Asking Easy Questions: A User-Friendly Approach to Active Reward Learning

TL;DR

This paper introduces a method for active reward learning in robots that prioritizes asking questions easy for humans to answer, improving learning efficiency by balancing robot and human uncertainties.

Contribution

It proposes an information gain approach that considers human answerability, enhancing question selection for faster reward learning in human-robot interaction.

Findings

Questions are easier for humans to answer.

Faster reward learning achieved in simulations.

User study confirms improved question quality.

Abstract

Robots can learn the right reward function by querying a human expert. Existing approaches attempt to choose questions where the robot is most uncertain about the human's response; however, they do not consider how easy it will be for the human to answer! In this paper we explore an information gain formulation for optimally selecting questions that naturally account for the human's ability to answer. Our approach identifies questions that optimize the trade-off between robot and human uncertainty, and determines when these questions become redundant or costly. Simulations and a user study show our method not only produces easy questions, but also ultimately results in faster reward learning.