Cognitive Interaction

The term "interaction" is used in a variety of contexts, but as a methodology in cognitive science, it refers to "a fundamental theory for designing artifacts that can interact with people naturally and continuously, depending on the situation," which has not been done by conventional cognitive science or information science, including AI. (Cognitive Interaction. By Kazuhiro Ueda, Yoshimasa Omoto, and Yugo Takeuchi, (Ohmsha, Inc., 2022)

People basically have the ability to share intentions and preferences with others and accomplish cooperative tasks, which is said to be the difference between people and chimpanzees. (p.30) Similarly, AI will be able to interact more smoothly if it can estimate people's intentions, preferences, and beliefs, and show behavior in line with their interests, goals, and other orientations.

To this end, the book describes a methodology aimed at realizing how to make AI successfully estimate people's intentions and how to make it respond in accordance with people's objectives.

What struck me when I read the book was that it is very important to understand (p.2) what kind of models people have of others (i.e., how they learn) and how they update those models. That is difficult to be aware of. So we need to deepen our understanding of methodology there.

The book has annotations on each page with clear definitions of technical terms such as ethnography, agents, the difference between cognitive computing and cognitive interaction, and so on.

In the latter half of the book, analytical methods (time series analysis, machine learning, etc.) are systematized, and if I thought it sounded like a textbook, it was.