This study was pursued to determine the essential characteristics of a fifth-grade technology education unit of study utilizing an expert system for bridge design. A panel of artificial intelligence experts was consulted to determine the feasibility of using expert system software to model bridge design, and a panel of education experts was consulted to determine the critical components for the study of bridge design in the fifth grade.It was determined that the most important components of such a study were integration with other curricular areas, most notably mathematics and social studies, and the incorporation of cooperative and group work among students. It was also determined that bridge design is a suitable knowledge domain for expert system consultation.Finally, recommendations for further study were stated. / Department of Industry & Technology

本研究的目的，是在實作一個以開放式問答題的作答方式為基礎的線上輔助學習系統，藉以提升評量及學習的效果。我們選定的問題領域在地球科學，以極光的發生為例，將問題所涉及的觀念進一步拆解細分成許多小觀念，建立階層化使用者知識模型，以評估學生於各個觀念的分別表現。學生於作答時，系統會即時分析學生的答案，以評估學生在各觀念的表現，並在適當的時機提供回饋，嘗試協助學生思考原本未能答好的觀念。我們以不同的回饋時機及回饋內容為獨立變因，嘗試瞭解對使用者答題表現的影響。我們也以個案分析的方式將學生的對於系統學習態度分類，藉以瞭解不同類型學生對此類系統的接受程度。我們認為此系統可以提供學生作答歷程予教學者進行分析，用以找出學生缺漏的觀念，進而協助提升教學效果。 / The purpose of this research is to use an open-ended based on-line learning system to improve the effects of evaluation and learning. We set the question domain at Earth Science. Taking the occurrence of polar lights, we divide the related concept into many small concepts, and build the hierarchical user knowledge model to evaluate student’s reactions to each concept. While students take the test, the system will analyze the answers to evaluate the students, and provide feedbacks timely to help the student think over the missing concept. Taking the different feedback timing and feedbacks as independent factors, we try to understand the influence to the users’ performance. By using case study, we classify the students’ learning attitudes to understand the different types of students’ acceptance of this system. We think that this system can provide student’s answering process to teachers for analyzing. By using this, teachers can find out students’ missing concept and to provide teaching effectiveness.

ITSs (Intelligent Tutoring Systems) provide a way of addressing some of the issues that the more traditional CAI (Computer Aided Instruction) systems do not address - the individual learning needs and individual learning abilities and levels of users - so that the user is in control of their learning experience. An ITS needs to be able to provide an explanation, for a real world situation, that successfully meets the needs of the user. To ensure relevant explanation content requires the ITS be based on sound planning principles and tutoring knowledge as well as knowledge of the domain and the user. To ensure a coherent explanation structure requires that the tutoring knowledge be applied with full recognition of the knowledge of the domain and the user. For a model of the user's knowledge to be effective, the system should be able to use it to enhance the flexibility and responsiveness of explanations generated. A user model should guide the generation of explanations so they are pitched at the correct level of the user's existing knowledge; models should be able to actively support the needs of the user so that the user's efforts in seeking out information are minimised. The aim of this research is to generate effective, flexible and responsive explanations, in educational software systems, through developing better explanation facilities than exist in currently available ITS software. In achieving this aim, I am advancing research into dialogue planning and user modelling. The explanation facilities described meet the requirements of an explanation that is tailored to the user's needs, a sound theory from which particular explanations are constructed, and a user model that can accurately represent the behaviour and beliefs of the user. My research contributions include explicitly and formally representing discourse planning / reasoning, from both the user's view and the tutor's view so that they can be clearly understood and represented in the ITS. More recent planners have adopted approaches that can be characterised as using adaptations of the classical planning approach, with informally specified planning algorithms and planning languages. Without clear, explicit and full descriptions of actions and the planning algorithm we can not be certain of the plans that such planners produce. I adopt a theoretically rigorous approach based on classical planning theory - the actions available to the planner, the planning language and algorithm should be explicitly represented to ensure that plans are complete and consistent. Classical regression planning uses dynamic planning thus enabling the system to be flexible in a variety of situations and providing the responsiveness required for an ITS. I take a theoretically rigorous approach in constructing a well specified model of discourse, building upon existing research in the area. I present a tutoring module that is able to find a way to motivate the user to take a recommended action, by relating the action to the user's goals, and that is able to reason about the text structure to generate an effective explanation - putting together several clauses of text whilst maintaining coherency. As part of developing such constructs for motivating, enabling and recommending, as well as constructs for structuring text, I use a pedagogic model based on the principled approach of (i) advising the user to take an action (ii) motivating the user to want to take the action and (iii) ensuring the user knows how to do the action. I take a clear and realistic approach to user modelling, making explicit models of the user's behaviour and beliefs. I adopt a theoretically rigorous approach, formally distinguishing between the user's reasoning and their actions, so they can be focused on separately. Formally making this distinction, more easily enables models of the user's reasoning to be tailored to the individual user. To enable the tutor to consider the full impact on the user, of the information to be delivered to the user, I use different plan spaces. I explicitly identify the different perspectives of the user and the tutor so that they can be focused on separately to generate an explanation that is tailored to the user. In my approach, reasoning about the user's skills, rules and knowledge is independent from reasoning about those of the tutor.

