What Is Worth Representing? Representational Empowerment for Continual Model Construction

The first problem of modeling the world is not just estimating the right parameters or causal structure, but deciding *what should be represented at all*. We frame this as *continual model construction*: an agent maintains an environment-specific model $M$ of an inaccessible world $W$ and curates a persistent library $\mathcal{L}$ of reusable representational elements across environments. We propose *Representational Empowerment* ($\mathrm{RepEmp}$) to score candidate elements by how much they expand the agent's future capacity to model and plan—a counterpart to classical environmental empowerment, redirected from control over external states to control over internal representations. We realize the framework as a hierarchical Actor-Curator architecture and test it across three experiments. In a finite-vocabulary causal-learning task, human participants construct causal models at varying abstraction levels to maximize goal reachability rather than fidelity to the world—a signature better predicted by $\mathrm{RepEmp}$ over information-gain or novelty alternatives. Matched simulations reveal $\mathrm{RepEmp}$-guided construction, not exploration, to be the driver of sufficient structure recovery and cross-task transfer. Finally, in an open-vocabulary planning domain, an LLM-augmented Curator builds more compact symbolic libraries, which also generalize better than baselines. Ablating $\mathrm{RepEmp}$ eliminates these benefits. Together, these results identify $\mathrm{RepEmp}$ as a potential principle for continual model construction: deciding what to build, retain, and reuse under bounded resources.