Intelligent Tutoring Systems

Evermore widespread “flipped classroom” learning model is associated with increased independence of learning. The problem is the independence of learning math by students with visual impairments, especially the blind. Mathematical content includes spatial objects such as formulas and graphics, inaccessible to blind students and hardly accessible to low vision students. They prevent independent learning. The article presents a method that increases students’ independence in recognising mathematical content in textbooks and worksheets. The method consists in introducing into the document elements of Augmented Reality (AR), that is texts and sounds extending information about the mathematical objects encountered in the content, beyond the information provided by WCAG guidelines and recommendations of the WAI-ARIA standard under development by the W3C consortium. Access to AR elements is gained through multi-sensual User Interface - hearing, the touch of a braille display, touch screen and touch gestures. The method was developed in cooperation with students with visual impairment and math teachers. It is currently undergoing valorisation in Poland, the Netherlands and Ireland.

The article presents a methodological approach thought to monitor, and assess adaptive learning impact on teaching and learning French to a diversity of pupils. First, the development process of the web platform is detailed, including the didactic foundation. Second, the single-case experimental design elaborated for the testing phase is presented. Four cohorts of pupils with various profiles – including special needs – will participate. So, the assessment of the adaptive learning potential to respond to every need will be systematically documented.

ITSs are developing into more and more complex systems. The classic components of the general purpose ITS software architecture are used widespread but its focus is on the databases and the user interface. To reflect the functionalities and complexity needed to accomplish an ITS of today’s standards a new software architecture was developed which emphasis the definition of functionality components and component connections without ignoring the classical components. A 5 layer architecture with a detailed separation into components but still abstract definition of an ITS is being shown in UML. Upcoming general ideas of important functionality is included and split up into more abstract packages.

In the paper we address the research question of whether non-experienced and untrained end-users may efficiently design and customise their own interaction experiences as part of a tutoring system. To this end, we present the main end-user development (EUD) approaches highlighting user engagement, usability and user experience (UX) design principles. We examine aspects related to the triptych of efficiency, efficacy, satisfaction of all user development and learning activities. To this direction, we offer suggestions on how such a EUD approach for an Intelligent Tutoring System (ITS) setting can form part of a maker space for providing sustainable and hands-on learning experiences.

This paper presents a multimodal affect assessment protocol developed for aviation training, which consists of physiological, behavioral measures of affect and subjective self-report of affective correlates. Data convergence is examined by comparing physiological and behavioral data output with self-report variables. We found significant correlations between arousal inferred from electro-dermal activity (EDA) and self-reported workload, fatigue and effort. We also found that the intensities of emotions inferred from facial expression correlate with self-reported variables. These findings support the validity of EDA and facial expression as measures of affect in aviation training context.

Mentoring is the activity when an experienced person (the mentor) supports a less knowledgeable person (the mentee), in order to achieve the learning goal. In a perfect world, the mentor would be always available when the mentee needs it. However, in the real world higher education institutions work with limited resources. For this, we need to carefully design socio-technical infrastructures for scaling mentoring processes with the help of distributed artificial intelligence. Our approach allows universities to quickly set up a necessary data processing environment to support both mentors and mentees. The presented framework is based on open source standards and technologies. This will help leveraging the approach, despite the organizational and pedagogical challenges. The deployed infrastructure is already used by several universities.

This paper presents for the first time a detailed analysis of fine-grained navigation style identification in MOOCs backed by a large number of active learners. The result shows 1) whilst the sequential style is clearly in evidence, the global style is less prominent; 2) the majority of the learners do not belong to either category; 3) navigation styles are not as stable as believed in the literature; and 4) learners can, and do, swap between navigation styles with detrimental effects. The approach is promising, as it provides insight into online learners’ temporal engagement, as well as a tool to identify vulnerable learners, which potentially benefit personalised interventions (from teachers or automatic help) in Intelligent Tutoring Systems (ITS).

Many works in the literature show that positive emotions improve learning. However, in the educational context, the affective dimension is often not adopted in the teaching-learning process. One of them is that there are many students for a teacher, making the practice of adapting the didactics and individualized feedbacks practically impossible. The low or sometimes no emotion analysis of those involved in learning also becomes a obstacle. One possibility to circumvent this problem is the use of Intelligent Tutoring Systems (ITS), to understand the student individually and adapt environments according to their use. It also adds the theories of emotions so that the ITS can understand the affective dimension of the student during activities. This paper aims to present a way to infer changes in a student’s affective states to improve feedbacks in ITS For this, facial expressions and brain waves (using a low-cost equipment called openBCI) were studied for acquisition and emotions. In the initial tests, the methodology has met what was expected, however, more studies with experiments must be carried out.