Esta tese situa-se na área de IA (Inteligência Artificial) aplicada à educação incluindo características interdisciplinares tanto da própria IA como de IE (informática na Educação). Faz-se também necessário constarem, aspectos referentes à Ciência da Computação e Educação a fim de melhor situar a complexidade e a dimensão do trabalho desenvolvido. A utilização de técnicas de IA na elaboração e no desenvolvimento de ambientes de ensino-aprendizagem computadorizados tem se constituído em objeto de investigação por parte dos pesquisadores da área de Informática aplicada à Educação, devido as suas potencialidades. A utilização de agentes na modelagem e no projeto de STI permite-nos resgatar antigos problemas em aberto, como por exemplo a melhoria da interação entre tutor e aluno e a possibilidade de investigação dos processos mentais em nível mais estratificado. A arquitetura descrita nesta tese utiliza a metodologia que vem sendo aplicada ao projeto de STI, onde são contempladas diferentes formas de se trabalhar com um determinado conhecimento (estratégias de ensino e táticas associadas), levando-se em consideração o tipo de usuário que está interagindo com o sistema. A arquitetura, elaborada segundo uma abordagem construtivista, prevê que o tutor seja menos diretivo e menos controlador das ações do aluno. O controle é feito na forma de monitoração para que o tutor funcione como um parceiro, ou seja, como facilitador do trabalho do aluno. Contudo, devido às características da modalidade escolhida para construção do protótipo (jogo educacional), precisamos ter algumas atitudes no tutor que garantam que o sistema não entre em colapso. O que inviabilizaria o trabalho do aluno. Nestas situações críticas, o tutor vai se comportar de maneira mais diretiva. Cabe salientar que a abordagem construtivista não significa dar liberdade total ao aluno nem privá-lo de qualquer tipo de assistência. Portanto, o que deve ser destacado é o grau de interferência do tutor, i.e., o quanto ele interfere no trabalho do aluno e se ele permite ou não que o aluno siga um caminho alternativo àquele que ele tem como o ideal para resolver o problema (heurísticas do tutor sobre o problema e forma de solução). Além destes aspectos educacionais inerentes a todo o projeto de software educacional (necessários num trabalho desta natureza), esta tese está inserida no contexto da pesquisa em agentes cognitivos modelados através de seus estados mentais. É importante salientar que os estados mentais utilizados neste trabalho (crenças, desejos, intenções e expectativas) funcionam como uma metáfora dos estudos mentais humanos. Por exemplo, quando se coloca a crença de um aluno a respeito de "lago", na realidade está se colocando a crença que temos a respeito da crença que o aluno possui a respeito de "lago". O mesmo acontece com os outros estados mentais aqui utilizados. Os diálogos reais foram registrados através de observação direta e posteriormente analisados a fim de se identificar os estados mentais relacionados. Perguntas adicionais foram feitas no sentido de obterem-se mais elementos para auxiliar na inferência do conjunto de estados mentais que o aluno possui naquele momento em que estava jogando. Após a observação de vários alunos jogando, identificou-se um certo padrão nas suas atitudes quando executavam uma ação. Observações sucessivas permitiram delinear o conjunto de estados mentais associados à ação do aluno. Tal conjunto foi utilizado como base para elaboração da coreografia. Estes dados servem de entrada para a construção do modelo do aluno mediante a interação com o tutor. No presente trabalho, nós apresentamos a modelagem de um STI através do uso da tecnologia de agentes utilizando a arquitetura de SMA (Sistemas Multiagentes). O STI é concebido como um SMA híbrido composto por um ambiente reativo (SMAR - Sistema Multiagente Reativo) e um "kernel" cognitivo (SMAC - Sistema Multiagente Cognitivo). O SMAR e o SMAC interagem entre si de para ampliar as informações quantitativas e qualitativas oferecidas aos alunos que utilizam o sistema. Estas informações disponíveis é que irão permitir ao tutor selecionar estratégias de ensino mais adequadas a um determinado tipo de aluno. A principal contribuição desta tese está centrada no "kernel" cognitivo. Nós propomos uma arquitetura para o tutor que permitirá a monitoração de dois alunos trabalhando conjuntamente. Além disso, propomos uma forma de selecionar o comportamento do tutor para oferecer auxílio personalizado aos alunos considerando o perfil de cada um. Esta arquitetura pretende ser uma alternativa de solução para uma questão importante na área de STI: Como o tutor pode selecionar, entre várias estratégias de ensino, a mais adequada para cada perfil de aluno? O grupo de pesquisa no qual este trabalho está inserido (GIA/UFRGS, sob orientação da Prof.a. Rosa Maria Viccari) tem realizado algumas avaliações experimentais, usando STI tradicionais e STI projetados e modelados através de sistemas multiagentes As contribuições científicas listadas no texto deste trabalho possibilitaram que o grupo avançasse sua pesquisa na abordagem mentalística através da criação de uma arquitetura para o tutor e favorecesse a integração do trabalho desenvolvido por Móra et al. [MOR97; MOR98]. A utilização do modelo computacional de agentes criado por Móra et al. gerou a implementação do "kernel" cognitivo. Os desafios inerentes a implementação da arquitetura proposta para o tutor ampliaram