B-EUR Model: Computational Framework for Design Option Value

A novel computational framework called Bayesian Expected Uncertainty Reduction (B-EUR) has been introduced to quantify the benefits of evaluating a candidate design action based on its anticipated decrease in epistemic uncertainty regarding action-outcome relationships. This model tackles a lingering issue in the Uncertainty Driven Action (UDA) framework about how variations in uncertainty perception influence action choices. The investigation focuses on two environmental aspects: generalizability, which assesses the extent to which insights from one trial apply to similar candidates, and outcome discriminability, which evaluates the clarity of distinguishing between different outcomes. Testing involved simulations and human trials using a graph-shape guessing task designed to isolate learning under a constrained trial budget. Results indicated that epistemic value exhibited an inverted-U relationship with generalizability and rose with greater outcome discriminability. The research is accessible on arXiv with the identifier 2608.05642, categorized as a 'new' announcement. This study enhances the understanding of decision-making in design scenarios, with implications for artificial intelligence and human-computer interaction.