Machine learning algorithms are increasingly being deployed into environments in which they must interact with other strategic agents like algorithms and people with potentially misaligned objectives. While the presence of these strategic interactions creates new challenges for learning algorithms, they also give rise to new opportunities for algorithm design. In this talk I will discuss how ideas from economics can give us new insights into the analysis and design of machine learning algorithms for these real-world environments. 

In this talk I will focus on equilibrium computation in Markov games.  A significant roadblock to the development of principled multi-agent reinforcement learning algorithms is the fact that the underlying problem of equilibrium computation is computationally intractable. To overcome this obstacle, I will take inspiration from behavioral economics and show that -- by imbuing agents with important features of human decision-making like risk aversion and bounded rationality -- a class of risk-averse quantal response equilibria (RQE) become tractable to compute in all n-player matrix and finite-horizon Markov games. In particular, I will show that they emerge as the endpoint of no-regret learning in suitably adjusted versions of the games. Crucially, the class of computationally tractable RQE is independent of the underlying game structure and only depends on agents' degree of risk-aversion and bounded rationality. To validate the richness of this class of solution concepts I will show that it captures peoples' patterns of play in a number of 2-player matrix games previously studied in experimental economics and present a first analysis of the sample complexity of approximating these equilibria using multi-agent reinforcement learning.