Our research work focuses on computer-aided grouping of students based on questions answered in an assessment for effective reading intervention in early education. The work can facilitate placement of students with similar reading disabilities in the same intervention group to optimize corrective actions. We collected ELA (English Language Arts) assessment data from two different schools in USA, involving 365 students. Each student performed three mock assessments. We formulated the problem as a matching problem—an assessment should be matched to other assessments performed by the same student in the feature space. In this paper, we present a study on a number of matching schemes with low-level features gauging the grade-level readability of a piece of writing. The matching criterion for assessments is the consistency measure of matched questions based on the students’ answers of the questions. An assessment is matched to other assessments using K-Nearest-Neighbor. The best result is achieved by the matching scheme that considers the best match for each question, and the success rate is 17.6%, for a highly imbalanced data of only about 5% belonging to the true class.

This study applies the state of the art in explainable AI techniques to shed light on the automated essay scoring (AES) process. By means of linear regression and Shapley values, SHAP (Shapley Additive Explanations) approximates a complex AES predictive model implemented as a deep neural network and an ensemble regression. This study delves into the essentials of the automated assessment of ‘organization’, a key rubric in writing. Specifically, it explores whether the organization and connections between ideas and/or events are clear and logically sequenced. Built on findings from previous work, this paper, in addition to improving the generalizability and interpretability of the AES model, highlights the means to identify important ‘writing features’ (both global and local) and hint at the best ranges of feature values. By associating ‘organization’ with ‘writing features’, it provides a mechanism to hypothesize causal relationships among variables and shape machine-learned formative feedback in human-friendly terms for the consumption of teachers and students. Finally, it offers an in-depth discussion on linguistic aspects implied by the findings.

A blocking state is a measurable state on an intelligent tutoring systems’ user interface, which mirrors a student’s cognitive state where she/he cannot temporarily make any progress toward finding a solution to a problem. In this paper, we present the development of four probabilistic models to detect a blocking state of students while they are solving a Canadian high school-level problem in Euclidean geometry on an ITS. Our methodology includes experimentation with a modified version of QED-Tutrix, an ITS, which we used to gather labelled datasets composed of sequences of mouse and keyboard actions. We developed four predicting models: an action-frequency model, a subsequence-detection model, a 1D convolutional neural network model and a hybrid model. The hybrid model outperforms the others with a $$F_1$$ score of 80.4% on the classification of blocking state on validation set while performing 77.3% on the test set.

Intrinsic motivation is the psychological construct that defines our reasons and interests to perform a set of actions. It has shown to be associated with positive outcomes across domains, especially in the academic context. Therefore, understanding and identifying peoples’ levels of intrinsic motivation can be crucial for professionals of many domains, e.g. teachers aiming to offer better support to students’ learning processes and enhance their academic outcomes. In a first attempt to tackle this issue, we propose an end-to-end approach for recognition of intrinsic motivation, using only facial expressions as input. Our results show that visual cues from students’ facial expressions are an important source of information to detect their levels of intrinsic motivation (AUC $$=0.570$$, $$F_1=0.556$$). We also show how to avoid potential bias that might be present in datasets. When dividing the training samples per gender, we achieved a substantial improvement for both genders (AUC $$=0.739$$ and $$F_1=0.852$$ for male students, AUC $$=0.721$$ and $$F_1=0.723$$ for female students).

This paper aims to present the design of a low-cost, open-source, robotic platform for use in Educational Robotics and SΤΕΜ as a holistic approach to the curriculum. In alignment with the research presented in [1] the robotic platform’s innovation is based on two axes: (a) its specifications came from the 1st cycle of participatory action research; (b) it is equipped with a visual programming language integrated into the robot’s “brain” itself so that it can be programmed by any device (smartphone, tablet, PC) with Wi-Fi connectivity, without the need for any software or app to be downloaded and installed in the device. The spark for this research arose from an educational robotics survey’s data evaluation handled at the municipality of Agia Varvara in Athens-Greece which, while showing a strong students interest in educational robotics, however few of them got involved because of the robotic platform’s high cost. So, this research’s motivation was to go on designing and developing a robotic platform suitable for the whole educational community that the specifications based on its members’ needs and extracted by quantitative and qualitative data collection and analysis tools.

An open learner model uses system’s representation of the student to support learning and reveal progress. The model contains information regarding learner’s characteristics such as level of knowledge, interests, involvement and other relevant cognitive aspects. The current paper presents an example of incorporating an open learner model in a peer assessment platform, more specifically LearnEval, and applying it in the context of a project-based learning scenario in a Web Applications Design course. The student is modeled based on several traits such as competence, involvement and assessment abilities. Furthermore, an aggregated overall score offers a general overview of the student capabilities. To incorporate the open learner model, a Scores module was integrated into LearnEval, offering intuitive, friendly and effective visualizations of the scores and a breakdown of the metrics composing them in the form of progress bars, gauges, column bars, trophies and medals. We offer a description of the context where the open learner model was put in practice as well as an example of how a learner could utilize it. An opinion survey regarding the experience with the open learner model was applied to the students at the end of the semester. The findings are encouraging, as the learners found the module easy to use, helpful and comprehensive and they examined it relatively often.

