Ecological Suboptimality in Naturalistic Foraging: Amplified Deviation from Optimality in a Mouse Model of Alzheimer Disease
Adaptive decision-making extends beyond selecting among discrete alternatives. It requires the dynamic organization of actions as costs, opportunities, and goals evolve over time. Foraging captures this complexity in an evolutionarily conserved behavior that integrates spatial, temporal, and reward-related information across successive actions. These demands make foraging an ecologically grounded framework for investigating Alzheimer disease, in which spatial cognition, temporal organization, cost evaluation, and behavioral flexibility are frequently disrupted. We examined foraging-related action selection in control C57BL/6J mice and APPNL-G-F knock-in mice using a controlled task in which travel distance, food texture, and pellet size altered the costs and benefits of available strategies. Bayesian multinomial modeling showed that control mice flexibly redistributed their behavior across conditions, whereas APPNL-G-F mice showed impaired cost integration, increased withdrawal, and reduced behavioral flexibility rather than a generalized performance impairment. Comparison with a classical reward-rate-maximization benchmark revealed substantial suboptimality in both groups. Control mice achieved 53% of the predicted optimal reward rate, whereas APPNL-G-F mice achieved 45%. These systematic departures suggest that behavior was shaped by cognitive, motivational, or contextual constraints not represented in the normative model, with Alzheimer-disease-relevant dysfunction associated with greater suboptimality. Group differences were greatest when optimal behavior was highly sensitive to competing costs. By preserving the distribution of behavior across strategies, this study establishes foraging-related action selection as a sensitive framework for detecting disease-associated disruptions that may be obscured by conventional measures of choice or overall task performance.