Using Pedagogical Agents to Support Collaborative Distance Learning Patrícia Jaques*
Adja Andrade**
João Jung*
Rafael Bordini* Rosa Vicari*
*
PPGC, **PGIE – Federal University of Rio Grande do Sul {pjaques, jjung, bordini, rosa}@inf.ufrgs.br, **
[email protected]
*
ABSTRACT In this paper, we describe an animated pedagogical agent that assists the interaction among students in a virtual class within a collaborative communication tool, either on-line or not, motivating them, correcting wrong ideas and providing new information. This guiding agent, which we call Collaboration Agent, will consider not only the cognitive capabilities of students, but also social and affective ones; this becomes a more qualitative mechanism for collaboration among students and learning. The considered agent is being modelled as part of the multi-agent architecture of the project “A Computational Model of Distance Learning Based on Socio-Cultural Pedagogical Approaches” (Andrade et al, 2001).
KEYWORDS Pedagogical Agents, Dialogue Analysis, Communication Tools
INTRODUCTION With the popularisation of computer nets and all technological capability they offer, we can observe the fast development of computer supported collaborative learning systems (CSCL). For students to feel involved in their study, they must interact with other students attending the course. As a classroom, students must work in groups, exchange ideas and help each other. To support the interaction among students, collaborative systems provide tools that facilitate the online interaction, such as chat, bulletin board and discussion lists. These software are good mechanisms of conversation among students, but they do not provide any guidance or direction for the student during or after the dialogue sessions (Soller, 2001). In this paper, we focus on the construction of an animated pedagogical agent that assists the interactions among students in a collaborative tool, either on-line or not, motivating them, correcting wrong concepts and providing new knowledge. This guiding agent, that we call Collaboration Agent, will consider not only the cognitive capabilities of students, but also social and affective ones, which becomes a more qualitative mechanism for collaboration among students and learning. The agent considered in this work, called Collaboration Agent, is being modelled as part of the multi-agent architecture of the project “A Computational Model of Distance Learning Based on Socio-Cultural Pedagogical Approaches” Piaget (1976, 1978, 1979, 1983), Freire (1980, 1995, 1996), Lévy (1999), Morin (1990) and Vygotsky (1998). This project was first presented in (Andrade et al., 2001). In the next section we describe the conceptual background of this research: Computer Supported Collaborative Learning, the Socio-Cultural Pedagogical Approaches, and Pedagogical Agents. In the Section 3, we describe the proposed system and the agents that are part of it. In Section 4, we describe the system working as a guiding tool in collaborative activities, the focus of this paper. Finally, in Section 5, we present some conclusions and future work.
CONCEPTUAL BACKGROUND The work presented in this paper is interdisciplinary and comprehends some research areas such as: Computer Supported Collaborative Learning and Distance Education, Socio-Cultural Pedagogical Approaches, and
1
Pedagogical Agents. In this section, we present the background theories which are necessary to understand the work presented in this paper. Computer Supported Collaborative Learning The development of high performance computing and of information technologies (such as e-mail, www, bulletin boards and chat) has made possible a model of “distance teaching” and not of “distance learning”. It is beyond doubt that teaching at distance has brought several benefits, facilitating the access to information, the individualised study and the gathering of data. However, its presuppositions are still focused on classical paradigms of instruction. What actually happens is that teaching has been interpreted as synonym for learning, which actually it is not. Teaching focuses on the transmission of knowledge, from the teacher through the didactic materials and distributed knowledge bases. Teaching does not focus on the student, or the social group. It is necessary to change this model of instructional training for a social model and collaborative learning. Clearly there is no teaching without learning, but it is necessary to look for a new approach, to establish new strategies, providing shared understanding and the distributed solution of problems. The role of technology will be essential for these changes. However, our purpose has been not just to use the technology to inform the individuals, but above all to form them in a permanent way and at distance. In this sense, the technologies of distributed artificial intelligence (in particular multi-agent systems), virtual reality, multimedia systems and the Web can constitute excellent allies in the construction of a learning model that favours social interaction among individuals, allowing for the formation of groups or virtual communities of learners. In terms of research in collaborative learning, Pierre Dillenbourg (Dillenbourg, 1994) analyses various conceptions of learning that can give foundation for a computational model: the constructivist approach, which focuses on the individual in the context of social interaction, the socio-cultural approach, which focalises the relationship between the individuals, and the shared cognition approach, which focalises the environment (which