This is an EC 2025 tutorial that will take place virtually on Tuesday, June 17, 2025, 2 pm – 4 pm ET. Registration is free but required. Please use this form for registration.
This tutorial aims to cover both classical foundations and recent advances on calibrated forecasting and decision making. Calibration, which ensures that predicted probabilities align with long-run empirical frequencies, has emerged as a key evaluation metric for forecasting. We will explore two modern perspectives on forecasting: one viewing calibration as a condition that forecasts should satisfy in order to provide a reliable information for downstream decision-making, and another viewing calibration as a behavioral assumption for agents in games.
Part 1 (45 mins): Forecasting for decision-making -- guarantees and algorithms
Part 2 (45 mins): Calibration in multi-agent decision-making
We will upload the slides after the tutorial.