as características do ambiente criado por Móra et al. e favorecem a junção de dois trabalhos de tese supervisionados pela mesma orientadora [MOR99]. Portanto, a nova arquitetura proporcionou ganhos tanto para tais pesquisas, como para o avanço das pesquisas desenvolvidas pelo nosso grupo. / The present thesis has been elaborated within the AI (Artificial Intelligence) applied to Education realm, and it brings specific contributions to the STI (Intelligent Tutoring System) area. The use of AI techniques has been investigated by researchers of Computer Science applied to Education, due to its potentialities to improve educational systems. The agents' techniques used in the design of STI allow us to solve old problems opened in the area. For instance, the improvement of the interaction between tutor and student, and the possibility of tracing the mental processes in a more stratified way. The architecture described in this thesis uses the methodology applied to the modern STI projects: multiple strategies for the tutor (i.e., teaching strategies and associated tactics). This approach considers different forms of working with a certain piece of knowledge, and is taken into consideration to create the user profile, as well as to monitor the student interaction with the system. The architecture, designed according to a constructivist approach, expects the tutor to be less directive, and less controller of the student's actions. The control is made by an observation of students' actions by the tutor. The tutor works either as a student's partner or as a facilitator. However, due to the characteristics of the modality chosen for construction of the prototype (educational game), we needed to take some attitudes in the tutor in order to avoid the system to collapse. What would make unfeasible the student's work. In these critical situations, the tutor will behave in more directive way. It fits to point out that the construtivist approach does not mean to give total freedom to the student or to deprive it of any kind of attendance. Therefore, what should be note here is the degree of the tutor's interference, i.e., how it interferes with the student's work using its own set of heuristics. Besides these educational aspects, inherent to a project of Educational software, this thesis is inserted in the context of the research in cognitive agents modelled through their mental states (Believe, Desire, Intention, and Expectation). It is important to point out that the mental states used in this work are as a metaphor of the human mental states. For example, when the student's has a believe regarding "lake", in fact we area talking about belief that we have regarding the belief that the student possesses regarding about "lake". The same happens here with the other mental states used. The real dialogues were registered through direct observation in real situation (students playing with the game). They were analysed in order to identify the mental states connected with the student's actions. Additional questions were asked to obtain more elements to aid us to inference the group of mental states possessed by the student when he/she was playing. Successive observations allowed us to delineate the group of mental states associated to the student's action. Such group was used as a base for the choreography. These data were used as input for the construction to the student's model during the interaction with the tutor. In this work, we have presented the design of an STI with the use of architecture of MAS (Multi-agent System Architecture). The ITS is conceived as a hybrid MAS composed by a RMAS (Reactive Multi-agent System) and a "cognitive kernel" using the CMAS (Cognitive Multi-agent System). The RMAS and the CMAS interact with each other to enlarge the quantitative and qualitative information offered to the students that uses the system. These available information allow the tutor to select teaching strategies more adapted to a certain student type. The main contribution of this thesis is centred in the "cognitive kernel". We propose an architecture for the tutor that will allow the two students to work together. Besides, we propose a way to select the tutor behaviour in order to aid the students considering their personal profile. This architecture intends to be an alternative solution for an important question in ITS research: How can the tutor select, among several teaching strategies, the one that is more suitable for each student profile? The research group in which this work is placed (GIA/UFRGS - Artificial Intelligence Group), under the supervision of Profa. Rosa Maria Viccari, has been doing executing some experimental evaluations, using traditional ITS and ITS modelled through multi-agents systems techniques. The scientific contributions listed in this work allow the group to achive some interesting results in the research of STI using a mental approach. The tutor architecture favoured the integration of the X-BDI (eXecutable Belief Desire, and Intention model)) developed by Móra et al. The use the X-BDI allowed us to implement the "cognitive kernel". The inherent challenges posed by the implementation of the architecture of the tutor refined the XBDI environment. It favours the junction of two thesis works supervised done under the same supervisor[MOR99]. Therefore, the new architecture provided many gains for both researches, as well as for the progress of the research developed by our group.

