Controllability in preference-conditioned multi-objective reinforcement learning

TL;DR

This paper introduces a new metric for assessing controllability in preference-conditioned multi-objective reinforcement learning, addressing limitations of existing evaluation methods.

Contribution

It proposes a specific controllability metric for preference-conditioned agents, enhancing evaluation beyond standard MORL metrics.

Findings

Standard MORL metrics do not reliably measure controllability.

A new metric is proposed to evaluate how well agents respond to preference changes.

Results aim to improve evaluation protocols for complex MORL problems.

Abstract

Multi-objective reinforcement learning (MORL) allows a user to express preference over outcomes in terms of the relative importance of the objectives, but standard metrics cannot capture whether changes in preference reliably change the agent's behavior in the intended way, a property termed controllability. As a result, preference-conditioned agents can score well on standard MORL metrics while being insensitive to the preference input. If the ability to control agents cannot be reliably assessed, the symbolic interface that MORL provides between user intent and agent behavior is broken. Mainstream MORL metrics alone fail to measure the controllability of preference-conditioned agents, motivating a complementary metric specifically designed to that end. We hope the results spur discussion in the community on existing evaluation protocols to consolidate advances in preference adaptation in MORL to larger and more complex problems.