Victoria Jacoby¹, Christine Massey¹, Philip Kellman¹; ¹University of California, Los Angeles

Processes of perceptual learning can be accelerated through direct comparison of items. Previous work showed that adaptive learning methods can enhance the benefits of comparison by creating targeted contrasts between confused categories (Jacoby, Massey, & Kellman, 2024). Here, we tested whether adaptive comparisons could improve learning in the difficult task of skin lesion differentiation. Undergraduate participants were assigned to learn 10 categories of skin lesions in one of three conditions. In the Single-Item Condition, participants classified individual exemplars sequentially using a standard category learning format. The ARTS system (Mettler, Massey & Kellman, 2016), an adaptive interleaving system, determined category spacing based on learner performance. In the Adaptively Triggered Comparisons (ATC) Condition, participants completed the single-item trials with their responses monitored for confusions. Repeated misclassifications between categories triggered a comparison trial, in which two exemplars from the confused categories were presented simultaneously for discrimination. Finally, the Non-Adaptive Comparison (NAC) Condition included a similar number of comparison trials as the ATC Condition, but these occurred at fixed intervals with randomly selected categories. Participants completed trials until they met mastery criteria based on accuracy and response times, followed by immediate and one-week delayed posttests. Results: Average time to mastery was lowest in the ATC condition, followed by the Single-Item and NAC conditions. Posttest accuracy and learning efficiency (calculated as posttest accuracy divided by time invested) was numerically highest in the ATC condition, followed by the Single-Item and NAC conditions. An ANOVA revealed significant differences for time to mastery and posttest accuracy, with marginal effects for efficiency. Follow-up contrasts indicated that the ATC Condition significantly outperformed the NAC Condition across all measures and reached mastery reliably faster than the Single-Item Condition. These results provide initial evidence that adaptive comparisons can enhance the perceptual learning of skin lesion classifications, demonstrated primarily through quicker learning rates.

Acknowledgements: We gratefully acknowledge support for this research from National Institutes of Health grant R01CA236791 to PK and CM.