Santa Fe Institute Working Group: Economic Agent-Based Models: Crossing Over the Tipping Point

Summary

Month-long working group focusing on Agent-based Modelling (ABM).

Abstract

This talk explores the integration of Multi-Agent Reinforcement Learning (MARL) into economic agent-based models. Using the “AI Economist” framework as a case study for taxation policy, I introduce the PEARL algorithm—a method for recovering underlying agent utility functions from observed data using input-concave neural networks. By moving beyond fixed behavioural assumptions, these techniques enable more realistic modelling of co-adapting agents and social planners in complex economic environments.

Marta Grześkiewicz

/ˈmɑːtə ɡrɛskɛvɪtʃ/\

I am a researcher at the intersection of economics and machine learning / AI.