Recent research indicates that paired comparisons can accelerate perceptual learning of challenging dermatological lesion categories. Here we investigated whether the role of object categories as targets or distractors differentially influences learning outcomes. The frequency with which a given category occupied the target position was manipulated across three learning conditions: Always-Never, where half of 10 categories were always shown as target and the other half never shown as target; Often-Rarely, where half of categories appeared 75% as targets and 25% as distractors, with reversed presentation frequency for the other half; and Equal Split learning, in which all categories appeared as targets or distractors equally often. After learning, transfer results indicated that all conditions yielded equivalent overall learning, but categories prioritized more often as targets exhibited greater learning gains. These findings implicate differential processing of images in comparisons, even when no information regarding target vs. distractor was given prior to feedback.

Combining perceptual learning techniques with adaptive learning algorithms has been shown to accelerate the development of expertise in medical and STEM learning domains (Kellman & Massey, 2013; Kellman, Jacoby, Massey & Krasne, 2022). Virtually all adaptive learning systems have relied on simple accuracy data that does not take into account response bias, a problem that may be especially consequential in multi-category perceptual classifications. We investigated whether adaptive perceptual learning in skin cancer screening can be enhanced by incorporating signal detection theory (SDT) methods that separate sensitivity from criterion. SDT-style concepts were used to alter sequencing, and separately to define mastery (category retirement). SDT retirement used a running d’ estimate calculated from a recent window of trials based on hit and false alarm rates. Undergraduate participants used a Skin Cancer PALM (perceptual adaptive learning module) to learn classification of 10 cancerous and readily-confused non-cancerous skin lesion types. Four adaptive conditions varied either the type of adaptive sequencing (standard vs. SDT) or retirement criteria (standard vs. SDT). A non-adaptive control condition presented didactic instruction on dermatologic screening in video form, including images, classification schemes, and detailed explanations. All adaptive conditions robustly outperformed the non-adaptive control in both learning efficiency and fluency (large effect sizes). Between adaptive conditions, SDT retirement criteria produced greater learning efficiency than standard, accuracy-based mastery criteria at both immediate and delayed posttests (medium effect sizes). SDT sequencing and standard adaptive sequencing did not differ. SDT enhancements to adaptive perceptual learning procedures have potential to enhance learning efficiency.

Research has shown that estimation of correlation from scatter plots is done poorly by both novices and experts. We tested whether proficiency in correlation estimation could be improved by perceptual learning interventions, in the form of perceptual-adaptive learning modules (PALMs). We also tested learning effects of alternative category structures in perceptual learning. We organized the same set of 252 scatter plot displays either into a PALM that implemented spacing in learning by shape categories or one in which the categories were ranges of correlation strength. Both PALMs produced markedly reduced errors, and both led trained participants to classify near transfer items as accurately as trained items. Differences in category organization produced modest effects on learning; there was some indication of more consistent reduction of absolute error when learning categories were organized by shape, whereas average bias of judgments was best reduced by categories organized by different numerical ranges of correlation.

Adaptive generation of spacing intervals in learning usingresponse times improves learning relative to both adaptivesystems that do not use response times and fixed spacingschemes (Mettler, Massey & Kellman, 2016). Studies haveoften used limited presentations (e.g., 4) of each learningitem. Does adaptive practice benefit learning if items arepresented until attainment of objective mastery criteria? Doesit matter if mastered items drop out of the active learning set? We compared adaptive and non-adaptive spacing underconditions of mastery and dropout. Experiment 1 comparedrandom presentation order with no dropout to adaptivespacing and mastery using the ARTS (AdaptiveResponse-time-based Sequencing) system. Adaptive spacingproduced better retention than random presentation.Experiment 2 showed clear learning advantages for adaptivespacing compared to random schedules that also includeddropout. Adaptive spacing performs better than randomschedules of practice, including when learning proceeds tomastery and items drop out when mastered.

