Since the 1960s, many studies on computational models for emotions have been proposed in the research on artificial intelligence (Itoh et al., 1994; Tokosumi et al., 1994). They have been done based on the assumption that information processing systems for emotions are realized by cooperation between many subsystems specialized individually, as well as systems for reasoning, thinking, learning, and memory in human. Although their aims are different in individual cases, a basic one is to clarify mechanisms of emotions by implementing computational models that are assumed to be maintained in systems for emotions, and verifying the basic assumption through observation for actions of the models. These studies basically represent processes of emotions in one individual and focus on how emotions in the individual change and actions are affected according to stimuli from environments including social situations. They are divided into two kinds of models, one where symbolic states of emotions are given in advance (e.g., Frijda & Moffat, Elliott in Tokosumi et al (1994)), and another where explicit emotional state does not exist in advance and they are implicitly and emergently determined through information processes (e.g., Ono & Sato, 1995, Pfeifer in Tokosumi et al(1994)).
On the other hand, some computational models for emotions have recently been studied from the viewpoint of group dynamics, that is, interaction processes that social structures and individuals' emotions affect and constrain each other (Fujisawa, 1997; Nomura, 1997. These models represent bidirectional processese between the change of emotions in individuals by social relations and the change of social relations by emotional states. The aim of these models is to verify social significance of emotions from the viewpoint of social psychology. We are aiming to apply these computational models for emotions to a new field of sociology, sociology for emotions (Hochschild, 1979; Okahara et al., 1997), from the viewpoint of group dynamics. It is Gergen's generativis (Gergen, 1994) that provides a direction for the research.
Sociology of emotions is an attempt to develop a new perspective in sociology by focusing on emotions. In the traditional sociology, emotions have actively not been dealt with because social phenomena and organizations have been analyzed based on the assumption that important decision should be rationally made. Moreover, since analogical models from natural science like physics have suceeeded to explain phenomena even in sociology, emotions have been regarded as those which are impossible to objectively observe and deal with.
There are three reasons why emotions can be subject in sociological research under the above situation. First, focusing on emotions leads to more realistic understanding of social actions. Second, it can become a strategy to develop a new perspective in sociology for a paradigm change. Third, there has been a variety of emotional phenomena that cannot be neglected in the modern societies, such as excess and/or deficiency of feelings in families, commoditization of feelings, mental fatigue due to emotion management, and so on. Under this situation, sociology of emotions has appeared for requirement for practical efforts.
One of important concepts in sociology of emotions is ``feeling rules'' proposed by Hochschild(1979). For example, we often speak of ``having the right'' to feel angry at someone, or we say that we ``should feel more grateful'' to a benefactor. Feeling rules are a set of socially shared guidelines that direct how we want to try to feel and not to feel emotions according to given situations. Based on this concept, Hochschild showed a problem of emotional workers like airline stewardess who are requiered a high degree of emotion management. Moreover, Yamada showed the fact that emotions have become a strategy for social control in modern society, and Okahara claimed a cultural trend of extreme desire for emotions and its evil practice in modern society (Okahara et al., 1997).
On the other hand, Gergen criticized the fact that logical empiricism has been adopted in social psychology as a meta thoery, from the viewpoints of the fundamental difference between the phenomena in natural and social sciences (complexity of the phenomena and affection of observers to objects), the difficulty of objective description of human phenomena (not ostensive description but one including actors' motivations), and the problem of theories and observation (pseudo-relation between hypothetical constructs like attitudes and motivations, and observed results which are another hypothetical constructs) (Gergen, 1994). He claimed that the purpose of social psychology should not be the ones of logical empiricism (understanding, prediction, and control of phenomena), and proposed a new criteria for determining quality of theories in social psychology, ``capability of generativity''. Generative theories have capablity of doubt premises and common senses in societies, proposing fresh alternatives for societies, and changing the current social order as a result. In addition, as concrete methodologies for the purpose, he proposed specification of minor opinions in societies, extreme extension of common sense frameworks, exploration of antithesis, and exploration of new metaphors.
In Gergen's generativism, mathematical and computational models in social psychology should not be used to understand, predict, or control phenomena. They should and can be tools not only to verify whether presumptions in models for social phenomena with a natural language really derives results implicated from the models, but also to derive an implication different from results predicated in the models with the natural language for constructing generative theories. Models of artificial societies (Epstein & Axtell, 1996) are representative ones in the view of group dynamics.
This fact connects between computational models for emotions and sociology for emotions. Sociology of emotions unconsciouslly adopts Gergen's generativism as a meta theory because one of its aims is to develop a new perspective on emotions in sociology like the problems of emotional workers. When models with natural languages have a limit for social implications, computational models for emotions can be useful tools for implicating a new social perspective beyond the limit. In particular, models of artificial societies including emotions and distributed aritificial intelligence in which feeling rules are implemented in agents are appropriate for sociological implication. Moreover, as Yamada suggested that the concept of ``double bind'' can be explained from the viewpoint of feeling rules (Okahara et al., 1997), we consider the possibility of application of computational approach to psychopathology by implementing feeling rules in the context of double bind with the concept of ``affection-cognition systems'' (Ciompi, 1994).