Search

For the forty-fifth time, the Association for Educational Communications and Technology (AECT) is sponsoring the publication of these Proceedings. Papers published in this volume were presented online and onsite during the annual AECT Convention. The Proceedings of AECT's Convention are published in two volumes. Volume #1 contains papers dealing primarily with research and development topics. Papers dealing with the practice of instructional technology including instruction and training issues are contained in Volume #2. This year, both volumes are included in one document.

This paper presents the best practices used by institutional review boards (IRBs) and human research protections programs (HRPPs) to prepare online graduate student investigators for human research protections specific to research within online graduate degree programs or where research supervisors are not proximal to graduate student investigators and their research protocols. In recent years, advances in artificial intelligence (AI), machine learning (ML), and other data mining/scraping forms have adversely impacted individual privacy and the unintended sharing of personally identifiable information (PII). With this growth of ubiquitous digital technologies, such as AI, ML, and data mining/scraping, used across online graduate degree programs, specialized training and preparation are needed to best prepare graduate student researchers for human research protections involving data with PII. Implications for IRBs and HRPPs are also addressed in this rapidly evolving climate, with recommendations for the design of online graduate degree programs that include graduate research and the best strategies to prepare online graduate student investigators for human research protections. [This paper was published in: "1st Annual Virtual Fall National Conference on Creativity, Innovation, and Technology (NCCiT) Proceedings," November 15-16, 2023, pp. 84-108.]

There is a growing concern that AI is likely to replace the work done face to face in the classroom by teachers. The concerns also extend to the students use of AI to complete assignments which could impact on their grades either positive or negative and in some cases, when a student work is detected with high AI the work could be classified as plagiarism if AI usage is not declared. On another note, there are increasing debates about the use of AI as a valid tool to support work completed by human beings. Whatever maybe the growing concerns, many researchers have argued that AI is not likely to replace education jobs such as teaching and learning done by teachers and education administrators. The purpose of this paper is to explore debates around the use of AI in education, mostly in teaching and learning and assessment of students university misconduct policy. Teachers work and the link to the opinions on integrating AI in the classroom are illuminated by empirical evidence gathered via interviews. A lot of educators respond to AI in different ways. Some of the debates falls around AI as God of the admin work and assessment of students s sent tools that can help reduce some work such as helping with multiple choice questions, on the other hand, some students have been penalised and in some cases failed their work due to use of AI in completing their assignments without acknowledging the use. In addition, others have argued that AI has come to replace the work done by teachers and are anxious about AI in education jobs done by teachers, hence would not bear the idea for classroom teaching and learning. [For the complete Volume 22 proceedings, see ED656158.]

The integration of ChatGPT in Personal Learning Environments (PLEs) has emerged as a promising approach to personalized learning in tertiary education. ChatGPT is believed to have the potential to transform traditional higher education into a more personalized, quality-driven, and student-centered learning experience that fosters critical thinking, self-regulated learning, and creativity. While recent studies have highlighted the potential benefits of ChatGPT in enhancing personalized learning experiences, there are several risks and challenges that need to be addressed. This paper reviews relevant literature on ChatGPT and PLEs and identifies key risks and challenges associated with their integration, including ethical concerns, data privacy, technical issues, and user acceptance. Meanwhile, the paper also proposes ways and thoughts for the future implementation of ChatGPT in PLEs. The paper concludes that ChatGPT has significant potential to facilitate a new round of educational revolution which pushes educators to reconsider why to teach, how to teach, and what to teach. [For the full proceedings, see ED654100.]

This paper provides a possible strategy for integrating large language artificial intelligence models (LLMs) in supporting students' education in artistic or design activities. We outline the methodological foundations concerning the integration of CHATGPT LLM in the educational approach aimed at enhancing artistic conception and design ideation. We also present the knowledge and system architecture for integrating LLM in the °'°Kobi system. Finally, this paper discusses some relevant aspects concerning the system's application in a real educational context and briefly reports its preliminary assessment. [For the full proceedings, see ED636095.]

The growth of Artificial Intelligence (AI) chatbots has created a great deal of discussion in the education community. While many have gravitated towards the ability of these bots to make learning more interactive, others have grave concerns that student created essays, long used as a means of assessing the subject comprehension of students, may be at risk. The bot's ability to quickly create high quality papers, sometimes complete with reference material, has led to concern that these programs will make students too reliant on their ability and not develop the critical thinking skills necessary to succeed. The rise in these applications has led to the need for the development of detection programs that are able to read the students submitted work and return an accurate estimation of if the paper is human or computer created. These detection programs use natural language processing's (NLP) ideas of perplexity, or randomness of the text, and burstiness, or the tendency for certain words and phrases to appear together, plus sophisticated algorithms to compare the essays to preexisting literature to generate an accurate estimation on the likely author of the paper. The use of these systems has been found to be highly effective in reducing plagiarism among students, however concerns have been raised about the limitations of these systems. False positives, false negatives, and cross language identification are three areas of concern amongst faculty and have led to reduced usage of the detection engines. Despite the limitations however, these systems are a valuable tool for educational institutions to maintain academic integrity and ensure that students are submitting original work. [For the full proceedings, see ED656038.]