Intelligent tutoring systems become more and more common in assisting human learners. Distinct advantages of intelligent tutoring systems are personalized teaching tailored to each student, on-demand availability not depending on working hour regulations and standardized evaluation not subjective to the experience and biases of human individuals. A virtual driving instructor that supports driver training in a virtual world could conduct on-demand personalized teaching and standardized evaluation. We propose an architectural design of a virtual driving instructor system that can comprehend and explain complex traffic situations. The architecture is based on a multi-agent system capable of reasoning about traffic situations and explaining them at an arbitrary level of detail in real-time. The agents process real-time data to produce instances of concepts and relations in an ever-evolving knowledge graph. The concepts and relations are defined in a traffic situation ontology. Finally, we demonstrate the process of reasoning and generating explanations on an overtake scenario.

[Context] The use of MOOCs has generated an increasing amount of data that, if properly explored, can provide an understanding of how students interact throughout their learning, in addition to identifying strategies that can be used in the process of building environments. that enhance the construction of knowledge in a shared way. [Objective] The objective of this article is to propose a Conceptual Collaborative Framework (MOOColab), based on Learning Analytics mechanisms and Recommendation Systems to improve collaborative learning in the environment. [Methodology] For the development of the Framework, the Design Science Research model was used for the analysis, development and evaluation of MOOColab. [Results] An experiment was carried out with two samples: a control group (which did not use the Framework) and an experimental group (which went through the same course using MOOColab). [Conclusion] From the results obtained in the research, it is evident that the implementation of Framework identifying the individualities of each student with the discovery of behavioral patterns and their respective skills to adapt the environment with the recommendation of peers, in order to improve the mutual exchange of information between students involved in the learning process.

Computerized Assisted Testing (CAT) has supported the development of numerous adaptive testing approaches. Such approach as Item Response Theory (IRT) estimates a student’s competency level by modeling a test as a function of the individual’s knowledge ability, and the parameters of the question (i.e. item). Multidimensional Item Response Theory (MIRT) extends IRT so that each item depends on multiple competency areas (i.e., knowledge dimensions). MIRT models consider two opposing types of relationship between knowledge dimensions: compensatory and noncompensatory. In a compensatory model, having a higher competency with one knowledge dimension compensates for having a lower competence in another dimension. Conversely, in a noncompensatory model all the knowledge dimensions are independent and do not compensate for each other. However, using only one type of relationship at a time restricts the use of MIRT in practice. In this work, we generalize MIRT to a mixed-compensation multidimensional item response theory (MCMIRT) model that incorporates both types of relationships. We also relax the MIRT assumption that each item must include every knowledge dimension. Thus, the MCMIRT can better represent real-world curricula. We show that our approach outperforms random item selection with synthetic data.

Welfare and economic development is directly dependent on the availability of highly skilled and educated individuals in society. In the UK, higher education is accessed by a large percentage of high school graduates (50% in 2017). Still, in Brazil, a limited number of pupils leaving high schools continue their education (up to 20%). Initial pioneering efforts of universities and companies to support pupils from underprivileged backgrounds, to be able to succeed in being accepted by universities include personalised learning solutions. However, initial findings show that typical distance learning problems occur with the pupil population: isolation, demotivation, and lack of engagement. Thus, researchers and companies proposed gamification. However, gamification design is traditionally exclusively based on theory-driven approaches and usually ignore the data itself. This paper takes a different approach, presenting a large-scale study that analysed, statistically and via machine learning (deep and shallow), the first batch of students trained with a Brazilian gamified intelligent learning software (called CamaleOn), to establish, via a grassroots method based on learning analytics, how gamification elements impact on student engagement. The exercise results in a novel proposal for real-time measurement on Massive Open Online Courses (MOOCs), potentially leading to iterative improvements of student support. It also specifically analyses the engagement patterns of an underserved community.

Investigating the factors affecting students’ academic failure in online and/or blended courses by analyzing students’ learning behavior data gathered from Learning Management Systems (LMS) is a challenging area in intelligent learning analytics and education data mining area. It has been argued that the actual course design and the instructor’s intentions is critical to determine which variables meaningfully represent student effort that should be included/excluded from the list of predicting factors. In this paper we describe such an approach for identifying students at risk of failure in online courses. For the proof of our concept we used the data of two cohorts of an online course implemented in Moodle LMS. Using the data of the first cohort we developed a prediction model by experimenting with certain base classifiers available in Weka. To improve the observed performance of the experimented base classifiers, we enhanced further our model with the Majority Voting ensemble classifier. The final model was used at the next cohort of students in order to identify those at risk of failure before the final exam. The prediction accuracy of the model was high which show that the findings of such a process can be generalized.

Identifying users’ demographic characteristics is called Author Profiling task (AP), which is a useful task in providing a robust automatic prediction for different social user aspects, and subsequently supporting decision making on massive information systems. For example, in MOOCs, it used to provide personalised recommendation systems for learners. In this paper, we explore intelligent techniques and strategies for solving the task, and mainly we focus on predicting the employment status of users on a MOOC platform. For this, we compare sequential with parallel ensemble deep learning (DL) architectures. Importantly, we show that our prediction model can achieve high accuracy even though not many stylistic text features that are usually used for the AP task are employed (only tokens of words are used). To address our highly unbalanced data, we compare widely used oversampling method with a generative paraphrasing method. We obtained an average of 96.4% high accuracy for our best method, involving sequential DL with paraphrasing overall, as well as per-individual class (employment statuses of users).

