In-context learning agents are asymmetric belief updaters
This research addresses how problem framing influences learning in AI agents, offering insights into cognitive parallels with humans, but it is incremental as it builds on existing in-context learning studies.
The study investigated in-context learning in large language models using cognitive psychology tasks, finding that they update beliefs asymmetrically, learning more from better-than-expected outcomes than worse ones, with effects reversing for counterfactual feedback and disappearing without agency.
We study the in-context learning dynamics of large language models (LLMs) using three instrumental learning tasks adapted from cognitive psychology. We find that LLMs update their beliefs in an asymmetric manner and learn more from better-than-expected outcomes than from worse-than-expected ones. Furthermore, we show that this effect reverses when learning about counterfactual feedback and disappears when no agency is implied. We corroborate these findings by investigating idealized in-context learning agents derived through meta-reinforcement learning, where we observe similar patterns. Taken together, our results contribute to our understanding of how in-context learning works by highlighting that the framing of a problem significantly influences how learning occurs, a phenomenon also observed in human cognition.