There were two projects at the National Defence University of Finland (NDU), which both ended by the end of 2022. One of them tried to find the answers to the main question: How artificial intelligence (AI) could be used to improve learning, teaching, and planning? The other tried to find the answer to the main question: What new skills do officers need when artificial intelligence is coming? We did literature reviews and found out that intelligent technology combined with data analytics can offer several improvements to traditional classroom teaching. From literature reviews, we also found some new skills that officers might need to be able to handle AI-based technologies. This is a position paper presenting the arguable opinions of the writers. We have found lots of benefits that the use of intelligent learning technology can bring, mainly by supporting individual learning paths. There is also an obvious need for AI officers who should have a deeper understanding of the AI-supported technology than normal officers. This project and some other similar projects have raised a lot of discussions, one seminar series about artificial intelligence and we do have some trained AI officers as well. [For the full proceedings, see ED629086.]

Eight scholarly papers and seven abstracts comprise the content of the twenty-third annual Brick & Click Libraries Conference, held annually at Northwest Missouri State University in Maryville, Missouri. The 2023 paper and abstract titles include: (1) The Reliability and Usability of ChatGPT for Library Metadata (Jenny Bodenhamer); (2) A Balancing Act in the Archives: Increasing Access to the Great Plains Black History Museum Collections (Wendy Guerra and Lori Schwartz); (3) Developing Info Students Where They Are: Personalizing Instruction to Increase Literacy Skills to Meet Engagement (Jorge A. León); (4) Empowering Undergraduates: Building Confidence in Primary Source Literacy (Jaycie Vos and Jess Cruz); (5) Quest for the Best: An Info Lit Strategy for First Year Seminars (Stephanie Hallam, Mary Bangert, and Michael Bezushko); (6) Are We Putting Our Values into Practice? Chat Reference Assessment (Mardi Mahaffy); (7) A Pilot Workshop on AI Art and Libraries at the University of Mississippi (Alex Watson); (8) New Expansions of Open Access to Benefit Research and Researchers (Barbara Pope); (9) Zettelkasten Note-Taking in Zotero for Grounded Writing (Rachel Brekhus); (10) Building Community: Library Leadership of a Common Book Program (Jill Becker); (11) Digital Media and Innovation Lab: A Must Have for Academic Libraries (Navadeep Khanal and Joseph Sabo); (12) Digital Libraries as Digital Third Place: Virtual Programming in the Age of Loneliness (Craig Finlay and Jenny Haddon); (13) Community Engagement: Academic and School Library Partnerships (Melissa Dennis); (14) Launching a Ticketing System With Asana (Hong Li); and (15) Meeting the Needs of Student Parents (Sarah Hebert). [For the 2022 proceedings, see ED623765.]

The expanding domain of Artificial Intelligence (AI) offers a diverse array of educational applications and tools. However, the scholarly exploration of AI's suitability for enhancing English as a Foreign Language (EFL) instruction at the university level remains notably limited. This research gap impedes educators from fully harnessing AI's pedagogical potential. Given the inclusion of linguistic and literary disciplines in preservice EFL teacher training in Slovakia, it is increasingly imperative for educators to acquaint themselves with various AI tools, enabling the development of effective methodologies for enhancing EFL teaching and learning. Integrating AI into teacher training programs equips future EFL educators with essential skills for 21st-century classrooms and meets the evolving needs of digitally proficient students. This paper aims to provide a concise yet comprehensive overview of AI's relevance to pre-service EFL teacher training, encompassing linguistic and literary domains, by categorising six prominent AI forms: a) Natural Language Processing (NLP) Tools, b) Content Creation and Personalisation tools, c) Content Recommendation Systems, d) Emotion and Sentiment Analysers, e) Text Summarisation and Analysis tools, and f) Chatbots and Virtual Assistants. Furthermore, it highlights the research gap in AI's implementation in EFL education and emphasises the need to explore pedagogical and ethical implications while outlining future research directions to enhance our understanding of this dynamic field. [For the full proceedings, see ED652261.]