includes a physical context as well as a social one). Grounded on these conceptions, and aiming for a computational model of social learning, this paper proposes an approach of multi-agent systems based on the socio-cultural perspective proposed by Piaget (1976, 1978, 1979, 1983), Paulo Freire (1980, 1995, 1996), Pierre Lévy (1999), Morin (1990) e Vygotsky (1998). The Socio-Cultural Approaches Currently, one of the greatest challenges of research in artificial intelligence applied to Education is the creation of a framework or adequate model of collaborative learning that privileges the collective. In order to construct such a model, not only computational tools are necessary; we also need suitable architectures and adequate learning theories, strategies, and tactics. We base our studies on theories that privilege the social level, so that the proposed computational framework is in accordance to an accepted pedagogical model. Some social-interactionists theoreticians will serve as inspiration for this research. Piaget (1976, 1978, 1979, 1983) Freire (1980, 1995, 1996), Lévy (1999), Morin (1990), Vygotsky (1998), and others supply a theoretical support to this model, because they consider the interaction as a bi-directional process and a joint action among the participants. The choice of Vygotsky as the main reference of this work was made because of the emphasis that he gives to the process of mediation and to the interaction among students, a basic requirement in collaborative learning. One of the most important concepts in Vygotsky’s theory is that mental activities are based on social relationships between the individual and the environment in a historical process and that this relationship is mediated by symbolic systems, through instruments and signs. For Vygotsky (1998), the signs are artificial incentives with the purpose of mnemonic aid; they work as middle ground for adaptation, driven by the individual's own control. A sign act as instruments of the psychological activity. They are auxiliary means to solve a given problem and they are guided internally. The function of an instrument is to serve as tool between the worker (in the case of this research, the student) and the object of his work, providing help in some activity; these are guided externally. Both have in common the mediation function. Another fundamental concept in Vygotsky’s theory is the Zone of Proximal Development (ZPD). In mentioning the ZPD, it is necessary to define which are the levels of the student's development: the Level of Real Development (LRD) refers to the functions that the student already possesses. The Level of Potential Development (LPD) determines the functions that a student can develop, through an adult's aid or from the collaboration of more experienced colleagues. ZPD is "the distance between the real level of development, determined by the capacity to solve a problem independently, and the level of potential development, determined by the resolution of a problem 2
under an adult's orientation or in collaboration with other more capable students" (Vygotsky, 1998). Besides these concepts, Vygotsky defends that cognitive functions happen first at the social level for later to happen at an individual level: firstly among people (inter-psychological) and, later, within the person (intra-psychological). The psychological abilities, in the same way, happen initially in social relationships (inter-mental) and afterwards within the child (intra-mental). Pedagogical Agents Since the 70’s, with the arising of Intelligent Tutoring Systems, researchers in Computer Science in Education observed the necessity to use techniques from Artificial Intelligence to make educational systems more flexible and customised to their users. The collaborative approach to distance education has produced a great number of problems and activities where the technology of agents can employed, not only in assisting and monitoring the students, but also in providing information to the teacher. Agents are stand-alone entities that have knowledge about themselves and about other agents in the society and, therefore, can collaborate with each other to reach a common objective in a shared environment (Shoham, 1997) (Jennings et al., 1998) (Wooldridge and Jennings, 1997). In this agent-oriented approach, the modular architecture of the system is substituted by a society of agents who work in a cooperative form using diverse techniques of Artificial Intelligence and integrated to an Intelligent Tutoring Systems (ITS). According to Viccari (1990), ITS are programs that, interacting with the student, modify its knowledge through the capacity of learning and customise the learning strategies to the cognitive model of the students based on their actions. The main objective of these systems is to provide a suitable instruction to the student in terms of content and form. An educational system that has learning strategies and that is formed by intelligent agents is called Pedagogical Agent. In these systems, agents can be used as personal assistant and as animated characters that interact with the user, or as cooperative agents who work in the background as part of the architecture of the educational system (Gürer, 1998). The use of agents in the conception of educational systems has some advantages, such as the ability to react to the actions of the user, credibility, modelling of multi-user collaborative systems, modularity and “openness”. In this way, as each agent is an independent component, it is easier to add other agents to the system (Giraffa, 1995).
