(A) study on the effect of one-shot learning on inductive and deductive Inference

Abstract

Information between one event and another, or a stimulus and an outcome, can be acquired by learning their causal relationship. There are at least two distinct learning strategies for identifying a causal relationship : incremental learning that occurs from repeated trial and error and one-shot learning that occurs from just one pairing. Despite researches on biological studies such as conditioned taste aversion or one-shot learning studies based on relationship between ‘uncertainty and learning rate’ or‘prediction error and learning rate’, little has been known about the effect of one-shot learning on real-world scenarios. In this study, we designed a new task using a real world scenario to induce inductive and deductive. To further investigate the effect of one-shot learning on inductive and deductive reasoning, we carefully modulated uncertainty. When analyzing score and confidence according to factors of incremental/one-shot, inductive/deductive, and learning/inference, we found that the group that induced one-shot learning developed an inference rule better than a control group. Subjects who reported rule development achieved high scores with high confidence in the inference test, and also achieved high scores in the one-shot inference test.