Using electronic portfolios (e-portfolios) to assist learning is an important component of future educational models. A portfolio is a purposeful collection of student work that exhibits the student's efforts, progress and achievements in one or more areas. An e-portfolio contains a variety of information about a person's learning outcomes, such as artifacts, assertions from others, self-reflective information and presentation for different purposes. E-portfolios become sources of evidence for claims about prior conceptual knowledge or skills. This thesis investigates using the information contained in e-portfolios to initialize the learner model for an intelligent tutoring system. We examine the information model from the e-portfolio standardized specification and present a method that may assist users in initializing learner models using e-portfolios as evidence for claims about prior conceptual knowledge or skills. We developed the EP-LM system for testing how accurately a learner model can be built and how beneficial this approach can be for reflective and personalized learning. Experimental results are presented aiming at testing whether accurate learner models can be created through this approach and whether learners can gain benefits in reflective and personalized learning. Monitoring this process can also help ITS developers and experts identify how an initial learner model can automatically arise from an e-portfolio. Additionally, a well-structured learner model, generated by an intelligent tutoring system also can be attached to an e-portfolio for further use by the owner and others.

Programming is one of the core skills required by Computer Science undergraduates in tertiary institutions worldwide, whether for study itself, or to be used as a tool to explore other relevant areas. Unfortunately, programming can be incredibly difficult; this is for several reasons, including the youth, depth, and variety of the field, as well as the youth of the technology that frames it. It can be especially problematic for computing neophytes, with some students repeating programming courses not due to academic laziness, but due to an inability to grasp the core concepts. The research outlined by this thesis focuses on our proposed solution to this problem, a constraint-based intelligent tutoring system for teaching the Java programming language, named J-LATTE. J-LATTE (Java Language Acquisition Tile Tutoring Environment) is designed to solve this problem by providing a problem-solving environment for students to work through programming problems. This environment is unique in that it partitions interaction into a concept mode and a coding mode. Concept mode allows the student to form solutions using high-level Java concepts (in the form of tiles), and coding mode allows the student to enter Java code into these tiles to form a complete Java program. The student can, at any time, ask for feedback on a solution or partial solution that they have formed. A pilot study and two full evaluations were carried out to test the effectiveness of the system. The pilot study was run with an assignment given to a postgraduate Computer Science course, and because of the advanced knowledge level of the students, it was not designed to test teaching effectiveness, but instead was useful in determining usability issues and identifying any software errors. The full evaluations of the system were designed to give insight into the teaching effectiveness of J-LATTE, by comparing the results of using the system against a simulated classroom situation. Unfortunately, the participant base was small, for several reasons that are explained in the thesis. However, the results prove interesting otherwise and for the most part are positive towards the effectiveness of J-LATTE. The participants’ knowledge did improve while interacting with the system, and the subjective data collected shows that students like the interaction style and value the feedback obtained.

Emotions play a significant role in many human mental activities, including decision-making, motivation, and cognition. Various intelligent and expert systems can be empowered with emotionally intelligent capabilities, especially systems that interact with humans and mimic human behaviour. However, most current methods in affect recognition studies use intrusive, lab-based, and expensive tools which are unsuitable for real-world situations. Inspired by studies on keystrokes dynamics, this thesis investigates the effectiveness of diagnosing users’ affect through their typing behaviour in an educational context. To collect users’ typing patterns, a field study was conducted in which subjects used a dialogue-based tutoring system built by the researcher. Eighteen dialogue features associated with subjective and objective ratings for users’ emotions were collected. Several classification techniques were assessed in diagnosing users’ affect, including discrimination analysis, Bayesian analysis, decision trees, and neural networks. An artificial neural network approach was ultimately chosen as it yielded the highest accuracy compared with the other methods. To lower the error rate, a hierarchical classification was implemented to first classify user emotions based on their valence (positive or negative) and then perform a finer classification step to determining which emotions the user experienced (delighted, neutral, confused, bored, and frustrated). The hierarchical classifier was successfully able to diagnose users' emotional valence, while it was moderately able to classify users’ emotional states. The overall accuracy obtained from the hierarchical classifier significantly outperformed previous dialogue-based approaches and in line with some affective computing methods.