DESCRIPTION OF THE ENVIRONMENT Five types of artificial agents are part of the architecture of the system – Diagnostic Agent, Mediating Agent, Collaboration Agent, Social Agent and Semiotic Agent. The human agents that interact with the artificial ones are learners and teachers. The Mediating Agent is an interface agent that is responsible for presenting pedagogical contents to the student. All information on user actions will be gathered by the Mediating Agent and sent to the Diagnostic Agent. The Diagnostic Agent updates the information in the student model and verifies, according to received data, if it is necessary to use a new educational tactic; in this case, it sends this information to the Mediating Agent. If this tactic is, for example, the presentation of an instructional content, the Mediating Agent makes a request to the Semiotic Agent. The Semiotic Agent searches the requested sign or instrument in the knowledge base (for example: contents, animations, videos) and sends them to the Mediating Agent for them to be shown to the student. When the Diagnostic Agent finds a deficiency in the student’s learning and considers it would be interesting to perform an activity in group, it will make a request to the Social Agent. The Social Agent will create a Collaboration Agent and form a study group of students. The Collaboration Agent will monitor and mediate the interactions among them. There is a Diagnostic Agent and a Mediating Agent for each student, a Semiotic and Collaboration Agent for the whole society, and a Collaboration Agent for each group of students that has been formed. Note that the tutoring system may function as an individual tutor, where the Mediating Agent presents pedagogical contents to the student in accordance to his/her profile and cognitive style, or as a facilitating system of collaboration, where the Collaboration Agent monitor and mediate the interaction among the students in a collaborative tool. The general architecture of the system can be seen in Figure 1. In this article, we shall focus on the system as a collaborative tool. In this case, the Collaboration Agent has an essential role; its function is to promote and to mediate the interactions of groups of students using communication tools (for example chat, discussion list, and distribution list). In such a way, it attends the students during the interactions, stimulating them when they look unmotivated, presenting new ideas and correcting wrong ones.
3
In the next sections, we describe the functionalities of each of the agents who are part of the proposed system. Later on, we describe the functionalities of the system as a collaborative tool.
Social Agent
Semiotic Agent
Pedagogical Content
Pedagogical Content Request ZPD Agents
Knowledge Base Access
Pedagogical Content Request
WW W Exercises Examples
Pedagogical Tactic Request Collaborative Group
Diagnostic Agent Cognitive Profile
Student Behavior
Mediating Agent
Diagnostic Agent
Mediating Agent
Mediating Agent Actions
Student Model Affective Profile
Pedagogical Tactic
...
Collaboration Tool User Action
Pedagogical Content
Student’s Message Create Collaboration Agent Group Profile
Group Profile
Message to the student Message to Collaboration the student Agent
Student
Student
Group Model
Figure 1: A society of Social Agents for a Learning Environment Diagnostic Agent The system has Diagnostic Agents that are responsible for observing the real development of the students and for proposing activities which would make their real capacities (Real Development Level) as close as possible to the desired ones (Potential Development Level). As each of these agents monitors only one student, there is one Diagnostic Agent for each student. The Diagnostic Agent, based on Vygotsky’s theory (1998), is responsible for stimulating those functions of the student that have still not matured, but are in a development process. A Diagnostic Agent can have functions such as: to vary the degree of control of the joint activities, to consider tasks gradually, or to modify the offered forms of help/support. In the Figure 2, we can see the internal architecture of the Diagnostic Agent.
