H.B. Keller Colloquium
The ability to learn from data and make decisions in real-time has led to the rapid deployment of machine learning algorithms across many aspects of everyday life. While this has enabled new services and technologies, the fact that algorithms are increasingly interacting with people and other algorithms marks a distinct shift away from the traditional machine learning paradigm. Indeed, little is known about how these algorithms--- that were designed to operate in isolation--- behave when confronted with strategic behaviors on the part of people, and the extent to which strategic agents can game the algorithms to achieve better outcomes.
In this talk, I will give an overview of my work on learning in the presence of strategic agents. In the first part of the talk, I will discuss the necessary building blocks for optimal, decentralized, and coordination-free algorithms for learning while competing in matching markets. In the second, I will present a new dynamic model of strategic classification that allows for more realistic interactions between the strategic agents and the decision-maker. This model allows us to make insights into what features of algorithmic decision-making determine the order-of-play in the underlying Stackelberg game--- leading to simple procedures that decision-makers can use to find better equilibria for everyone in the game.