Temporal Diversity as a design principle for within-person collective intelligence

Response times (RTs) are often treated as a scalar cue: faster responses tend to be more accurate, and recent work shows that RT can be exploited at the group level to select better answers from a crowd. However, within-person collective intelligence poses a different challenge—whether a single individual can generate meaningful diversity across repeated judgments. We argue that temporal variability in response times across repeated estimates (RT-diversity), rather than mean RT, indexes switches between inferential modes and thereby helps generate error-canceling diversity within an individual. We test this idea using two previously collected laboratory datasets: a five-estimate dataset and a three-estimate dataset. In the five-estimate dataset, individuals with higher RT-diversity showed better outcomes when their own estimates were aggregated, whereas mean RT was not predictive. Importantly, aggregation often worsened accuracy relative to the first estimate, and RT-diversity primarily predicted who avoided such deterioration. Extensive robustness analyses rule out alternative explanations. In the three-estimate dataset, analogous relationships were weak, revealing a boundary condition. We situate these findings within research on response times, rational metareasoning, diversity-driven collective intelligence, and cognitive boosts, and discuss implications for identifying and fostering within-person wisdom.