Intelligent Tutoring Systems (ITSs) are softwares that aims to represent human behavior inherent in the teaching process in any particular field in order to support the activities of teachers and offer a adapted teaching to students. ITSs have great potential both in the classroom teaching as the distance, however, its construction is a complex and expensive task that requires the presence of specialized professionals in computing and system domain being compounded when you need to build ITSs large-scale and adapted to each area. In this sense, the approach Software Product Lines (SPL) allows to build large-scale ITS. Complementarily, ontologies can be used to allow such a construction is automatically adaptable to different domains. However, in the application engineering phase of the SPL, when ITS are instantiated, is required manipulation of ontologies and different software artifacts, demanding the presence of professionals with expertise in ontology and software engineering, making it difficult the accomplishment of this task by authors / designers. Given the above, we propose an ontology-based model to automate the SPL application engineering for intelligent tutoring systems. Specifically, it is intended to automate the processes of customization, instantiation and deployment of an ITS of a SPL, making the manipulation of ontologies and software artifacts transparent to the user. The proposed model uses ontologies to represent the features and constraints of a generic SPL, the specific features for ITSs, the decisions of the author in terms of which features will be part of the ITS to be generated and the information of the student. The SPL application engineering process is performed by components that lead author by the steps of authentication in the system, selection of the SPL to be instantiated, customization/configuration of the features of ITS, validation, generation and deployment of ITS on a Web server. In order to validate the proposed model has been built a tool that automates the generation of products in a SPL. This tool was used on a case study involving the application engineering of an ITS from a SPL. The obtained results showed to be adequate singled out as major contributions, the design and development of a semantic model for the SPL application engineering for ITSs, this model guides the author through the process making transparent the use of ontologies and SPL, helps reduce complexity and effort (i.e., workload) in the construction of ITSs from semantic SPL, reduces the skills required to instantiate ITSs what can enable more people to perform this task and allows properly validate the configuration of the features of the ITS to be instantiated, allowing only products without misconfigurations can be generated. / Sistemas Tutores Inteligentes (STIs) são softwares que buscam representar o comportamento humano inerente ao processo de ensino em algum domínio específico com o objetivo de dar suporte às atividades de professores e oferecer um ensino adaptado aos estudantes. Os STIs têm grande potencialidade tanto no ensino presencial quanto a distância, entretanto, sua construção é uma tarefa complexa e dispendiosa que demanda a presença de profissionais especializados em computação e domínio do sistema sendo agravado quando se necessita construir STIs em larga escala e adaptados a cada domínio. Nesse sentido, a abordagem de Linhas de Produto de Software (LPS) possibilita construir STIs em larga escala. De forma complementar, ontologias podem ser utilizadas para permitir que tal construção seja automaticamente adaptável para diferentes domínios. Porém, na fase de engenharia da aplicação da LPS, momento em que os STIs são instanciados, faz-se necessária a manipulação de ontologias e artefatos de software distintos, demandando a presença de profissionais com conhecimentos em ontologia e engenharia de software, o que dificulta a realização dessa tarefa por autores/projetistas. Com base no exposto propõe-se neste trabalho um modelo baseado em ontologia para automatizar a engenharia de aplicação de LPS para sistemas tutores inteligentes. De maneira específica pretende-se automatizar o processo de customização, instanciação e implantação de STIs de uma LPS, tornando transparente ao usuário a manipulação de ontologias e artefatos de software. O modelo proposto utiliza ontologias para representar as funcionalidades e restrições de uma LPS genérica, as funcionalidades específicas para STIs, as decisões do autor em termos de quais funcionalidades farão parte do STI a ser gerado e as informações do aluno. O processo de engenharia de aplicação da LPS é realizado por componentes que conduzem o autor pelas etapas de autenticação no sistema, seleção da LPS a ser instanciada, customização/configuração das funcionalidades do STI, validação, geração e implantação de um STI em um servidor Web. Para validar o modelo proposto foi construída uma ferramenta que automatiza a geração de produtos em uma LPS. Tal ferramenta foi utilizada em um estudo de caso abrangendo a engenharia de aplicação de um STI a partir de uma LPS. Os resultados obtidos mostram-se adequados apontando como principais contribuições a concepção e desenvolvimento de um modelo semântico para a engenharia de aplicação de LPS para STIs, este modelo guia o autor pelo processo tornando transparente o uso de ontologias e LPS, auxilia na redução da complexidade e do esforço empregado (i.e., carga de trabalho) na construção de STIs a partir de LPS semântica, reduz as qualificações exigidas para instanciar STIs ao qual pode possibilitar que mais pessoas realizem essa tarefa e permite validar corretamente a configuração das funcionalidades do STI a ser instanciado de forma que apenas produtos sem erros de configuração sejam gerados.

