Item

Abstract

Humans continuously categorise inputs but only rarely receive explicit feedback as to whether or not they are correct. This implies that they may be engaged in semi-supervised learning, which integrates supervised and unsupervised adaption. However, experiments testing semi-supervised learning in humans are sparse, and bedevilled with conflicting results about the benefits of unsupervised information. Here, we suggest that one important factor that has been paid insufficient attention is the alignment between subjects' internal representations of the stimulus material, which is shaped by prior biases and experience, and the experimenter-defined representations that determine success in the tasks. Only if these representations match will unsupervised learning be successful. To test this hypothesis, we designed a series of behavioural categorisation experiments in which subjects initially categorise items along a salient, but task-irrelevant, dimension, and only recover the correct categories when sufficient feedback draws their attention to the subtle, task-relevant, stimulus dimensions. Withdrawing feedback at different stages of this learning curve tests whether unsupervised learning can improve performance when psychological stimulus space and task are sufficiently aligned, and whether the opposite is true if they are misaligned. Our predictions fit with work in perceptual and language learning where task proficiency has been reported to be a crucial predictor of the effects of further unsupervised learning. Our work implies that predicting and understanding human category learning in particular tasks requires assessment and consideration of the representational spaces that subjects entertain for the materials involved in those tasks. These considerations not only apply to studies in the lab, but could also help improve the design of tutoring systems and instruction.