ASQ-IT: Interactive Explanations for Reinforcement-Learning Agents

TL;DR

ASQ-IT is an interactive tool that enables users to understand reinforcement learning agents through video-based explanations driven by formal logic queries, fostering active user engagement and improved fault detection.

Contribution

This paper introduces ASQ-IT, a novel interactive explanation system for reinforcement learning agents that uses formal logic-based queries to facilitate user understanding and diagnosis.

Findings

Users can effectively formulate queries in ASQ-IT.

ASQ-IT helps users identify faulty agent behaviors.

User studies validate the system's usability and effectiveness.

Abstract

As reinforcement learning methods increasingly amass accomplishments, the need for comprehending their solutions becomes more crucial. Most explainable reinforcement learning (XRL) methods generate a static explanation depicting their developers' intuition of what should be explained and how. In contrast, literature from the social sciences proposes that meaningful explanations are structured as a dialog between the explainer and the explainee, suggesting a more active role for the user and her communication with the agent. In this paper, we present ASQ-IT -- an interactive tool that presents video clips of the agent acting in its environment based on queries given by the user that describe temporal properties of behaviors of interest. Our approach is based on formal methods: queries in ASQ-IT's user interface map to a fragment of Linear Temporal Logic over finite traces (LTLf), which we developed, and our algorithm for query processing is based on automata theory. User studies show that end-users can understand and formulate queries in ASQ-IT, and that using ASQ-IT assists users in identifying faulty agent behaviors.