Research On Modeling And Feedback Tactics Of The Emotional And Pedagogical Agent

Intelligent agents, one of the very important research fields of artificial intelligence, are widely applied due to their characteristics of autonomy, adaptability, pro-activeness and social ability. More and more intelligent tutoring systems begin to adopt intelligent agents as their key technology of the architectures, out of which a new research trend, pedagogical agents come into being. Modern cognitive science and neurophysiologic researches show that emotions exert great influence on human cognition, and pedagogical researches pay continuous attention to emotional experience in students' learning process. Therefore, one very clear research objective in this field is to endow the pedagogical agents with the ability of how to handle emotions.Inspired by emotional model research in psychology, the role of the emotional elements in human-machine interaction in the modeling of intelligent systems attracts attention in the intelligent agent research field, and some models of emotions come out one after another. However, most of the existent models still focus on the external reactive behaviors, not the internal states of the intelligent agents. Therefore, in order to synthesize emotions, these agents are more based on some static rules or some pre-determined domain knowledge, then, they lack adaptability. At the same time, the existent models of emotions handle emotions with a black-and-white matter, which don't show any fuzzy and dynamic nature of emotions.The findings of a comparative study on the existent representative pedagogical agents show that although the present researches have attempted to make use of students' emotional states to choose the teaching methods more suitable to students, there are few researches that can arouse and motivate students' positive emotions and then form an emotional support for the students. The findings still show that even though the present pedagogical agents have touched upon the modeling of emotions, they only have done that on the level of increasing believability, not endowing pedagogical agents with the ability of emotions recognition and the mechanism of emotion synthesis.In this dissertation, fuzzy rules are adopted to explore the role of fuzzy logic in the emotional modeling of pedagogical agents, which embodies the fuzzy and dynamic nature of emotions; machine learning algorithm is utilized to realize the experience information learning of pedagogical agents, which strengthens the adaptability of the model. Based on these, in this dissertation, an emotional and pedagogical agent (EPA) is designed, who can recognize students' emotions, own the ability of showing emotions, provide emotional support for students during their learning process, and then promote students' learning performance.The main achievements of this dissertation include:(1) Modeling of emotions based on fuzzy logic and machine learning.This dissertation adopts fuzzy set to represent the degree of impact of events upon objectives, the degree of importance of objectives, the degree of desirability. It explores the reasoning method of the degree of desirability based on fuzzy logic rules as so to realize the smooth transition of observations and emotional states with comparatively less rules, which at the same time is more in accordance with the fuzzy nature of human emotions. Using machine learning algorithm realizes the pedagogical agents' learning of the experience information such as event expectation, which makes pedagogical agents be able to react emotionally and with dynamic adaptability.(2) Recognition of emotions based on students' learning motivation types.The findings of students' study motivation show that students' studying motivation type (performance-oriented or mastery-oriented) can exert great influence on students' possible emotional states during their learning. Therefore, students' motivation orientation is regarded as a key factor during the emotion recognition so as to realize the individual differences of emotional recognition.(3) Affective feedback tactics of pedagogical agents which aim at arousing students' positive emotions.The researches in psychology and pedagogy show that students' negative emotions are hindrance to their learning and their positive emotions are impetus to their learning. Based on the recognition of students' emotional states and acquisition of their learning motivation types, a series of emotional feedback strategies and decision-making mechanisms are designed for emotional and pedagogical agents so as to encourage students and arouse their positive emotions and support them emotionally.