Uncertainty-aware QD framework optimizes 24-hour time-use for health

Daily time allocation—how we split 24 hours between sleep, work, exercise, and leisure—is strongly linked to physical, mental, and cognitive health. But most optimization models focus purely on maximizing expected benefits, ignoring the uncertainty inherent in data-driven predictions. This can produce unrealistic health recommendations. A new paper from Aneta Neumann, Ty Stanford, Dorothea Dumuid, and Frank Neumann introduces a Quality Diversity (QD) framework that explicitly incorporates predictive uncertainty into the optimization process. Using compositional data analysis on a child cohort with over 1,000 participants, the model learns relationships between time-use compositions and health indicators such as body mass index, life satisfaction, and cognition. The framework explores the solution space through both variable-based and objective-based behavioral representations, uncovering diverse, high-quality time-use schedules while quantifying trade-offs between health outcomes under uncertainty.

By embedding uncertainty directly into QD, the approach shifts recommendations toward regions where the model is more confident—while preserving high-quality structures in the solution space. This yields more reliable time-use suggestions that balance expected health gains with prediction robustness. The work, accepted at the Parallel Problem Solving from Nature (PPSN) 2026 conference, is particularly relevant for personalized health coaching, wearable-device recommendations, and public health policy where real-world decisions depend on trustworthy guidance. Unlike previous methods that return a single optimal schedule, this framework generates a diverse set of near-optimal options, allowing users or clinicians to choose based on risk tolerance and lifestyle constraints. The researchers' integration of uncertainty quantification into evolutionary QD techniques marks a practical step toward data-driven behavioral health recommendations that users can actually trust.