Position: Human-like reinforcement learning ıs facilitated by structured and expressive goals
Abstract
Goals play a central role in the study of agentic behavior. But what is a goal, and how should we best represent them? The traditional reinforcement learning answer is that all goals are expressible as the maximization of future rewards. While parsimonious, such a definition seems insufficient when viewed from both the perspective of humans specifying goals to machines and autotelic agents that self-propose tasks to explore and learn. We offer a critical perspective on the distillation of all goals directly into reward functions. We identify key features we believe goal representations ought to have, and then discuss a proposal we believe meets those considerations. This position paper argues that to specify human-like problems and construct agents to pursue them, the basic notion of reward in RL is impoverished and instead should be augmented by structured and expressive goal representations.
Type
Add the full text or supplementary notes for the publication here using Markdown formatting.