4
Diagnostic Agent Actions Assistance degree
Mediating Agent Actions
Pedagogical Content
Consolidate Abilities Support Types: Moderate Intensive Guided
Student Cognitive Model
Diagnostic
Student ZPD
Suggested Tactics
Collaboration Agent
Abilities in Consolidation
Tactics Module Tactic 1 Tactic 2 … Tactic n
Collaborative Tactics
Figure 2: The Internal Architecture of the Diagnostic Agent In order to assist the student in the learning process, a Diagnostic Agent must have a student model, identifying his/her abilities and deficiencies, which is constructed by the observation of the student’s interaction with other students and with instructional contents. This way, it is able to indicate how the student’s cognitive abilities have expanded; it does so by having access and modifying this model. Mediating Agent The Mediating Agent is responsible for the interface between the system and the student. The main difference between a Mediating Agent and a Diagnostic Agent is that the first one carries through all the tasks of interface and communication with the user. The Diagnostic Agent, in turn, is responsible for the learning process (that is, the construction of the student’s model), and for the identification of deficiencies in learning through the observation of the actions of the user. Besides, the Mediating Agent must have access to the student’s model, helping the prediction of the behaviour of that student. This allows for determining the best actions to be executed in order to assist in the student’s learning process. This agent will be implemented as an animated pedagogical agent. Semiotic Agent For the Mediating Agent to fulfil its role, it is necessary the intervention in the students’ interactions by means of external stimuli (like instruments and signs). The Semiotic Agent assists the cognitive activity of the student by introducing these elements, for example in order to help the student in solving a problem. For that purpose, the agent will use signs such as pictures, sounds, texts, and numbers. The Semiotic Agent dynamically constructs a Web page to be presented to the student, showing pedagogical contents at increasing levels of difficulty when necessary. Moreover, it is also responsible for the feedback to the knowledge base, verification of the existence of links for future use, and it has rules for making decisions on which contents should be presented and how they should be presented. The Semiotic Agent can also allow to the interactions among students and teachers. Although the Collaboration Agent is the one responsible for monitoring and mediating the interactions among students, the Semiotic Agent must make available the instruments for social interaction, such as e-mail tools, meeting scheduling, chat, and FAQ. Moreover, the Semiotic Agent can also record how long a student uses a certain contents page. In the Figure 3, we present the architecture of the Semiotic Agent.
5
Semiotic Agent
Pedagogical Content Request
Social Agent
KQML Message
Check’s Student Tactics and Preferences
Search Pedagogical Content
Knowledge Knowledge Base Base Access
Build Content Mediating HTML Page Agent
Figure 3: The Internal Architecture of the Semiotic Agent Moreover, as the Semiotic Agent is responsible for constructing dynamically the contents page to be shown to the student, we can see it as an interface designer. The Semiotic Agent has the means to decide which signs fit better in each particular situation, since it is implemented as an intelligent agent and it has features such as autonomy, communication, adaptability and rationality. The control of this process of sign presentation is made by the Semiotic Agent, which searches for them in the domain database based on the pedagogical tactics specified by the Diagnostic Agent based on the information received from the Mediating Agent. Collaboration Agent and Social Agent The Social Agent searches for peers that are capable of assisting a student in his/her learning process and creates a Collaboration Agent for mediating the interaction among the students. The Collaboration Agent will monitor and mediate the interaction between students in collaborative communication tools (for example, chat, discussion list and bulletin boards). It attends the students during the interactions, stimulating them when they look unmotivated, presenting new ideas and correcting wrong ones. In the figure 4, we show the internal architecture of the Collaboration Agent. During the interaction with the students in the collaborative tool, the Collaboration Agent interacts with the Diagnostic Agent to obtain new tactics to be used. In such a way, it must send the actions of the user, in this case, sent messages, so that the Diagnostic Agent decides which tactics must be carried out. The Collaboration Agent interacts with the Semiotic Agent to get the pedagogical content. For example, the Collaboration Agent can check, in accordance with statistical analyses of the students’ message, which students presented incorrect ideas. As the interactions progress, the Diagnostic Agent can decide if a more difficult subject can be presented. In that case, the Collaboration Agent requests that the Semiotic Agent sends certain contents at a more difficult level. The Collaboration Agent updates the affective model of the student. It is responsible for obtaining the affective state of the student and updating the student model, in order to reply to the student with an appropriate emotional behaviour. In collaborative learning, the group is an active entity; therefore, the system must contain information that refers to it as a whole. This information generates a group model, which is constructed and stored by the Collaboration Agent.
