2017 年 36 巻 1 号 p. 163-172
An ideal observer is theoretical device that performs a given task in an optimal manner provided the available information and some specified constraints. Comparing the performance of the ideal observer to that of a test observer in the given task, one can infer characteristics and/or deficit in a system of the test observer. Ideal observer theory has been applied to a wide range of problems, such as perception, object recognition, category learning, memory, attention, decision-making, and others. Recent application of Bayesian statistical theory enables us to investigate perceptual processes in more naturalistic and complicated scene and phenomena and to explore optical learning processes in many areas.
Here I first summarize the basic concepts and logic of ideal observer analysis and then briefly describe an application to a simple perceptual task.