Depuis quelques années, le domaine de la santé numérique est en pleine expansion tant sur les volets de la prévention, des soins ou de l’éducation à la santé. L’asthme, première maladie chronique chez l’enfant, a des conséquences médicales graves (hospitalisations, décès). Actuellement, un des enjeux est de fournir de l’éducation thérapeutique personnalisée dans le but d’améliorer les compétences des enfants en terme d’autogestion de la maladie, et par effet rebond d’améliorer l’évolution de la maladie. Dans ce contexte, l’objectif général de la thèse était de développer et de valider un outil numérique d’éducation à la santé ciblant de jeunes enfants avec asthme chronique, et ce en collaboration étroite avec le pôle de recherche et développement de la société ItWell. De là, trois volets d’études expérimentales ont été menés. - Volet « Conception » : dans une première étude (étude 1), les méthodes centrées utilisateur de Conception Participative ont été formalisées pour l’élaboration de serious games ciblant l’éducation thérapeutique d’enfants atteints de maladies chroniques. De là, KidBreath a été conçu pour proposer un programme d’éducation thérapeutique ludique centré sur l’asthme avec pour public cible des jeunes enfants. Celui-ci a fait l’objet d’évaluations auprès d’une population « contrôle » étayant son efficacité pédagogique (connaissance sur l’asthme) et son efficience pédagogique (type de motivation suscitée, expérience utilisateur, parcours d’apprentissage dans les modules de KidBreath). - Volet « Validation » : dans une seconde étude (étude 2a), KidBreath a été déployé pendant deux mois auprès d’enfants asthmatiques en condition réelle d’usage et dans un environnement non contrôlé, et ce avec des évaluations à court (1 mois) et à moyen (2 mois) termes. Les principaux résultats ont confirmé l’efficacité et l’efficience pédagogique auprès du public cible, avec un maintien à 2 mois. Néanmoins, l’amélioration des connaissances liées à l’asthme par KidBreath n’a pas permis de modifier les perceptions des enfants sur l’autogestion de la maladie. - Volet « Optimisation » : dans une troisième étude (étude 2b), l’automatisation d’un parcours d’apprentissage personnalisé a été implémentée et évaluée dans les mêmes conditions que l’étude précédente. L’efficience pédagogique de KidBreath a été à nouveau confirmée mais l’amélioration des connaissances liées à l’asthme par KidBreath n’a pas atteint la significativité. De manière intéressante, le niveau atteint de connaissance avec ou sans l’algorithme de personnalisation était similaire alors que le nombre de contenu et le temps consacré dans la condition « avec algorithme de personnalisation » étaient bien inférieurs à ceux observés dans la condition « sans algorithme ». La méthodologie de l’impact des algorithmes a également été utilisée pour les mathématiques (projet Kidlearn), où les résultats ont montré son efficacité pour personnaliser et adapter les contenus pédagogiques selon la progression de l’élève. Ces travaux ont permis d’extraire un modèle général de conception centrée utilisateur dédié aux serious games à visée thérapeutique détaillant la conception des contenus et formats pédagogiques, et les critères de jugement pour l’évaluation. Les études empiriques (études 1, 2) révèlent l’importance des relations entre efficacité pédagogique (connaissance de la maladie) et efficience pédagogique (motivation intrinsèque générée, expérience utilisateur, respect des parcours individualisés d’apprentissage) sans pour autant impacter à court terme les perceptions de la maladie chez l'enfant. L’ensemble des résultats est discuté à la lumière des différents facteurs (durée d’intervention, âge des participants, sévérité de l’asthme, etc) pouvant influencer la transformation connaissance-métaconnaissance. A cela, des perspectives de recherche sont proposées en regard des opportunités offertes par les Systèmes Tutoriels Intelligents (STI) pour la personnalisation des apprentissages. / In recent years, digital health is an expanding domain in both prevention, care as well as health education. Asthma is the first chronic disease in children and has severe medical outcomes (hospitalization, death). One of the key issues is the personalization of therapeutic education to improve disease self-management skills and as far outcome to improve the disease evolution. In this context, the general purpose of the thesis aimed to develop and to assess a serious game for health education of children with asthma in collaboration with the ItWell Company. To this end, experimental studies have been divided into three parts. - “Conception” part: in a first study (study 1), the user-centered methods of Participatory Design have been formalized to develop a design framework for therapeutic education-related serious games targeting children with chronic disease. From this, KidBreath has been designed to propose an edutainment therapeutic educational program for children suffering from asthma. The device has been successfully assessed amongst control children (without asthma) in terms of its pedagogical efficacy (asthma knowledge) and pedagogical efficiency (type of motivation elicited, user experience, and learning path in different contents). - “Validation” part: in a second study (study 2a), KidBreath has been used by children with asthma over a two-month period in a real-life setting. The main results supported pedagogical efficacy and efficiency of KidBreath for target audience at short-term (1 month) and their maintaining at middle-term (2 month). However, the increase of asthma knowledge after using KidBreath failed to modify children perceptions about disease self-management. - “Optimization” part: in a third study (study 2b), the implementation of a system that automates the personalization of learning path was assessed in similar condition than the previous one. The KidBreath pedagogical efficiency was also demonstrated without a significant increase of asthma knowledge. Despite this, the learning level achieved with the personalization algorithm was similar than