The Community Learning and Service Partnership (CLASP) being named the recipient of the Malcolm Knowles Award prompts retrospection and raises the question of what to present at the American Association for Adult and Continuing Education (AAACE) conference. What is the overarching narrative of CLASP? Given the international and intergenerational nature of the program, anecdotes abound--stories of transformation, rich relationships, and shared achievements; but what is the "metastory"? An unfinished, in-house film made years ago wherein program participants and observers spoke freely on camera may serve. From a detached perspective, we turn to artificial intelligence in the form of a qualitative data analytics program. This paper reports the results of that media analysis and conference attendees' reactions to the film. [For the full proceedings, see ED648717.]

Knowledge components (KCs) have many applications. In computing education, knowing the demonstration of specific KCs has been challenging. This paper introduces an entirely data-driven approach for: (1) discovering KCs; and (2) demonstrating KCs, using students' actual code submissions. Our system is based on two expected properties of KCs: (1) generate learning curves following the power law of practice; and (2) are predictive of response correctness. We train a neural architecture (named KC-Finder) that classifies the correctness of student code submissions and captures problem-KC relationships. Our evaluation on data from 351 students in an introductory Java course shows that the learned KCs can generate reasonable learning curves and predict code submission correctness. At the same time, some KCs can be interpreted to identify programming skills. We compare the learning curves described by our model to four baselines, showing that: (1) identifying KCs with naive methods is a difficult task; and (2) our learning curves exhibit a substantially better curve fit. Our work represents a first step in solving the data-driven KC discovery problem in computing education. [For the complete proceedings, see ED630829.]

In this paper, we present an extended evaluation of a course recommender system designed to support students who struggle in the first semesters of their studies and are at risk of dropping out. The system, which was developed in earlier work using a student-centered design and which is based on the explainable k-nearest neighbor algorithm, recommends a set of courses that have been passed by the majority of the student's nearest neighbors who have completed their studies. The present evaluation is based on the data of students from three different study programs. One result is that the recommendations do lower the dropout risk. We also discovered that while the recommended courses differed from those taken by students who dropped out, they matched quite well with courses taken by students who completed the degree program. Although the course recommender system targets primarily students at risk, students doing well could use it. Furthermore, we found that the number of recommended courses for struggling students is less than the number of courses they actually enrolled in. This suggests that the recommendations given indicate a different and hopefully feasible path through the study program for students at risk of dropping out. [For the complete proceedings, see ED630829.]

Massive Open Online Courses (MOOCs) make high-quality learning accessible to students from all over the world. On the other hand, they are known to exhibit low student performance and high dropout rates. Early prediction of student performance in MOOCs can help teachers intervene in time in order to improve learners' future performance. This is particularly important in healthcare courses, given the acute shortages of healthcare staff and the urgent need to train data-literate experts in the healthcare field. In this paper, we analysed a health data science MOOC taken by over 3,000 students. We developed a novel three-step pipeline to predict student performance in the early stages of the course. In the first step, we inferred the transitions between students' low-level actions from their clickstream interactions. In the second step, the transitions were fed into Artificial Neural Network (ANN) that predicted student performance. In the final step, we used two explanation methods to interpret the ANN result. Using this approach, we were able to predict learners' final performance in the course with an AUC ranging from 83% to 91%. We found that students who interacted predominately with lab, project, and discussion materials outperformed students who interacted predominately with lectures and quizzes. We used the DiCE counterfactual method to automatically suggest simple changes to the learning behaviour of low- and moderate-performance students in the course that could potentially improve their performance. Our method can be used by instructors to help identify and support struggling students during the course. [For the complete proceedings, see ED630829.]

We propose an innovative, effective, and data-agnostic method to train a deep-neural network model with an extremely small training dataset, called VELR (Voting-based Ensemble Learning with Rejection). In educational research and practice, providing valid labels for a sufficient amount of data to be used for supervised learning can be very costly and often impractical. The shortage of training data often results in deep neural networks being overfitting. There are many methods to avoid overfitting such as data augmentation and regularization. Though, data augmentation is considerably data dependent and does not usually work well for natural language processing tasks. Moreover, regularization is often quite task specific and costly. To address this issue, we propose an ensemble of overfitting models with uncertainty-based rejection. We hypothesize that misclassification can be identified by estimating the distribution of the class-posterior probability P(y|x) as a random variable. The proposed VELR method is data independent, and it does not require changes to the model structure or the re-training of the model. Empirical studies demonstrated that VELR achieved classification accuracy of 0.7 with only 200 samples per class on the CIFAR-10 dataset, but 75% of input samples were rejected. VELR was also applied to a question generation task using a BERT language model with only 350 training data points, which resulted in generating questions that are indistinguishable from human-generated questions. The paper concludes that VELR has potential applications to a broad range of real-world problems where misclassification is very costly, which is quite common in the educational domain. [For the complete proceedings, see ED630829.]

