Feedback Mechanism Based On Family Resemble Material In Category Learning

Category Learning is a process that learns to how to classify the stimulus through exercises. It is a essential part of many cognitive activities, such as inference, problem solving. We can divide category learning into supervised category learning and unsupervised category learning. Most laboratory research has focused on supervised category learning that is acquisition of a pre-determined category structure through feedback or instruction. In this regard, there are many theories and experimental paradigms have been developed. Although unsupervised category learning that we can learn how to categorize objects in the absence of "right" or "wrong" feedback is very common, it has attracted relatively little attention. Laboratory research on category learning main focused on materials, ways of category learning et al. However, feedback that is an important process of categorization has attracted little attention.Feedback is an important intermediary between subject and environment, and we can not adapt to the environment without feedback. There are two views on feedback mechanism. Reinforcement learning takes feedback as reinforcement, the views consider that the right response can increase the probability of action and the wrong response will decrease the probability of action recurrence. The view of Error-Driven highlights the role of wrong response, it held that the wrong response make learner find right answer. Making mistake provides more information that can tell learner which feature is irrelevant. From our points of view, the two mechanisms are simultaneous, but act on different systems. Reinforcement learning plays larger role for memory system, and Error-Driven for inference system. Based on these ideas, we study feedback mechanisms in category learning with family resemblance material.The paper included 3 experiments. We compared the difference between observational learning and feedback learning in experiment 1. In experiment 2, we operated timing of feedback, compared delayed feedback with immediate feedback. We study how the nature of feedback affects category learning. Draw conclusions as follows:1 Feedback learning is more efficient than observational learning. Because feedback make participants do deeper processing. Observational learning only uses memory strategies, but in feedback learning participants also make inference from errors.2 The differences between delayed and immediately feedback are not significant. Neither supports the reinforcement learning nor Error-Driven.3 Negative feedback are superior to positive feedback. Participants in negative feedback take deeper processing.4 Reinforcement learning and Error-Driven are simultaneous, they act on different systems. Reinforcement learning plays an important role in memory system and Error-Driven in inference system.