that obtained without the algorithm. Moreover, the number of contents and the time spent in the personalized condition were diminished compared to the non-personalized condition. The whole of works allowed to extract a general framework of user-centered design for therapeutic serious games which details both instructional content and media formats, as well as judgment criteria for the assessment. Empirical studies (study 1 & 2) revealed the strength of relationships between pedagogical efficacy (disease knowledge) and pedagogical efficiency (intrinsic motivation elicited, user experience, child’s learning path) without short-term impact on the children illness perception. Overall results were discussed in light of different factors, which can bias the transfer from disease knowledge to meta-knowledge (intervention period, age of children, asthma severity etc.). Finally, research perspectives are proposed according to opportunities provided by the Intelligent Tutorial System (ITS) for personalization of learning paths. In particular, a new project named Kidlearn is presented. It is methodologically inspired from the KidBreath project and aimed the algorithm assessment for learning mathematics at primary school level.

Esta tese situa-se na área de IA (Inteligência Artificial) aplicada à educação incluindo características interdisciplinares tanto da própria IA como de IE (informática na Educação). Faz-se também necessário constarem, aspectos referentes à Ciência da Computação e Educação a fim de melhor situar a complexidade e a dimensão do trabalho desenvolvido. A utilização de técnicas de IA na elaboração e no desenvolvimento de ambientes de ensino-aprendizagem computadorizados tem se constituído em objeto de investigação por parte dos pesquisadores da área de Informática aplicada à Educação, devido as suas potencialidades. A utilização de agentes na modelagem e no projeto de STI permite-nos resgatar antigos problemas em aberto, como por exemplo a melhoria da interação entre tutor e aluno e a possibilidade de investigação dos processos mentais em nível mais estratificado. A arquitetura descrita nesta tese utiliza a metodologia que vem sendo aplicada ao projeto de STI, onde são contempladas diferentes formas de se trabalhar com um determinado conhecimento (estratégias de ensino e táticas associadas), levando-se em consideração o tipo de usuário que está interagindo com o sistema. A arquitetura, elaborada segundo uma abordagem construtivista, prevê que o tutor seja menos diretivo e menos controlador das ações do aluno. O controle é feito na forma de monitoração para que o tutor funcione como um parceiro, ou seja, como facilitador do trabalho do aluno. Contudo, devido às características da modalidade escolhida para construção do protótipo (jogo educacional), precisamos ter algumas atitudes no tutor que garantam que o sistema não entre em colapso. O que inviabilizaria o trabalho do aluno. Nestas situações críticas, o tutor vai se comportar de maneira mais diretiva. Cabe salientar que a abordagem construtivista não significa dar liberdade total ao aluno nem privá-lo de qualquer tipo de assistência. Portanto, o que deve ser destacado é o grau de interferência do tutor, i.e., o quanto ele interfere no trabalho do aluno e se ele permite ou não que o aluno siga um caminho alternativo àquele que ele tem como o ideal para resolver o problema (heurísticas do tutor sobre o problema e forma de solução). Além destes aspectos educacionais inerentes a todo o projeto de software educacional (necessários num trabalho desta natureza), esta tese está inserida no contexto da pesquisa em agentes cognitivos modelados através de seus estados mentais. É importante salientar que os estados mentais utilizados neste trabalho (crenças, desejos, intenções e expectativas) funcionam como uma metáfora dos estudos mentais humanos. Por exemplo, quando se coloca a crença de um aluno a respeito de "lago", na realidade está se colocando a crença que temos a respeito da crença que o aluno possui a respeito de "lago". O mesmo acontece com os outros estados mentais aqui utilizados. Os diálogos reais foram registrados através de observação direta e posteriormente analisados a fim de se identificar os estados mentais relacionados. Perguntas adicionais foram feitas no sentido de obterem-se mais elementos para auxiliar na inferência do conjunto de estados mentais que o aluno possui naquele momento em que estava jogando. Após a observação de vários alunos jogando, identificou-se um certo padrão nas suas atitudes quando executavam uma ação. Observações sucessivas permitiram delinear o conjunto de estados mentais associados à ação do aluno. Tal conjunto foi utilizado como base para elaboração da coreografia. Estes dados servem de entrada para a construção do modelo do aluno mediante a interação com o tutor. No presente trabalho, nós apresentamos a modelagem de um STI através do uso da tecnologia de agentes utilizando a arquitetura de SMA (Sistemas Multiagentes). O STI é concebido como um SMA híbrido composto por um ambiente reativo (SMAR - Sistema Multiagente Reativo) e um "kernel" cognitivo (SMAC - Sistema Multiagente Cognitivo). O SMAR e o SMAC interagem entre si de para ampliar as informações quantitativas e qualitativas oferecidas aos alunos que utilizam o sistema. Estas informações disponíveis é que irão permitir ao tutor selecionar estratégias de ensino mais adequadas a um determinado tipo de aluno. A principal contribuição desta tese está centrada no "kernel" cognitivo. Nós propomos uma arquitetura para o tutor que permitirá a monitoração de dois alunos trabalhando conjuntamente. Além disso, propomos uma forma de selecionar o comportamento do tutor para oferecer auxílio personalizado aos alunos considerando o perfil de cada um. Esta arquitetura pretende ser uma alternativa de solução para uma questão importante na área de STI: Como o tutor pode selecionar, entre várias estratégias de ensino, a mais adequada para cada perfil de aluno? O grupo de pesquisa no qual este trabalho está inserido (GIA/UFRGS, sob orientação da Prof.a. Rosa