6
Knowledge Base
Collaboration Agent Messages
Extract Message’s Parts
Student’s message
Check Message Subject
Check Student’s Affective State
Check if content is correct
Semiotic Contents OK or Agent Send correct contents Student action
Generate Tactic
Collaborative Tool
Access Knowledge Base
Student
Student
Agent Action
Cognitive Tactic
Diagnostic Agent
Student
Cognitive Model
Affective Model
Student’s Affective Model Access and Update
Student’s Cognitive Model Access and Update
Student Model
Figure 4: The Internal Architecture of the Collaboration Agent
DESCRIPTION OF THE COLLABORATIVE ENVIRONMENT We can better understand the implementation of the collaborative approach through a scenario. Let us imagine a student, using his/her computer (at home or at work), connected to this system through the Internet. A Mediating Agent will be sent to the user's machine and it will monitor his/her activities. The Mediating Agent gathers the information with the student’s profile. The information gathered by the Mediating Agent is sent to the Diagnostic Agent. Initially, it will suggest the educational contents to be shown to the student. The Diagnostic Agent will suggest to the Mediating Agent a tactic to be used for the presentation of the proposed contents, based on the inferences on the student model. The Mediating Agent will send the request for new contents to the Semiotic Agent. The Semiotic Agent will send back to the student the actual contents page in HTML, which will be shown in the browser running on the student’s computer. When a Diagnostic Agent notices there is a problem with the student’s learning for which it is necessary the intervention of other facilitators or other more capable colleagues and/or teachers, it will make a request to the Collaboration Agent. Then, the Collaboration Agent will invite certain students with a suitable profile to participate in an interaction by means of a collaborative tool, which can be a chat, a discussion list, or a bulletin board, for example. The Collaboration Agent will monitor the discussions among the students, stimulating them when they look unmotivated, presenting new ideas and correcting wrong ones. This agent will be connected to the collaborative tool, just as any another user of the system, which gives greater realism to it. The information gathered by the Collaboration Agent during the interaction among the students will be stored in the group model, as well as in the individual student model if the information is on affective states. The information related to cognitive states will be sent to the Diagnostic Agent, which decides on what tactics should be used and updates the student model.
COLLABORATION AGENT IMPLEMENTATION Due to its social function – to communicate with students, to promote and monitor the interaction among students – it would be interesting for the Collaboration Agent to have an interface that would allow it to exploit students’ social nature. In fact, one of our main concerns is to better exploit the social potential of the students to improve their 7
learning. Studies demonstrate that people interacting with animated characters learn to interact with other humans (Huard, 1998). Therefore, we chose to represent it as an animated character who has a personality and which interacts with the student through messages in natural language. Thus, as in human social interactions, the Collaboration Agent must be able to show and perceive emotional responses. Learning is a comprehensive process which does not simply consists of the transmission and learning of contents. A tutor (in this case, the Collaboration Agent) must promote the student's emotional and affective development, enhancing his/her self-confidence and a positive mood, ideal to learning. The way in which emotional disturbances affect mental life has been discussed in the literature (Goleman, 1995). He recalls the well-known idea that depressed, bad-humoured and anxious students find greater difficulty in learning. In order to interact with the student in an adequate way, the agent has to interpret his/her emotions correctly. Therefore, it is necessary for the Collaboration Agent to have not only a student’s cognitive model, but also an affective one. We are going to use the student model proposed by (Bercht et al., 1999), which considers affective states such as effort, self-confidence and independence. In collaborative learning, the group is an active entity; therefore, the system must contain information that refers to the group as a whole. This information generates a group model, which is built and maintained by the Collaboration Agent. The Collaboration Agent can build the group model from the individual student models, which are obtained from the interactions between the students and their Mediating Agents, and updated by the Diagnostic Agents. The group model can also be obtained from the observation of the group as a whole. Nevertheless, it is necessary to have in mind the responsibility associated with the use of affective agent architecture for interaction with the user, especially in education. Often we observe that agents have attitudes that are not suitable to students’ mood (e.g., if an agent gets sad when the student could not carry out an exercise). This kind of attitude may generate a disturbed reaction in the student, making him/her more anxious and less self-confident. It is necessary to identify which behaviours are appropriate to promote a mood in the student that provides better learning conditions. The Collaboration Agent will carry out the analysis of the students’ conversations based on statistical methods, such as pattern matching, message categorisation and information retrieval (Soller, 2001). The messages will be generated in natural language, using dialogue models and frames. This analysis will be based on the work in (Jaques & Oliveira, 1999; Jaques et al., 2000).