Online collaborative learning environments open new research opportunities, for example, the analysis of learning outcomes, the identification of learning patterns, the prediction of students’ behaviors, and the modeling and visualization of social relations and trends among students. Moodle is an online educational platform which supports both students and teachers, and can be effectively employed to encourage collaborative learning. Moodle is often used to make inquiries on student homework, exams, to request clarifications, and to make announcements. Our goal is to predict student success based on Cohesion Network Analysis (CNA) and to identify interaction patterns between students (n = 71 who had a sufficient level of participation on the forum) and 4 tutors together with 19 teaching assistants in a Romanian Moodle course. CNA visualizations consider a hierarchical clustering that classifies members into central, active, and peripheral groups. Weekly snapshots are generated to better understand students’ evolution throughout the course, while correlating their activities with specific course events (e.g., homework deadlines, tests, holidays, exam, etc.). Several regression models were trained based on the generated CNA indices and the best model achieves a mean average error below .5 points when predicting partial course grades, prior to the final exam, on a 6-point scale.

Discussion is an effective way of improving teaching quality. Face-to-face discussion and online discussion have their own advantages and disadvantages respectively, but they both face the challenge that how to enhance students’ participation. There are many related studies on factors affecting students’ participation, most of which are around external factors, while studies on internal factors are fewer. Spiral of Silence (SOS) theory can be applied for explaining the factors affecting people engage in public discussion, and this study apply extended SOS framework for analyzing psychological factors that affect students’ participation willingness and outspoken willingness in face-to-face discussion and online synchronous discussion. The result shows that communication apprehension can significantly hinder participation, and fear of isolation can significantly reduce outspokenness willingness. Online discussion will cause less communication apprehension and more participation willingness and outspoken willingness. Contemporary, Chinese university students have the courage to face the competition and are not afraid of expressing minority opinions, but they may fear the authority of the government and their teachers.

Mobile game-based learning (MGBL) exploits an entertaining environment for providing digital education. Such an approach involves the construction of students’ groups for gaming towards advancing their knowledge. However, building adequate groups has important pedagogical implications, since the recommendation of appropriate collaborators could further enhance the students’ cognitive abilities. Towards this direction, this paper presents a MGBL application for the tutoring of computer programming. In this application, the system recommends to each student four peers to play with as competitors using genetic algorithm. The genetic algorithm finds the most adequate peers for each student by taking into consideration students’ learning modality, previous knowledge, current knowledge and misconceptions. As such, the student can select from the list one person from the proposed ones, who share common characteristics. The two main reasons why homogeneous groups are chosen to be formed are to promote fair competition and to provide adaptive game content based on players’ characteristics for improving their learning outcomes. Our MGBL application was evaluated using students’ t-test with promising results.

E-learning software is oriented to a heterogeneous group of learners. Thus, such systems need to provide personalization to students’ needs and preferences so that their knowledge acquisition could become more effective. One personalization mechanism is the adaptation to the students’ learning modalities. However, this process requires a lot of time when happening manually and is error-prone. In view of the above, this paper presents a novel technique for learning modalities detection. Our approach utilizes the Honey-Mumford model, which classifies students in activists, reflectors, theorists and pragmatists. Furthermore, the automatic detection uses the fuzzy logic technique taking as input the students’ interaction with the learning environment, namely the kind of learning units visited, their type of media, the comments made by students on learning units and their participation in discussions. Our novel technique was incorporated is a tutoring system for learning computer programming and was evaluated with very promising results.

With Alzheimer’s disease becoming more prevalent, finding effective treatment is imperative. While no pharmacological treatment has yet proven to be efficient, we explore how technology can be integrated into non-pharmacological intervention to enhance its benefits. We propose a new and unique version of Music Therapy, an already existing therapy known to be beneficial. Music therapy has been shown to improve emotions and certain cognitive functions, which is the main focus of our study. To this aim, we designed a virtual reality environment consisting of a music theatre in which participants are immersed among the audience. A meticulously chosen selection of songs is presented on stage accompanied by visual effects. Results show that the environment decreases negative emotions, increases positive emotions, and improved memory performances were observed in most participants following the immersive experience. We speculate that by improving emotions through adaptive music therapy, our environment facilitates memory recall. With virtual reality now being easily accessible and inexpensive, we believe this novel approach could help patients through the disease.

Patients suffering from Alzheimer’s Disease (AD) exhibit an impairment in performing tasks related to spatial navigation. Tasks which require navigational skills by building a cognitive map of the surrounding are found effective in cognitive training. In this paper we investigated the effect of cognitive training using a fully immersive 3D VR orientation game. We implemented an intelligent guidance system which helps to reduce the negative emotions if the participants experience difficulty completing the quests of the game. We found that after playing the orientation game, participants performed better in memory and in certain attention exercises. We also studied the effects of guidance system to reduce the frustration during cognitive training using VR environments.

