ICM 2026 Plenary Lecture - Éva Tardos

ICM 2026 Plenary Lecture - Éva Tardos

by Éva Tardos

Formal & Physical Sciences Mathematics PBUOptimizationPBUDGame theory
🎙 Éva Tardos 👥 58K 📅 August 17, 2026 ⏱ 51 min 👁 3 📄 expert opinion 🧭 2026-08-17
Available in: English (current) Français

Keywords

price of anarchyno-regret learningNash equilibriumnetwork routingalgorithmic game theory

Summary

Éva Tardos delivers a plenary lecture at ICM 2026 on the intersection of learning, AI, and game theory. She begins by introducing the concept of the price of anarchy, which measures the loss of efficiency when self-interested agents optimize their own outcomes, using the classic Braess’s paradox as an illustrative example. She then discusses the limitations of Nash equilibrium as a solution concept, particularly the computational hardness of finding it and the informational requirements for players. The lecture shifts focus to repeated games and no-regret learning, where players use simple algorithms to adapt their strategies over time. Tardos highlights that no-regret learning provides a behavioral assumption that can be used to extend price of anarchy results to dynamic settings. However, she notes that these results rely on the assumption that the game is static, which is often violated in real-world systems like network routing where the state evolves. She presents a simple model of packet routing with queues and discusses how no-regret learning may not be sufficient in such dynamic games, and explores the question of how much extra capacity is needed to mitigate inefficiency. The talk concludes by suggesting that more sophisticated learning models are needed for games with evolving states.

202 words

Critical Evaluation

Value of the Information & Strength of the Argument

The lecture provides a high-level overview of key concepts in algorithmic game theory, with a clear narrative that connects classical results to modern challenges. Tardos effectively uses examples like Braess’s paradox and network routing to illustrate abstract ideas. The argumentation is solid, building from the definition of price of anarchy to the extension to no-regret learning, and then to the limitations in dynamic games. She presents the material in an accessible yet rigorous manner, suitable for a mathematical audience. The talk is valuable for its synthesis of existing results and its motivation for future research directions.

105 words

Title / Content Match

The title accurately reflects the content: a plenary lecture by Éva Tardos at ICM 2026.

Quality & Reliability

8/10

Lecture by a leading researcher in algorithmic game theory, presenting established results and recent work with mathematical rigor. The content is well-structured and based on peer-reviewed research, though the format is an expert presentation rather than a peer-reviewed publication.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • The Price of Anarchy in Games — Introduced by Koutsoupias and Papadimitriou, central to the talk.
  • Braess's Paradox — Illustrates the counterintuitive effect of adding capacity.
  • No-Regret Learning — Key concept for the learning model discussed.
  • Nash Equilibrium — Classical solution concept in game theory.
  • Roughgarden's Paper on Price of Anarchy and Learning — Summarizes the extension of price of anarchy to learning outcomes.

Concurring Sources

  • Roughgarden's Paper on Price of Anarchy and Learning — Supports the extension of price of anarchy to learning outcomes.

Contribution & Novelties

The lecture synthesizes existing results and highlights open questions in the intersection of learning and game theory. It emphasizes the need to move beyond static games and consider dynamic settings where the state evolves. The talk provides a clear motivation for future research on learning in dynamic games.

Pour aller plus loin :

79 words

Radar Profile

The radar profile shows high scores in quality of information and technical level, reflecting the expert presentation. The quantity of information is also high, but the global reliability is slightly lower due to the format (lecture) and lack of detailed citations. The overall balance indicates a strong, informative talk.

Reliability 8/10