CONCLUSIONS AND FUTURE WORK One of the contributions of this paper is the possibility to apply elements of Vygotsky’s theory (ZPD analysis, mediation through symbols) for computational learning systems. Beyond this contribution, the pedagogical agents will make the role of mediating (i.e., the Diagnostic, Mediating, and Collaboration Agents) or agents that give support to the mediation (i.e., the Semiotic and Social Agents), assisting in a more complex process carried out by the teacher inside an distance education environment. From the point of view of the collaboration, the authors see in the use of social agents new promising perspectives to stimulate distance interaction and participation. The collaboration process begins when a student has difficulty in learning something or accomplishing some task. At this moment, the agents must apply adequate collaborative strategies. This research, based on Vygotsky’s theory (Vygotsky, 1998), intends to be applied for groups of students of different cognitive levels. So, a more capable student can assist in the learning of another student with difficulties. Therefore, our analysis unit will be the group and not only the individual. The activities and tools must stimulate cooperation. This emphasis in the group model, grounded on cognitive and affective points of view, is one of the distinguishing features of this work. This model will be constructed from the analysis of the student’s interactions through collaborative synchronous or asynchronous tools, observing elements such as leadership, participation, and autonomy. This research is currently in the phase of implementation design and specification. The phase of study of the pedagogical model and computational architecture is concluded. This work is the result of a research project in the area of Artificial Intelligence applied to Education that intends to create a computational framework (of which the pedagogical agents are part) to support collaborative learning.
8
REFERENCES Andrade, A. F.; Jaques, P. A.; Jung, J.; Bordini, R.; Vicari, R. M.; (2001) A Computational Model of Distance Learning Based on Vygotsky's Socio-Cultural Approach. Proceedings of the MABLE Workshop (Antonio, Texas, May 19-23). X International Conference on Artificial Intelligence on Education. Bates, Joseph. (1994) The Role of Emotion in Believable Agents. Communications of ACM, July. Bercht, M.; Moissa, H.; Viccari, R.M. (1999) Identificação de fatores motivacionais e afetivos em um ambiente de ensino e aprendizagem. Proceddings of SBIE’99 (Curitiba, November 23-25) (Poster). Boone, K. (2000) A Human-computer Dialogue Authoring System for Web Interaction. URL: http://citeseer.nj.nec.com/390875.html. Canamero, D.; Velde, W. de. (1997) Socially Emotional: Using Emotions to Ground Social Interaction. Socially Intelligent Agents. Proceedings of AAAI Fall Symposium. Menlo Park: AAAI Press. Castelfranchi, C.; Rosis, F.; Falcone, R. (1997) Social Attitudes and Personalities in Agents. Anais Socially Intelligent Agents. Proceedings of AAAI Fall Symposium. Menlo Park: AAAI Press. Dillenbourg, P.; Baker, M.; Blaye, A; O´Malley, C. (1994) The evolution of research on Collaborative Learning. In Spada, H. and Reimann, P., (eds.), Learning in Humans and Machines. D’amico, C. B.; Pereira, A. S.; Geyer, C. F. R.; Viccari, R. M. (1998) Adapting Teaching Strategies in a Learning Environment on WWW. Proceedings of World Conference of the WWW, Internet & Intranet (Orlando, USA). Elliott, C.; Rickel, J.; Lester, J. (1997). Integrating Affective Computing into Animated Tutoring Agents. Proceedings of Workshop on Animated Interface Agents: Making Them Intelligent. IJCAI '97 (Nagoya, Japan, August). pp. 113-121. Finin, T.; Labrou, Y.; Mayfield, J. (1997). KQML as an agent communication language. BRADSHAW, Jeffrey (Eds.) Software Agents. Menlo Park: AAAI Press/The MIT Press, 1997. p.291-316. Foner, L. (1997). Entertaining Agents: A Sociological Case Study. Proceedings of the First International Conference on Autonomous Agents (AA '97),( Marina del Rey, California). Freire, P.; Fagundes, A. (1995) Por