In recent years, massive open online courses (MOOCs) have become one of the most exciting innovations in e-learning environments. Thousands of learners around the world enroll on these online platforms to satisfy their learning needs (mostly) free of charge. However, despite the advantages MOOCs offer learners, dropout rates are high. Struggling learners often describe their feelings of confusion and need for help via forum posts. However, the often-huge numbers of posts on forums make it unlikely that instructors can respond to all learners and many of these urgent posts are overlooked or discarded. To overcome this, mining raw data for learners’ posts may provide a helpful way of classifying posts where learners require urgent intervention from instructors, to help learners and reduce the current high dropout rates. In this paper we propose, a method based on correlations of different dimensions of learners’ posts to determine the need for urgent intervention. Our initial statistical analysis found some interesting significant correlations between posts expressing sentiment, confusion, opinion, questions, and answers and the need for urgent intervention. Thus, we have developed a multidimensional deep learner model combining these features with natural language processing (NLP). To illustrate our method, we used a benchmark dataset of 29598 posts, from three different academic subject areas. The findings highlight that the combined, multi-dimensional features model is more effective than the text-only (NLP) analysis, showing that future models need to be optimised based on all these dimensions, when classifying urgent posts.

Intelligent Tutoring Systems (ITSs) usually make adaptation decisions based on user models that rely on students’ knowledge. However, there are other interesting indicators, which could be used for adaptation that need further exploration. Students’ efficiency (defined as whether they require a lot of time to achieve correctness in their exercises) and constancy (defined as whether they spend a similar time each day they take exercises in the ITS) are two of these indicators. This work aims to analyze 1) how these variables are distributed among students, 2) their evolution over time, and 3) how they are related to other outcomes. Results show that there are different profiles based on the efficiency; e.g., students with low efficiency that need a lot of time to solve exercises correctly, and low reflective students, among others. Furthermore, efficiency and constancy do not vary on average throughout the course. In addition, students are less constant in their daily time spent when their total time spent and average time per exercise is higher, and more efficient students tend to be more constant. Finally, it was found that neither efficiency nor constancy correlate with better grades. The existence of different profiles based on these variables and that they add a different dimension from student knowledge based on answer on exercises suggest that ITSs can make adaptation based on efficiency and constancy.

The development of Social and Emotional competences of students can significantly improve their learning and social outcomes. This prompts for tools to assist tutors in accomplishing social and emotional learning activities and evaluating the impact achieved. To do so, the blending of Recommender Systems with Machine Learning technologies can be proven beneficial for the design of intelligent and self-learning tools with the capacity to recommend activities, aligned with the social and emotional needs of educational groups. In the current manuscript, we detail a modeling approach for an interactive Recommender System that aims to suggest educational activities to tutors for improving the social and emotional competences of students, taking advantage of Reinforcement Learning techniques. A Reinforcement Learning model has been designed that considers the evolution of students’ social and emotional characteristics and the provided feedback through a set of interactions. Short evaluation of the detailed approach is provided, focusing on validating its appropriateness to serve educational needs.

The impact of gamification has been typically evaluated via self-report assessments (questionnaires, surveys, etc.). In this work, we analise the use of gamification elements as parameters, to predict whether students are going to fail or not in a programming course. Additionally, unlike prior research, we verify how usage of gamification features can predict student performance not only as a discrete, but as a continuous measure as well, via classification and regression, respectively. Moreover, we apply our approach onto two programming courses from two different universities and involve three gamification features, i.e., ranking, score, and attempts. Our results for both predictions are notable: by using data from only the first quarter of the course, we obtain 89% accuracy for the binary classification task, and explain 78% of the students’ final grade variance, with a mean absolute error of 1.05, for regression. Additionally and interestingly, initial observations point also to gamification elements used in the online judge encouraging competition and collaboration. For the former, students that solved more problems, with fewer attempts, achieved higher scores and ranking. For the latter, students formed groups to generate ideas, then implemented their own solution.

In this work, the Affective Logic (AFFLOG) Tutor is presented. An Affective Tutoring System that uses knowledge representation and reasoning tools such as Answer Set Programming and the Event Calculus (EC) in order to represent the main components of the tutor. AI Planning is used to select individual parts of a given course material (tutorials) in order to build a specific course tailored to the needs of each user according to the user’s learning preferences. This course can dynamically change during the teaching session responding to the user’s mental and emotional states, providing affective support by offering praise, consolation or encouragement depending on the current emotion of the user. The design and a functioning implementation of the system is presented. As a proof of concept, a course on how to play the Settlers of Catan(c) board game was designed and implemented.

In this paper we focus on the application of Knowledge Management (KM) techniques to the design and implementation of Smart Learning Systems. In this direction, we explore the application of Soft System Methodology (SSM) in the development of a Knowledge Management (KM) framework for Smart Learning Systems (SLS) in Higher Education (HE). The proposed framework aims at increasing the learning outcomes and the participation of teachers and students in academic life, at improving the services provided in the Smart University, and at establishing new innovative services. It also contributes to the bibliography of system thinking about Smart Learning and Smart Universities.