Artificial intelligence (AI) technology is already challenging a variety of societal areas, including education. It is transforming education to data driven. AI-enhanced technologies in education (abbreviated AIinED) will have a significant role in changing the teaching and learning methods, as well as impacting the behavior and organization of the educational system. It is considered that the AIinED will change the paradigm of education in the future. And yet, AIinED is still more in the lab than being practically implemented in education and training. We consider three major players in the implementation of AIinED -- students, teachers, and society. All three can benefit from AIinED and at the same time be a potential target of the risks that AIinED brings along with its promises -- may be one of the reasons why main stakeholders (UNESCO, EC etc.) have been developing guidelines and recommendations for ethical use of AIinED. The literature shows that the center of AIinED system will be the student, but we consider the student not only as a target but also as a source of ideas for AIinED development with the potential to accelerate the process of adoption of AIinED. Hence, one of the big questions should be how the students foresee the role of artificial intelligence in education. To initiate such a question, though, it is important to know the level of understanding among the students about what and where artificial intelligence is. There are three major aspects that AIinED must be considered accordingly -- technological, lawful, and ethical. This paper presents the results of a study on high school students' understanding of AI technologies and their attitudes to their application in education. A survey was used as a tool to elaborate. The conceptual model of the research was developed on the basis of established theories linking attitudes to behavior and the acceptance of artificial intelligence technologies in education. Each element of this concept is explored with a different part of the questionnaire, which contains a total of 12 questions (some of which with sub-questions). The survey was elaborated online within October-November 2021. A link to the questionnaire in Bulgarian was provided to 178 high and vocational high schools educating students aged 14-19 (grades 8-12) across the country (Bulgaria). 766 students submitted their replies through the online survey form. Descriptive statistics and analysis of the frequencies of the respondents' opinions were made based on the data. The results show that the students participating in the survey:(a) understand the essence of AI technologies; (b) they are convinced of the usefulness of the application of artificial intelligence technologies in their daily activities and strongly believe that it improves it; (c) are not entirely clear about the benefits of artificial intelligence enhanced technologies in learning and teaching; (d) do not demonstrate sufficient knowledge and understanding of the necessity of ethical use of AI technologies in education; The latter reduces the positive influence of the perceived usefulness of artificial intelligence technologies in the learning process on students' attitudes. [For the full proceedings, see ED639391.]

In order to keep up with the speed of science and technology in today's conditions, it is important for human beings to constantly renew themselves and adapt to the changing world and living conditions. In this content, education plays a significant role, and art education, in particular, hold a pioneering role in nurturing civilized individuals. Art education, which is considered important at a young age, takes on the locomotive task of developing a child's creative power and integrating the knowledge and skills they acquire into real-life applications by incorporating current techniques and tools. The importance of drawing in a child's development can be expressed by its role as a means for perceiving the external world and self-expression. Studies also demonstrate that children utilize the knowledge and skills they acquire on touch screens in their real lives. Drawing applications with fun themes for children attract attention. However, these applications are generally designed for entertainment purpose and targeted towards children who can read and write. The lack of an educational software that provides audio guidance and teaches drawing with the support of artificial intelligence has led to the development of the 'Artificial Intelligence-Assisted Directed Drawing Technique' as identified in the literature research. This technique supports a child's creativity, enhances their fine motor skills, increases their digital literacy, and provides an enjoyable learning experience. Additionally, it enables children to create original drawings and express their ideas while guiding their progress and supporting their learning process. The artificial intelligence-assisted directed drawing application is believed to contribute to the development of a child's self-drawing skills, beneficial use of technological devices like tablets by preschoolers, and increasing the interest and talent of future generations in artistic activities. [This paper was published in: "EJER Congress 2023 International Eurasian Educational Research Congress Conference Proceedings," Ani Publishing, 2023, pp. 511-519.]

This paper presents a mobile application aimed at enhancing the financial literacy of college students by monitoring their spending patterns and promoting better decision-making. The application is developed using the agile methodology with Android Studio and Flutter as development tools and Firebase as a database. The app is divided into sub-applications, with the home page serving as the program's integration point, displaying a summary of the user's financial progress. The app generates valuable insights into the user's current and future financial success, utilizing data analytics and machine learning to provide detailed and summary insights into the user's financial progress. The machine-learning algorithm used in this app is linear regression, which predicts the user's income and expenses for the upcoming month based on their historical spending data. In addition, the app highlights deals and student discounts in the user's vicinity and links to financial articles that promote better financial planning and decision-making. By promoting responsible spending habits and providing valuable financial insights, this mobile application aims to help students become financially literate and make informed financial decisions for future. [For the full proceedings, see ED654100.]