Recent work suggests that learning perceptual classifications can be enhanced by combining single item classifications with adaptive comparisons triggered by each learner’s confusions. Here, we asked whether learning might work equally well using all comparison trials. In a face identification paradigm, we tested single item classifications, paired comparisons, and dual instance classifications that resembled comparisons but required two identification responses. In initial results, the comparisons condition showed evidence of greater efficiency (learning gain divided by trials or time invested). We suspected that this effect may have been driven by easier attainment of mastery criteria in the comparisons condition, and a negatively accelerated learning curve. To test this idea, we fit learning curves and found data consistent with the same underlying learning rate in all conditions. These results suggest that paired comparison trials may be as effective in driving learning of multiple perceptual classifications as more demanding single item classifications.

Spacing presentations of learning items across time improvesmemory relative to massed schedules of practice – thewell-known spacing effect. Spaced practice can be furtherenhanced by adaptively scheduling the presentation of learningitems to deliver customized spacing intervals for individualitems and learners. ARTS - Adaptive Response-time-basedSequencing (Mettler, Massey, & Kellman 2016) determinesspacing dynamically in relation to each learner’s ongoing speedand accuracy in interactive learning trials. We demonstrate theeffectiveness of ARTS when applied to chemistry nomenclaturein community college chemistry courses by comparing adaptiveschedules to fixed schedules consisting of continuouslyexpanding spacing intervals. Adaptive spacing enhanced theefficiency and durability of learning, with learning gainspersisting after a two-week delay and generalizing to astandardized assessment of chemistry knowledge after 2-3months. Two additional experiments confirmed and extendedthese results in both laboratory and community college settings.

Recent work suggests that optimal spacing in learningrequires adaptive procedures (Mettler, Massey & Kellman,2016). Here, we studied how adaptive techniques might befurther enhanced by combining active and passive learningmodes. Participants learned geography facts that werescheduled using the ARTS (Adaptive Reaction-Time-basedScheduling) system under four conditions involving passiveand/or active trials. Conditions included: a) Passive Onlypresentations of learning items, b) Passive Initial Blocksfollowed by active adaptive scheduling, c) Passive InitialItems followed by active adaptive scheduling for each itemintroduced, or d) Active Only learning with no passivepresentations. We found an advantage for combinations ofactive and passive presentation (by blocks or items) overPassive Only or Active Only presentation. Passive trialspresented in blocks at the beginning of learning showed bestperformance. We discuss possible explanations for thesedifferences and suggest principles underlying optimalcombinations of active and passive modes in adaptivelearning.

Adaptive learning systems that generate spacing intervalsbased on learner performance enhance learning efficiency andretention (Mettler, Massey & Kellman, 2016). Recentresearch in factual learning suggests that initial blocks ofpassive trials, where learners observe correct answers withoutovertly responding, produce greater learning than passive oractive trials alone (Mettler, Massey, Burke, Garrigan &Kellman, 2018). Here we tested whether this passive + activeadvantage generalizes beyond factual learning to perceptuallearning. Participants studied and classified images ofbutterfly genera using either: 1) Passive Only presentations,2) Passive Initial Blocks followed by active, adaptivescheduling, 3) Passive Initial Category Exemplar followed byactive, adaptive scheduling, or 4) Active Only learning. Wefound an advantage for combinations of active and passivepresentations over Passive Only or Active Only presentations.Passive trials presented in initial blocks showed the bestperformance, paralleling earlier findings in factual learning.Combining active and passive learning produces greaterlearning gains than either alone, and these effects occur fordiverse forms of learning, including perceptual learning.

Decades of research have demonstrated that students facecritical conceptual challenges in learning mathematics. Asnew adaptive learning technologies become ubiquitous ineducation, they bring opportunities both to facilitateconceptual development in more focused ways and to gatherdata that may yield new insights into students’ learningprocesses. The present study analyzes data archives from aperceptual learning intervention designed to help studentsmaster key concepts related to linear measurement andfractions. Using algorithmic data coding on a database of78,034 errors from a sample of sixth graders, both conceptualerrors and other errors were captured and analyzed for changeover time. Results indicate that conceptual errors decreasedsignificantly. This approach suggests additional ways thatsuch datasets can be exploited to better understand how thesoftware impacts different students and how next generationsof adaptive software may be designed to code and respond tocommon error patterns in real time.