Optimal Teaching for Limited-Capacity Human Learners

摘要: Basic decisions, such as judging a person friend or foe, involve categorizing novel stimuli. Recent work finds that people's category judgments are guided by small set of examples retrieved from memory at decision time. This limited and stochastic retrieval places limits on human performance for probabilistic classification decisions. In light this capacity limitation, recent idealizing training items, the saliency ambiguous cases is reduced, improves test items. One shortcoming previous in idealization distributions were idealized an ad hoc heuristic fashion. contribution, we take first principles approach to constructing sets. We apply machine teaching procedure cognitive model either (as humans are) unlimited most learning systems are). As predicted, find teacher recommends also learners perform best when recommendations based limited-capacity model. extent used conforms true nature learners, prove effective. Our results provide normative basis (given constraints) procedures offer selection models learning.