When individuals make inferences based on multiple probabilistic cues, they differ in the extent to which they rely primarily on the most valid cue or integrate cues in a compensatory manner. Traditional accounts attribute non-compensatory decisions to prematurely terminated evidence accumulation or to the use of heuristics that ignore information. However, these accounts are challenged by recent findings on fast, parallel, and bidirectional integration of information as formalized in Parallel Constraint Satisfaction (PCS) models. This project evaluates the PCS sensitivity parameter P, which captures the degree of non-compensatory decision-making by describing how strongly highly valid cues are overweighted relative to less valid cues when objective validities are transformed into subjective weights. A parameter recovery simulation shows that P can be reliably estimated from choices, response times, and confidence ratings given sufficient sample size and diagnostic tasks. Reanalyses of three datasets including 688 participants and 94,792 decisions reveal that participants classified as using non-compensatory heuristics show higher fitted P values than participants classified as compensatory decision makers, indicating stronger overweighting of highly valid cues. A validation study tests whether experimentally induced changes in compensatory versus non-compensatory cue weighting are reflected in P. Overall, the results suggest that P can capture individual differences in compensatory and non-compensatory cue use as continuous variations in subjective cue weighting, consistent with evidence for fast, parallel processing.