Estimation and validation of a single model parameter governing compensatory and non-compensatory cue use in decision making

Research on multi-cue probabilistic inference distinguishes between compensatory and non-compensatory cue use. Traditional accounts infer from certain choice patterns that decision makers rely on non-compensatory cue integration, effectively ignoring less valid cues. In contrast, the Parallel Constraint Satisfaction (PCS) model attributes the same patterns not to the exclusion of cues, but to differences in their weighting within a parallel, bidirectional integration process. The present research tests this theoretical account by evaluating the PCS parameter P as a continuous measure of cue weighting. Using parameter-recovery simulations, we show that P can be estimated without bias from choices, decision times, and confidence, with improved precision under sufficient and diagnostic data. Reanalyses of two existing datasets reveal substantial overlap between predictions of heuristic models and those generated by PCS with varying P, indicating that apparently discrete strategies can be reconciled within a single continuous framework. In a preregistered experiment (N = 189), experimentally reinforcing compensatory versus non-compensatory cue use systematically shifted estimates of P, supporting its psychological validity and adaptivity. Across studies, variation in P was related to behavioral, process, and self-report measures. These findings suggest that apparent differences between compensatory and non-compensatory decisions reflect graded differences in cue weighting rather than qualitatively distinct strategies. We discuss the potential of PCS as a framework for integrating diverse cue use patterns within a single measure and outline boundary conditions of parallel, bidirectional cue integration.