uma Pedagogia da Pergunta. Ed. Paz e Terra, Rio de Janeiro. Freire, P.; Fagundes, A. (1996) Pedagogia da Autonomia; Saberes Necessário A Prática Educativa. Ed. Paz E Terra, São Paulo. Freire, P.; Fagundes, A. (1980) Conscientização: Teoria e Prática da Libertação: uma Introdução ao Pensamento De Paulo Freire. Editora Moraes, São Paulo. Fuks, H.; Laufer, C.; Schwabe, D. (1995) ACCORD – Representing Conversation Clichés for Cooperation. Proceedings of the International Workshop on the Design of Cooperative Systems (Juan-les-Pins, France, Janeiro), pp 351-369. Goleman, D. (1995). Inteligência Emocional. Objetiva, 19995. Gürer, D. (1998). The Use of Distributed Agents in Intelligent Tutoring. Proceedings of Workshop On Pedagogical Agents, ITS’98 (San Antonio) p.20-25. Huard, R (1998). Character Mastery with the Improvisational Puppets Program. Technical Report (KSL-98-11) – Stanford University. Jaques, P.; Andrade, A.; Moraes, M.; Móra, M. C (2000). Uma Arquitetura de Agente para a Análise Qualitativa da Interação em Ambientes de Educação a Distância. Proceedings of SBIA (Maceió, November). Jaques, P.; Oliveira, F. M. (1999). Software Agents for Analysis of Collaboration in a Virtual Classroom. Proceedings of AI-ED99 (Le mans) Le mans, L’Université du Maine. Jennings, N.; Sycara, K.; Wooldridge, M. (1998) A Roadmap of agent research and development. Journal of Autonomous Agents and Multi-Agent Systems, 1:275-306. Levy, P (1999). Inteligência Coletiva. Por uma antropologia do ciberespaço. Edições Loyola, São Paulo. Loyal, A.; Bates, J. Personality-Rich Believable Agents that Use Language. Proceedings of Agents’97. (Marina del Rey, CA, USA). ACM Press. Móra, M. C.; Viccari, R. M.; Lopes, J. G. P.; Coelho, H (1998). BDI models and systems: reducing the gap . Proceedings of ATAL’98 (Paris, July). Morin, J. F; Lelouche, R (1997). Tutoring Knowledge Modelling as Pedagogical Agents in a ITS. Proceedings of Workshop V – Pedagogical Agents. AI-ED 97 (Kobe, Japan, August). Morin, E. (1990) Introdução ao Pensamento Complexo. Lisboa: Publicações Instituto Piaget. Ogata, H.; Yano, Y. (2000). Combining Knowlegde Awareness and Information Filtering in an Open-ended Collaborative Learning Environment. Intelligence Journal of Artificial Intelligence in Education, 11, 33-46.
9
Piaget, J. (1976) A Equilibração das Estruturas Cognitivas - Problema Central do Desenvolvimento. Zahar Editores, Rio de Janeiro. Piaget, J. (1978) A Tomada de Consciência. Edusp/Melhoramentos, São Paulo. Piaget, J. (1979) O Nascimento da Inteligência na Criança. Ed. Zahar, Rio de Janeiro. Piaget, J. (1983) A Epistemologia Genética /Sabedoria e ilusões da Filosofia; Problemas de Psicologia Genética; 2. Ed. – São Paulo: Abril Cultural. Pilkington, R. (2001). Analysing Educational Dialogue Interaction: Towards Models that Support Learning. Intelligence Journal of Artificial Intelligence in Education, 12. Porayska-Pomsta, K.; Melish, C.; Pain, H. (2000). Aspects of Speech Act Categorization: Towards Generating Teachers’Language. Intelligence Journal of Artificial Intelligence in Education, 11, 254-272. Rouke, L.; Anderson, T.; Garrison, D. R.; Archer, W. (2000) Methodological Issues in the Content Analysis of Computer Conference Transcripts. Intelligence Journal of Artificial Intelligence in Education, 11. Shoham, Y. (1997). An Overview of Agent-Oriented Programming. Bradshaw, Jeffrey (Ed.). Software Agents. Menlo Park: AAAI Press/The IT Press, p.271-290. Soller, A. (2001) Supporting Social Interaction in an Intelligent Collaborative Learning System. Intelligence Journal of Artificial Intelligence in Education, 11. Viccari, R. (1990) Um Tutor Inteligente para a programação em Lógica - Idealização, Projeto e Desenvolvimento, Universidade de Coimbra. (PhD Thesis). Vygotsky, L. S. (1998) A Formação Social da Mente: o Desenvolvimento dos Processos Psicológicos Superiores. Editora Martins Fontes, São Paulo. Wooldridge, M.; Jennings, N. (1997) Intelligent Agents: Theory and Practice. URL: http://www.doc.mmu.ac.uk/ STAFF/mike/ker95/ker95-html.html.
10
