Supervised Reward Inference
This work addresses the challenge of reward inference for AI systems when human behavior does not follow standard models, potentially improving human-robot interaction.
The paper tackles the problem of inferring reward functions from diverse human behaviors, which may be suboptimal or communicative, by proposing a supervised learning framework that is asymptotically Bayes-optimal. Experiments on simulated robotic manipulation tasks demonstrate that the method efficiently infers rewards from arbitrarily suboptimal demonstrations.
Existing approaches to reward inference from behavior typically assume that humans provide demonstrations according to specific models of behavior. However, humans often indicate their goals through a wide range of behaviors, from actions that are suboptimal due to poor planning or execution to behaviors which are intended to communicate goals rather than achieve them. We propose that supervised learning offers a unified framework to infer reward functions from any class of behavior, and show that such an approach is asymptotically Bayes-optimal under mild assumptions. Experiments on simulated robotic manipulation tasks show that our method can efficiently infer rewards from a wide variety of arbitrarily suboptimal demonstrations.