Claim

Subjective mental effort may track how much an internal model must be revised, rather than how much raw information is processed.

Reasoning

People often find predictable but information-rich streams easier to handle than short bursts of anomalous input. This suggests that effort is not driven primarily by volume, but by the computational cost of updating expectations, attention policies, and action plans when the world stops behaving as forecast.

Test

Create matched tasks where total information volume is held constant, but one condition preserves a stable generative pattern while the other forces repeated belief updates. Measure subjective effort, pupillary response, and control-related slowing.

Implication

If supported, this would strengthen predictive-processing accounts that treat cognition as model maintenance under uncertainty, not passive storage under load.