International trends in education show a shift to student-centered approaches. This paper presents the Learner centered Learning (LcL) environment which is activity oriented, serves learning and assessment, provides multiple types of informative and tutoring feedback components at activity and question level and deploys open learner model to promote self-regulation. The use of the environment in higher education reveals positive results in learning and a promising blending of activities, feedback and open learner model.

This work presents a novel approach for employing conversational agent technology in the context of Massive Open Online Courses (MOOCs), aiming to support learners that work in groups to sustain productive forms of peer dialogue. An exploratory study is presented featuring 56 undergraduate computer science students, who interacted with a conversational agent in the context of an online course. The study investigates the practicability of using configurable conversational agents to provide collaborative learning support and serves as an opportunity to compare two intervention strategies: (a) converging agent interventions, presented in the form of tips relating closely to the topic of the activity and (b) diverging agent interventions, which do not relate directly to the activity topic and pose new domain-relevant questions. The study findings suggest that a convergent agent is often perceived as more helpful by the students. A discourse analysis also reveals a series of interesting interaction patterns, facilitating improvements in the design of future conversational agent systems.

Students’ assessment based on score is no longer relevant today, this is after the major changes of students’ way of life and how they receive knowledge, especially after being immersed in social medias and modern technologies. In light of these significant changes, new methods of students’ assessment must be adopted in order to respond to the requirements of students of the 21st century, and to provide a real evaluation that reflects their knowledge, performance, and skills. Therefore, the Competency-Based Assessment might be a good candidate to meet these requirements. The scope of this paper is to address the issue of competency modelling in technology-enhanced learning systems in order to discover implicit competencies hidden behind students’ activities and how to translate them into acquired competencies. To face these challenges, the authors proposed an approach of semantic analytics of students’ activities data. Therefore, they modelled all knowledge about students and their competencies by Semantic Web and ontological representation; then students’ models have been subjected to a set of learning analytics approaches in order to analyze and evaluate the generated data according to the assessment model. An experimental study indicates that this approach is efficient and expected to show great advantages in evaluating students’ competencies.

Computer Supported Collaborative Learning (CSCL) scripts are important strategies for organizing and conducting collaborative learning. A recent line of research aims at the specification of scripts with tools from formal logic and constraint satisfaction systems. This work proposes a novel inductive method for the automatic generation of formal descriptions of CSCL scripts from examples. The method has been successfully applied in the generation of formal descriptions of well-known scripts from the related literature and educational practice.

The need for effective Intelligent Tutoring Systems (ITSs) and automated assessment is increasing. One area of ITSs has become urgent is that of the automated assessment of reflective writing. The reflective writing has been promoted, in higher education, in order to encourage students to think critically about their learning. However, many frameworks have been developed for assessing student’s reflective writing. Up to our knowledge, there is no empirical studies to validate reflective writing frameworks that used in Computer Science (CS) education. This paper presents the validation of reflective Writing Framework (RWF) by CS educators. The expert panelists validated the RWF. Subsequently, we proposed an ITS model for automating reflective writing analysis. The RWF was accepted that it received a level of consensus from the experts who reported obtaining from good to appropriate results using it.

The accuracy of a framework for annotating reflective writing can be increased through the evaluation and revision of the annotation scheme to ensure the reliability and validity of the framework. To our knowledge, there is a lack of literature related to the accuracy of any reflective writing framework in Computer Science (CS) education. This paper describes a manual annotation scheme, applied during four pilot studies, to validate the authors’ novel Reflective Writing Framework (RWF) for CS education. The results show, through the pilot studies, that the accuracy of Inter-Rater Reliability (IRR) increases from 0.5 to 0.8, which was substantial and close to an almost perfect agreement. This paper contributes to CS education through the reliability and validity of the RWF that can be potentially used for generating an Intelligent Tutoring Systems (ITS) using machine learning algorithms.

Current educational recommender systems (RS) represent an essential support tool and the most used system to interpret patterns and human interaction, which supports deeper learning and provides users with fast and accurate data. In an e-learning environment supported by RS, the learner often needs additional educational resources to enrich his learning scenario to meet his needs and deepen his knowledge and skills. So he spends a lot of time identifying his need, selecting the most convenient data sources and finding the appropriate resources to the current content of his activity. However, in the era of Big Data, apart from the services offered by RS and other data filtering tools, data sources are currently experiencing a significant evolution in terms of volume and variety of available resources. Given the importance of these data, the quality of the recommended content is decreasing significantly, which implies poor knowledge and a failed learning experience.To enhance the quality of student’s learning, we propose an approach of recommending quality educational resources, in accordance to the learner’s learning progress and his individual needs.The quality assessment module is integrated into the recommendation process to judge the level of quality of the resources. To help the quality assessment module make a better decision and improve analytics, we used artificial intelligence technique, Fuzzy Logic to simulate the human reasoning process and aid to deal with the uncertain data in engineering.

