🤖 AI Summary
Why do individuals reach divergent conclusions despite identical observations? This work formalizes cognitive disagreement as a problem of structural non-identifiability in world model learning, distinguishing between two types: θ-level (reasoning profile) and W-level (world model) non-identifiability. It introduces a reasoning profile θ characterized by reference frames, exploration strategies, stability criteria, and temporal horizons to elucidate the mechanisms underlying such divergence. By integrating formal reasoning, hierarchical representation learning, and latent state estimation, the study demonstrates that disagreements arise from the projection of computational, observational, and coordination constraints onto abstract bases. This framework not only accounts for polarization in real-world debates—such as those surrounding AI regulation—but also provides a unified theoretical foundation for understanding cognitive differences between humans and artificial agents.
📝 Abstract
When people share the same documents and observations yet reach different conclusions, the disagreement often shifts into a judgment that the other party is cognitively defective, irrational, or acting in bad faith. This paper argues that such divergence is better described as a form of non-identifiability inherent in inference and learning, rather than as a defect of the other party. We organize the phenomenon into two levels: (i) $θ$-level non-identifiability, where conclusions diverge under the same world model $W$ because inference settings differ; and (ii) $W$-level non-identifiability, where repeated use of an inference setting $θ$ biases data exposure and update rules, causing the learned world model $W$ itself to diverge. We introduce an inference profile $θ= (R, E, S, D)$, consisting of Reference, Exploration, Stabilization, and Horizon, and show how outputs can split even for the same observation $o$ and the same $W$. We further explain why disagreements tend to project onto a small number of bases -- abstract versus concrete, externalizability, and order versus freedom -- as a consequence of general constraints on learning systems: computational, observational, and coordination constraints. Finally, we relate the framework to deep representation learning, including representation hierarchy, latent-state estimation, and regularization-exploration trade-offs, and illustrate the framework through a case study on AI regulation debates.