What happens when an AI system faithfully follows a flawed goal? In this episode, we explore how even well-designed algorithms can produce dangerous outcomes — from amplifying hate speech to mismanaging infrastructure — simply by optimising a reward function which, like all reward functions, fails to encode all that matters. We discuss the hidden risks of reinforcement learning, why over-optimisation can backfire, and how game theory helps us rethink what it means for AI to act "rationally" in complex, real-world environments.

Jobst Heitzig is a mathematician at the Potsdam Institute for Climate Impact Research and an expert in AI safety and decision design.

In this episode, we discuss how game theory sheds light on modern electoral competition. Our guest, Adam Meirowitz, introduces the classical model of candidate convergence and explains why real elections often depart from this prediction. We then explore the role of echo chambers and selective exposure, and why these surprisingly do not necessarily lead to greater polarisation. We also touch on how bounded rationality among voters can further impact these dynamics.

Adam Meirowitz is the Damon Wells Professor of Political Science with a courtesy appointment in Economics at Yale University. His referenced research paper on echo chambers "Selective Exposure and Electoral Competition" is published in the Journal of Politics (2025) and is joint work with Avidit Acharya, Peter Buisseret, and Floyd Zhang. You can find it here.

Talk is cheap – literally. But if words cost nothing, why do we ever trust them? In this episode, Yale Professor Navin Kartik breaks down the game theory of communication.

We kick things off with the basics: what is the academic definition of "cheap talk", and why does cheap talk sometimes work but other times break down? From there, we transition to Navin's own research, which explores deeper facets of communication. We learn about lying costs and "inflated language": why does a reference letter need to say "excellent" to just mean "competent"? Finally, we dive into topics like pandering (why experts lie to agree with you) and muddled information (why test scores tell you more about gaming skills than natural ability).