Explaining by Imitating: Understanding Decisions by Interpretable Policy Learning

TL;DR

This paper introduces Interpole, a Bayesian method for learning interpretable decision policies from offline data, enabling transparent understanding of human decision-making in complex, real-world scenarios like healthcare.

Contribution

The paper presents a novel Bayesian approach that jointly models belief updates and decision policies, addressing challenges of partial observability and offline learning for interpretability.

Findings

Effective in simulated environments

Successfully applied to Alzheimer's diagnosis data

Provides insights into human decision-making processes

Abstract

Understanding human behavior from observed data is critical for transparency and accountability in decision-making. Consider real-world settings such as healthcare, in which modeling a decision-maker's policy is challenging -- with no access to underlying states, no knowledge of environment dynamics, and no allowance for live experimentation. We desire learning a data-driven representation of decision-making behavior that (1) inheres transparency by design, (2) accommodates partial observability, and (3) operates completely offline. To satisfy these key criteria, we propose a novel model-based Bayesian method for interpretable policy learning ("Interpole") that jointly estimates an agent's (possibly biased) belief-update process together with their (possibly suboptimal) belief-action mapping. Through experiments on both simulated and real-world data for the problem of Alzheimer's disease diagnosis, we illustrate the potential of our approach as an investigative device for auditing, quantifying, and understanding human decision-making behavior.