The abstract is read at the publisher; the summary is JClub's.
Cognitive grounding shifts seven LLMs toward human-calibrated behavior in 231 Beer Distribution Game sessions, improving trajectory fit while changing inferred decision parameters.
- Why it matters: LLMs are used as human stand-ins, but plausible outputs may not reflect human decision processes, especially in interacting settings where local choices compound into collective outcomes.
- What they did: The study tests whether qualitative behavioral-model content alone—cognitive grounding—can drive LLM agents, comparing grounded versus ungrounded agents across 231 supply-chain sessions.
- The result: Grounding improves trajectory fit for every model and generally moves inferred decision parameters toward human-calibrated references, with uneven effects by role and model, and it holds under alternative prompting and memory designs.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.