Joint Conference on Artificial Intelligence (IJCAI) annually recognizes outstanding papers presented at the conference. This year, during the conference’s opening ceremony, three papers were honored as distinguished papers.
And the winners are…
Online Combinatorial Optimization with Group Fairness Constraints
Negin Golrezaei, Rad Niazadeh, Kumar Kshitij Patel, and Fransisca Susan
Abstract: As digital markets and services continue to expand, ensuring a safe and fair environment for all users is crucial. This requires implementing fairness constraints in the sequential decision-making processes of these platforms to guarantee equal treatment. However, this can be challenging as these processes often need to solve NP-complete problems with exponentially large decision spaces at each time step. To address this issue, we propose a general framework that integrates robustness and fairness into NP-complete problems, such as product ranking optimization and submodular function maximization. Our framework presents the problem as a max-min game between a principal player aiming to maximize the platform’s objective and an adversarial player responsible for group fairness constraints. We demonstrate that it is possible to trace the entire Pareto fairness curve by adjusting the fairness constraint thresholds. We provide theoretical guarantees for our method and empirically evaluate its effectiveness.
Read the full paper here.
Enhancing Controlled Query Evaluation with Epistemic Policies
Gianluca Cima, Domenico Lembo, Lorenzo Marconi, Riccardo Rosati, and Domenico Fabio Savo
Abstract: In this paper, we propose the use of epistemic dependencies to express data protection policies in controlled query evaluation (CQE), a form of privacy-preserving query answering over ontologies and databases. The resulting policy language goes far beyond those proposed so far in the CQE literature, enabling very rich and practically interesting forms of data protection rules. We demonstrate the expressive capabilities of our framework and study the data complexity of CQE for (unions of) conjunctive queries when ontologies are specified in the Description Logic DL-LiteR. Interestingly, while we show that the problem is generally undecidable, we prove tractability in the case of acyclic epistemic dependencies by providing a suitable query rewriting algorithm. This latter result paves the way for the implementation and practical application of this new approach to CQE.
Read the full paper here.
Online Learning of Capacity-Based Preference Models
Margot Hérin, Patrice Perny, and Nataliya Sokolovska
Abstract: In multi-criteria decision making, sophisticated decision models often involve a non-additive set function (called capacity) to define the weights of all subsets of criteria. This allows for modeling interactions between criteria, accommodating a variety of attitudes in criteria aggregation. Adapting a capacity-based decision model to a given decision-maker is a challenging problem, and several batch learning methods have been proposed in the literature to derive the capacity from a database of preference examples. In this paper, we introduce an online algorithm for learning a sparse representation of the capacity, designed for decision contexts where preference examples become available sequentially. Our method, based on a regularized double averaging approach, is also well-suited for decision contexts involving a large number of preference examples or a large number of criteria. Additionally, we propose a variant that allows for the inclusion of normative constraints on the capacity (e.g., monotonicity, supermodularity) while preserving scalability, based on the alternating direction method of multipliers.
Read the full paper here.
Lucy Smith, Editor-in-Chief of AIhub.