Maria Viccari) tem realizado algumas avaliações experimentais, usando STI tradicionais e STI projetados e modelados através de sistemas multiagentes As contribuições científicas listadas no texto deste trabalho possibilitaram que o grupo avançasse sua pesquisa na abordagem mentalística através da criação de uma arquitetura para o tutor e favorecesse a integração do trabalho desenvolvido por Móra et al. [MOR97; MOR98]. A utilização do modelo computacional de agentes criado por Móra et al. gerou a implementação do "kernel" cognitivo. Os desafios inerentes a implementação da arquitetura proposta para o tutor ampliaram as características do ambiente criado por Móra et al. e favorecem a junção de dois trabalhos de tese supervisionados pela mesma orientadora [MOR99]. Portanto, a nova arquitetura proporcionou ganhos tanto para tais pesquisas, como para o avanço das pesquisas desenvolvidas pelo nosso grupo. / The present thesis has been elaborated within the AI (Artificial Intelligence) applied to Education realm, and it brings specific contributions to the STI (Intelligent Tutoring System) area. The use of AI techniques has been investigated by researchers of Computer Science applied to Education, due to its potentialities to improve educational systems. The agents' techniques used in the design of STI allow us to solve old problems opened in the area. For instance, the improvement of the interaction between tutor and student, and the possibility of tracing the mental processes in a more stratified way. The architecture described in this thesis uses the methodology applied to the modern STI projects: multiple strategies for the tutor (i.e., teaching strategies and associated tactics). This approach considers different forms of working with a certain piece of knowledge, and is taken into consideration to create the user profile, as well as to monitor the student interaction with the system. The architecture, designed according to a constructivist approach, expects the tutor to be less directive, and less controller of the student's actions. The control is made by an observation of students' actions by the tutor. The tutor works either as a student's partner or as a facilitator. However, due to the characteristics of the modality chosen for construction of the prototype (educational game), we needed to take some attitudes in the tutor in order to avoid the system to collapse. What would make unfeasible the student's work. In these critical situations, the tutor will behave in more directive way. It fits to point out that the construtivist approach does not mean to give total freedom to the student or to deprive it of any kind of attendance. Therefore, what should be note here is the degree of the tutor's interference, i.e., how it interferes with the student's work using its own set of heuristics. Besides these educational aspects, inherent to a project of Educational software, this thesis is inserted in the context of the research in cognitive agents modelled through their mental states (Believe, Desire, Intention, and Expectation). It is important to point out that the mental states used in this work are as a metaphor of the human mental states. For example, when the student's has a believe regarding "lake", in fact we area talking about belief that we have regarding the belief that the student possesses regarding about "lake". The same happens here with the other mental states used. The real dialogues were registered through direct observation in real situation (students playing with the game). They were analysed in order to identify the mental states connected with the student's actions. Additional questions were asked to obtain more elements to aid us to inference the group of mental states possessed by the student when he/she was playing. Successive observations allowed us to delineate the group of mental states associated to the student's action. Such group was used as a base for the choreography. These data were used as input for the construction to the student's model during the interaction with the tutor. In this work, we have presented the design of an STI with the use of architecture of MAS (Multi-agent System Architecture). The ITS is conceived as a hybrid MAS composed by a RMAS (Reactive Multi-agent System) and a "cognitive kernel" using the CMAS (Cognitive Multi-agent System). The RMAS and the CMAS interact with each other to enlarge the quantitative and qualitative information offered to the students that uses the system. These available information allow the tutor to select teaching strategies more adapted to a certain student type. The main contribution of this thesis is centred in the "cognitive kernel". We propose an architecture for the tutor that will allow the two students to work together. Besides, we propose a way to select the tutor behaviour in order to aid the students considering their personal profile. This architecture intends to be an alternative solution for an important question in ITS research: How can the tutor select, among several teaching strategies, the one that is more suitable for each student profile? The research group in which this work is placed (GIA/UFRGS - Artificial Intelligence Group), under the supervision of Profa. Rosa Maria Viccari, has been doing executing some experimental evaluations, using traditional ITS and ITS modelled through multi-agents systems techniques. The scientific contributions listed in this work allow the group to achive some interesting results in the research of STI using a mental approach. The tutor architecture favoured the integration of the X-BDI (eXecutable Belief Desire, and Intention model)) developed by Móra et al. The use the X-BDI allowed us to implement the "cognitive kernel". The inherent challenges posed by the implementation of the architecture of the tutor refined the XBDI environment. It favours the junction of two thesis works supervised done under the same supervisor[MOR99]. Therefore, the new architecture provided many gains for both researches, as well as for the progress of the research developed by our group.

