Beyond World Models: Rethinking Understanding in AI Models

ACL ARR 2025 May Submission5759 Authors

Abstract: The AI community has shown substantial interest in the concept of world models: internal representations that simulate aspects of the external world, track entities and states, capture causal relationships, and enable prediction of consequences. This contrasts with representations based solely on statistical correlations. A key motivation behind this research direction is the argument that humans possess such mental world models, and finding evidence of similar representations in AI models might indicate that these models truly "understand" the world in a human-like way. In this position paper, we argue that human-level understanding extends beyond mental world models alone. We examine illuminating cases from prior philosophical work on understanding—including analyses of computational systems, physical theories, and mathematical proofs—to demonstrate how human understanding goes beyond just mental world models. By highlighting these distinctions, we hope to stimulate deeper discussion about what constitutes true understanding in both human and artificial contexts.

Paper Type: Short

Research Area: Linguistic theories, Cognitive Modeling and Psycholinguistics

Research Area Keywords: cognitive modeling

Contribution Types: Position papers

Languages Studied: English

Submission Number: 5759