?
Подходы к распознаванию эмоций в интеллектуальных системах
The article examines approaches to emotion recognition in intelligent systems from the point of view of methodological grounds. In interdisciplinary research, where it is necessary to combine approaches from different fields, we face terminological uncertainty, since the problem of describing the emotional sphere is solved by different researchers in line with several main approaches.
First, the description of recognized and produced emotions can be conducted through the selection of fundamental (basic) emotions, the combination of which generates a variety of emotional phenomena. According to the discrete (categorical) approach, all emotional components are equal. Secondly, the emotional sphere can be described by highlighting the dimensions by which emotions are defined. According to the dimensional approach, emotions are located in n-dimensional space and can be characterized using a certain set of characteristics. Third, we can distinguish the component approach, which refers to the cognitive assessment of situations and events, the result of which leads to a pattern of emotional reactions. Fourth, a meaning-oriented approach that compares semantic and psychological structures and categories. Fifth, the constructivist approach, which involves the сonstitution of emotions depending on behavioral or value models. Sixth, an adaptive approach that defines emotions as ways to automatically respond to significant events.
Another classification of approaches to emotion recognition identifies four types of emotion theories: biological, cognitive, activity-based, and formal. The article offers a brief description of cognitive theories that explore the mechanisms of communication between emotions and cognitive processes. Fixing theoretical approaches is necessary to clarify the methodological basis for experiments related to the recognition of emotions by an intelligent agent in intelligent systems.
We consider the case when an intelligent system consists of two components: an intelligent agent (robot or algorithm) and a human. The actions related to emotion recognition that are performed by intelligent agents are listed: inducing and recognizing emotions, as well as creating the design of the desired emotion – a specific emotional contour.