It Is What It Isn’t: Introducing a Constraint-Based Approach To Structure Learning

Biological cognition depends on learning structured representations in ambiguous environments. Computational models of structure learning typically overlook the temporally extended dynamics that shape learning trajectories under such ambiguity. In this paper, we reframe structure learning as an emergent consequence of constraint-based dynamics. Informed by a literature on the role of constraints in complex biological systems, we build a framework for modelling constrain-based dynamics and provide a proof-of-concept computational cognitive model. The model consists of an ensemble of components, each comprising an individual learning process, whose internal updates are locally constrained by both external observations and system-level relational constraints. This is formalised using Bayesian probability as a description of constraint satisfaction. Representational structure is not encoded directly in the model equations but emerges over time through the interaction, stabilisation, and elimination of components under these constraints. Through a series of simulations in environments with varying degrees of ambiguity, we demonstrate that the model reliably differentiates the observation space into stable representational categories. We further analyse how global parameters controlling internal constraint and initial component precision shape learning trajectories and long-term behavioural alignment with the environment. The results suggest that constraint-based dynamics offer a viable and conceptually distinct foundation for modelling structure learning in adaptive systems. We further analyse how global parameters controlling internal constraint and initial component precision shape learning trajectories and long-term behavioural alignment with the environment. We show that this allows to capture structure learning even in cases where it is maladaptive, such as delusion-like belief updating.