Emotions play a significant role in many human mental activities, including decision-making, motivation, and cognition. Various intelligent and expert systems can be empowered with emotionally intelligent capabilities, especially systems that interact with humans and mimic human behaviour. However, most current methods in affect recognition studies use intrusive, lab-based, and expensive tools which are unsuitable for real-world situations. Inspired by studies on keystrokes dynamics, this thesis investigates the effectiveness of diagnosing users’ affect through their typing behaviour in an educational context. To collect users’ typing patterns, a field study was conducted in which subjects used a dialogue-based tutoring system built by the researcher. Eighteen dialogue features associated with subjective and objective ratings for users’ emotions were collected. Several classification techniques were assessed in diagnosing users’ affect, including discrimination analysis, Bayesian analysis, decision trees, and neural networks. An artificial neural network approach was ultimately chosen as it yielded the highest accuracy compared with the other methods. To lower the error rate, a hierarchical classification was implemented to first classify user emotions based on their valence (positive or negative) and then perform a finer classification step to determining which emotions the user experienced (delighted, neutral, confused, bored, and frustrated). The hierarchical classifier was successfully able to diagnose users' emotional valence, while it was moderately able to classify users’ emotional states. The overall accuracy obtained from the hierarchical classifier significantly outperformed previous dialogue-based approaches and in line with some affective computing methods.

In a technology driven world, efforts are being made to make education/learning available to individuals at any time with no compromise in the quality of teaching/training. To make learning flexible, different techniques such as distributing learning material, uploading audio lectures on the web, and creating intelligent tutoring systems (ITS) are being used. The technique considered here is an adaptive ITS, a system that replicates the learning that occurs in a student teacher relationship. This thesis develops an adaptive intelligent tutoring system architecture prototype where the addition, modification and removal of educational material are relatively easy. The resulting software will take into account: the goals of the specific educational experience, the concepts to be covered, the preferred learning style of the student, measures to detect misuse of the system functionality, behavior based on the student's performance and the generation of hint sequences and feedback messages to improve learning gain. The system will accomplish these objectives by assessing the student's prior knowledge level, observing the actions performed by the student and by adapting to his/her learning abilities. The ITS will attempt to be more intelligent by performing some actions traditionally done by a human teacher - such as diagnosing misconceptions, identifying the most suitable learning style, stressing concepts that the student is finding difficult to understand, switching back to the learning material if the student shows no improvement after a set of trials. The system makes sure that the student is getting feedback where appropriate. Using this prototype system, the student will be tutored to acquire declarative knowledge. A problem based learning (PBL) approach will be used to strengthen the acquired knowledge by providing a high degree of personal attention to the student. To show how the prototype system works, an example of analysis of a control system problem using bode plot technique will be used to assist the student in using the technique to perform the stability analysis of an analog, linear, time-invariant control system problems and to recommend a controller to attain stability (if the system is not stable). Ideas of porting the system from standalone to web-based architecture and features required for collaborative learning will be discussed and an architecture for a web-based tutoring system for supporting multiple students enabling communication between students and sharing data among them will be proposed. / Electrical and Computer Engineering

Intelligent Tutoring Systems (ITS) have been proven to be efficient in providing students assistance and assessing their performance when they do their homework. Many research projects have been done to analyze how students' knowledge grows and to predict their performance from within intelligent tutoring system. Most of them focus on using correctness of the previous question or the number of hints and attempts students need to predict their future performance, but ignore how they ask for hints and make attempts. In this research work, we build a Sequence of Actions (SOA) model taking advantage of the sequence of hints and attempts a student needed for previous question to predict students' performance. A two step modeling methodology is put forward in the work, which is a combination of Tabling method and the Logistic Regression. We used an ASSISTments dataset of 66 students answering a total of 34,973 problems generated from 5010 questions over the course of two years. The experimental results showed that the Sequence of Action model has reliable predictive accuracy than Knowledge Tracing and Assistance Model and its performance of prediction is improved after combining with Knowledge Tracing.

Predicting student performance is an important part of the student modeling task in Intelligent Tutoring System (ITS). The state-of-art model for predicting student performance - Bayesian Knowledge Tracing (KT) has many critical limitations. One specific limitation is that KT has no underlying mechanism for memory decay represented in the model, which means that no forgetting is happening in the learning process. In addition we notice that numerous modification to the KT model have been proposed and evaluated, however many of these are often based on a combination of intuition and experience in the domain, leading to models without performance improvement. Moreover, KT is computationally expensive, model fitting procedures can take hours or days to run on large datasets. The goal of this research work is to improve the accuracy of student performance prediction by incorporating the memory decay factor which the standard Bayesian Knowledge Tracing had ignored. We also propose a completely data driven and inexpensive approach to model improvement. This alternative allows for researchers to evaluate which aspects of a model are most likely to result in model performance improvements based purely on the dataset features that are computed from ITS system logs.

Student gaming behavior in intelligent tutoring systems (ITS) has been correlated with lower learning rates. The goal of this work is to identify such behavior, produce interventions to discourage this behavior, and by doing so hopefully improve the learning rate of students who would normally display gaming behavior. Detectors have been built to identify gaming behavior. Interventions have been designed to discourage the behavior and their evaluation is discussed.

Thesis (Ph. D.)--Open University. BLDSC no. DX174754.

Thesis (M.S.)--Worcester Polytechnic Institute. / Keywords: intelligent tutoring systems; its; authoring tool; artificial intelligence. Includes bibliographical references (p. 81-82).