Trust is one of the attitudes that can affect the intentions and the behavior of a human using a system. Misusing a system can have safety- as well as economic consequences, this is why it is important that a user develops calibrated trust towards a system. In this report, the research question is: how much trust cross country skiers have towards a physical intelligent tutoring system (PITS)? Six biological males (age 24 to 50) roller skied on a mechanical treadmill and received feedback from the PITS on a TV-screen. Experience of using the PITS was evaluated with an instrument in a semi-structured interview. The instrument measured the participants overall perceived trust (OPT) for the system, and the participants were asked to further their thoughts about the statements. The data was transcribed, coded, and categorized in a thematic analysis. The result showed that a majority of the participants had low OPT for the PITS, and the thematic analysis showed that the minority with higher levels of OPT focused on the choice of an elite skier as the reference skier. One of the problems with the instrument was that it was developed for evaluation in long term usage, and not first time usage as in this study. The result of this report can be used for further development of the PITS and a reminder on why trust needs to be considered when creating user experiences.

abstract: This study empirically evaluated the effectiveness of the instructional design, learning tools, and role of the teacher in three versions of a semester-long, high-school remedial Algebra I course to determine what impact self-regulated learning skills and learning pattern training have on students' self-regulation, math achievement, and motivation. The 1st version was a business-as-usual traditional classroom teaching mathematics with direct instruction. The 2rd version of the course provided students with self-paced, individualized Algebra instruction with a web-based, intelligent tutor. The 3rd version of the course coupled self-paced, individualized instruction on the web-based, intelligent Algebra tutor coupled with a series of e-learning modules on self-regulated learning knowledge and skills that were distributed throughout the semester. A quasi-experimental, mixed methods evaluation design was used by assigning pre-registered, high-school remedial Algebra I class periods made up of an approximately equal number of students to one of the three study conditions or course versions: (a) the control course design, (b) web-based, intelligent tutor only course design, and (c) web-based, intelligent tutor + SRL e-learning modules course design. While no statistically significant differences on SRL skills, math achievement or motivation were found between the three conditions, effect-size estimates provide suggestive evidence that using the SRL e-learning modules based on ARCS motivation model (Keller, 2010) and Let Me Learn learning pattern instruction (Dawkins, Kottkamp, & Johnston, 2010) may help students regulate their learning and improve their study skills while using a web-based, intelligent Algebra tutor as evidenced by positive impacts on math achievement, motivation, and self-regulated learning skills. The study also explored predictive analyses using multiple regression and found that predictive models based on independent variables aligned to student demographics, learning mastery skills, and ARCS motivational factors are helpful in defining how to further refine course design and design learning evaluations that measure achievement, motivation, and self-regulated learning in web-based learning environments, including intelligent tutoring systems. / Dissertation/Thesis / Ph.D. Educational Technology 2013

abstract: Students' ability to regulate and control their behaviors during learning has been shown to be a critical skill for academic success. However, researchers often struggle with ways to capture the nuances of this ability, often solely relying on self-report measures. This thesis proposal employs a novel approach to investigating variations in students' ability to self-regulate by using process data from the game-based Intelligent Tutoring System (ITS) iSTART-ME. This approach affords a nuanced examination of how students' regulate their interactions with game-based features at both a coarse-grained and fine-grain levels and the ultimate impact that those behaviors have on in-system performance and learning outcomes (i.e., self-explanation quality). This thesis is comprised of two submitted manuscripts that examined how a group of 40 high school students chose to engage with game-based features and how those interactions influenced their target skill performance. Findings suggest that in-system log data has the potential to provide stealth assessments of students' self-regulation while learning. / Dissertation/Thesis / M.A. Psychology 2014