Intelligent Tutoring Systems (ITSs) were proven efficient for supporting cognitive tasks; however, their potential contribution to support the development of other human skills is less known, even though they could be of high relevance in dealing with health issues. Lack of physical activity is considered the fourth leading cause for preventable death worldwide. Thus, our current work is aimed at assessing the amount and quality of research conducted to design and develop ITSs for training psychomotor abilities. A systemic literature search was conducted on the most reputable online data sources. After methodological filtering, a short-list of these studies was analyzed in-depth, and the results of their efficacy were considered through a structured comparative grid. The developed tutors use ITS architectures or variations, and capture several sub-fields of the psychomotor domain, from medicine (e.g., surgery, radiology), to military (e.g., training marksmanship), or learning to drive.

Social network analysis (SNA) is widely used to analyze social interactions in asynchronous discussions. A well established model that represents a clear picture of the knowledge-building processes occurring in online discussions is the Community of Inquiry. SNA has recently emerged as an analytical tool for exploring the development of a Community of Inquiry in online learning. However, the application of SNA has been limited and lacks appropriate pedagogical grounding. In this paper, we investigate a) the relationship between social networks and community learning as well as b) the impact of gamification, based on social network metrics, on community learning. According to the results, strong relationships have been revealed mainly between SNA metrics and the social presence, as well as the exploration phase of the cognitive presence of a Community of Inquiry. The evidence also highlights the potential of gamification for empowering the evolution of a Community of Inquiry.

Millions of people have enrolled and enrol (especially in the Covid-19 pandemic world) in MOOCs. However, the retention rate of learners is notoriously low. The majority of the research work on this issue focuses on predicting the dropout rate, but very few use explainable learning patterns as part of this analysis. However, visual representation of learning patterns could provide deeper insights into learners’ behaviour across different courses, whilst numerical analyses can – and arguably, should – be used to confirm the latter. Thus, this paper proposes and compares different granularity visualisations for learning patterns (based on clickstream data) for both course completers and non-completers. In the large-scale MOOCs we analysed, across various domains, our fine-grained, fish-eye visualisation approach showed that non-completers are more likely to jump forward in their learning sessions, often on a ‘catch-up’ path, whilst completers exhibit linear behaviour. For coarser, bird-eye granularity visualisation, we observed learners’ transition between types of learning activity, obtaining typed transition graphs. The results, backed up by statistical significance analysis and machine learning, provide insights for course instructors to maintain engagement of learners by adapting the course design to not just ‘dry’ predicted values, but explainable, visually viable paths extracted.

In mathematics, sciences and economics, understanding and working with graphs are important skills. However, developing these skills has been shown to be a challenge in secondary and higher education as it involves high order thinking processes such as analysis, reflection and creativity. In this study, we present Interactive Virtual Math, a tool that supports the learning of a specific kind of graphs: dynamic graphs which represent the relation between at least two quantities that covary. The tool supports learners in visualizing abstract relations through enabling them to draw, move and modify graphs, and by combining graphs with other representations, especially interactive animations and textual explanations. This paper reports a design experiment about students’ learning graphs with this tool. Results show that students with difficulty in generating acceptable graphs improve their ability while working with the tool.

This paper presents a web application for Association Rules Mining (ARM). It utilizes Apriori that is the most widely used algorithm for this type of data mining tasks. The web application is called WebApriori and offers a modern responsive web interface and a web service to scientific communities working in the field of ARM. It is also appropriate for educational purposes. WebApriori implements an Apriori engine that can efficiently discover the hidden associations in data and it is capable to process different types of datasets. Part of the process involves the removal of redundant associations rules. The asynchronous communication between the front-end, back-end, web service and Apriori engine layers efficiently handles multiple concurrent user requests.

This study counts pairs of successive problem-solving dialogue acts in typed-chat student group problem-solving. In ordinary two-person dialogue a pair of successive turns frequently corresponds to an exchange. It is hypothesized that by characterizing and counting dialogue exchanges, it will be possible to evaluate the productivity of the problem-solving dialogue. However these small-group dialogues contain more than two participants, and participants can type-chat simultaneously in overlapping dialogue exchanges. Successive turns in the linearized transcript do not necessarily correspond to two-person dialogue exchanges. This work reports on whether counting such pairs is likely to reveal dialogue behavior.

Do students retain the programming concepts they have learned using software tutors over the long term? In order to answer this question, we analyzed the data collected by a software tutor on selection statements. We used the data of the students who used the tutor more than once to see whether they had retained for the second session what they had learned during the first session. We found that students retained over 71% of selection concepts that they had learned during the first session. The more problems students solved during the first session, the greater the percentage of retention. Even when students already knew a concept and did not benefit from using the tutor, a small percentage of concepts were forgotten from the first session to the next, corresponding to transience of learning. Transience of learning varied with concepts. We list confounding factors of the study.

An attractive approach to sharing knowledge in Higher Education Institutions (HEIs) is offered by Best Teaching Practices (BTPs), which results from the accumulation of academics’ experiences gained through years of teaching. Many universities still struggle in documenting, sharing and applying the knowledge gained by instructors. Low quality BTP documentation can deter the successful implementation of teaching expertise, and this can affect instructors’ teaching performance and may result in lower levels of learner achievement than could otherwise be achieved. In order to address this issue, this paper proposes a comprehensive and practical Computer-based Teaching Practices Management System (TPMS) for supporting the capturing of instructors’ BTPs. A design science research methodology was adopted in order to understand how instructors share their knowledge, and in order to map instructors’ behavior to system features. A mixed-methods evaluation of the users’ experiences of using the system has shown that the instructors were satisfied with the BTP Document Template (a key element of the TPMS) and were mostly positive about its attributes. The results of this evaluation were promising but also highlighted some drawbacks of the system.

This paper describes the development of a knowledge sharing system architecture based on the knowledge sharing behavior (KSB) of real academicians in public institutions. Semi-structured interviews were conducted with 22 academics who work in Saudi universities to investigate current academics’ KSB and explore academics’ needs for a new knowledge management system (KMS). An inductive content analysis approach was used to help the researchers to extract themes that were frequently mentioned by the interviewees. The analysis and findings will expand an area of KMS in academic institutions, particularly universities which might still theoretically and empirically not sufficiently covered.

Students of human anatomy face domain-specific and generational challenges in their acquisition of the subject material. Anatomy is detail-heavy and overwhelming, as the larger conceptual framework is built upon interdependent components connected by several types of multidirectional relationships in four domains, all acting simultaneously. Adding to the difficulty of the material are changes in the educational and social environment, including “teaching to the test” and extensive cell phone usage among students, which has been found to correlate positively with inattention. Proper scaffolding of student development of higher-level cognitive processes has the potential to create positive learner outcomes in tomorrow’s medical workforce. Based on over 100 hours of expert interviews, our new ITS for teaching human anatomy and physiology uses a hierarchical concept map with unlockable content, shows multiple types of relationships, and is color-coded by domain to help students master the complex material.

The untamed big data era raises opportunities in learning analytics sector for the provision of enhanced educational material to learners. Nevertheless, big data analytics, brings big troubles in exploration, validation and predictive model development. In this paper, the authors present a data-driven methodology for greater utilization of learning analytics datasets, with the purpose to improve the knowledge of instructors about learners performance and provide better personalization with optimized intelligent tutoring systems. The proposed methodology is unfolded in three stages. First, the learning analytics summarization for initial exploratory purposes of learners experience and their behavior in e-learning environments. Subsequently, the exploration of possible interrelationships between metrics and the validation of the proposed learning analytics schemas, takes place. Lastly, the development of predictive models and simulation both on an aggregated and micro-level perspective through agent-based modeling is recommended, with the purpose to reinforce the feedback for instructors and intelligent tutoring systems. The study contributes to the knowledge expansion both for researchers and practitioners with the purpose to optimize the provided online learning experience.

Deploying Learning Analytics that significantly improve learning outcomes remains a challenge. Motivation has been found to be related to academic achievement and is argued to play an essential role in efficient learning. We developed a Learning Analytics dashboard and designed an intervention that relies on goal orientation and social comparison. Subjects can see a prediction of their final grade in a course as well as how they perform in comparison to classmates with similar goal grades. Those with access to the dashboard ended up more motivated than those without access, outperformed their peers as the course progressed and achieved higher final grades. Our results indicate that learner-oriented dashboards are technically feasible and may have tangible benefits for learners.

Modelling driving dynamics in experimental educational scenarios represents a key enhancement of a SMART city, where citizen-oriented politics promote traffic rules knowledge retention and awareness. We propose an instructional design of mapping simulations of real-world urban networks as use cases for practicing traffic rules. Traffic simulations have been implemented through Reinforcement Learning agents, using a modified Policy Proximal Optimization (PPO) strategy, demonstrating a good sample efficiency. The proposed objective function and the selected policy positive and negative rewards empower the car agent to reach a predefined destination, from a predefined start position, while adapting to the route line. Results validate the applicability of the proposed approach to educational simulations, within a generic gamified environment. The approach proposes a further extension towards adaption to complex lane design (e.g. traffic signs) and player’s in-game behavior.

In this paper, we propose and discuss a conceptual framework, based on knowledge management and selective literature review, for enhancing the contribution of students in the Institutional Quality Assurance processes (IQA). The framework is related to the design of a mechanism for IQA, which ensures the improvement of the assessment and uses various methods, indicators, and criteria of auditing. The mechanism is supported by an integrated system for the evaluation of the learning practice, and it also includes a subsystem for evaluation based on Social Networking, a subsystem of learning analytics, and a subsystem for increasing the commitment to the IQA.

The actual data availability, readiness and publicity has slowed down the research of making use of computational intelligence to improve the knowledge delivery in an emerging learning mode, namely adaptive micro open learning, which naturally has high demand in quality and quantity of data to be fed. In this study, we contribute a novel approach to tackle the current scarcity of both data and rules in micro open learning, by adopting evolutionary algorithm to produce association rules with both rare and negative associations taken into account. These rules further drive the generation and optimization of learner profiles through refinement and augmentation, in order to maintain them in a low-dimensional, descriptive and interpretable form.

The original version of the chapter was inadvertently published without incorporating the author’s proof corrections. The chapter has now been corrected and approved by the author.