Artificial intelligence has potentiated changes in higher education. In this position paper, we propose to discuss dimensions that shape higher education and that can be transformed by the mobilization of Artificial Intelligence: teaching-learning processes (from the perspectives of scholars and students), research (from the perspective of professional researchers, students and advisors), management of organizations (from the perspective of managers, subordinates and students), and the relationship with the outside (the look from inside to outside, as well as from outside to inside). To this end, a document collection and analysis was carried out to support our argument about the new challenges and the potential that emerges from the existence and application of artificial intelligence in higher education. As this change, like any transformation, will have positive and negative aspects, some implications of both a theoretical and more practical nature of these challenges for academics, students and other stakeholders in higher education institutions are also discussed. [For the full proceedings, see ED654100.]

Although information technology (ICT) education is being strengthened based on the national context, there are reports suggesting a decline in young people's communication skills. This phenomenon can be attributed to the rapid development of informatisation, which includes the diversification and spread of information tools, as well as the prevalence of nonverbal communication, such as pictograms in social networking services. In addition, the COVID-19 pandemic has drastically reduced face-to-face communication opportunities, making interactive communication in on-demand classes challenging. Even in assignments and short tests completed during class, many instances of content being copied and pasted from the web or written in a disorganized manner have been observed. For instance, students entering ICT-related careers, particularly those graduating from the faculty of informatics, must possess the ability to communicate with engineers and clients while implementing ICT advancements. Alongside programming skills, strong communication abilities are essential. Moreover, the emergence of generative artificial intelligence (AI) tools, such as ChatGPT and Bing AI, has considerably diminished the opportunities for independent thinking. In the current era of enhanced ICT education, AI, and IoT, the Faculty of Informatics at the Kanagawa Institute of Technology has been engaged in discussions regarding learning materials that aim to strengthen students' ability to think and communicate in their own words, preventing the mechanisation of individuals. This paper presents the development and implementation of learning materials designed to enhance students' verbal communication skills through the description and re-production of mathematical graphs. [For the full proceedings, see ED636095.]

This paper delves into statistical disparities between human-written and ChatGPT-generated texts, utilizing an analysis of Shannon's equitability values, and token frequency. Our findings indicate that Shannon's equitability can potentially be a differentiating factor between texts produced by humans and those generated by ChatGPT. Additionally, we uncover substantial distinctions when studying the most frequent tokens. [For the full proceedings, see ED636095.]

Due to the advances of artificial intelligence (AI) and natural language processing, new AI-powered writing tools have emerged. They can be used by students among other things for text translation, to improve spelling or to generate new texts. In academic writing, AI-powered writing tools are posing challenges but also opportunities for teaching and learning. It is an open question in which way to sensibly deal with these tools. To address the issue, this paper investigates, what interests different stakeholders (students, lecturers, university administration) pursue in relation to AI-powered writing tools. Building on this, tensions between different stakeholders are identified and (teaching) strategies proposed to deal with these tensions. To discuss the findings in light of recent developments around ChatGPT, semi-structured expert interviews were conducted in April 2023 with five academic writing lecturers at the University of St. Gallen. The results suggest that as writing tools become more and more powerful, the need for strategies to ensure their reasonable and transparent use also increases. [For the full proceedings, see ED636095.]

This paper proposes a didactic design that is centered around an immersive, multilinear narrative in virtual reality as a means of illustrating human life on the edge of technological singularity. It explores the potential of narrative scenarios to trigger a discourse from users' perspective. Affective Computing is taken as a use case. It is a subfield of AI that focuses on identifying, understanding, and appropriately responding to human emotions. Its goal is to create more personalized and emotionally engaging human-machine interactions. To explore what life might be like if an emotionally intelligent AI became our best friend, a multilinear scenario was created. This scenario takes the reader through different stages of the protagonist's life, starting from the first day of secondary school and ending with the loss of a loved one in midlife. The systematic approach to create and validate the multilinear interactive scenario is described and the results of an experiment with 164 participants are presented. The ultimate goal is the application of this approach for educational purposes regarding ethical thinking and responsible innovations. [For the full proceedings, see ED636095.]

The exponential growth of scholarly publications in recent years has presented a daunting challenge for researchers to keep track of relevant articles within their research field. To address this issue, we examined the capabilities of InfraNodus, an AI-Powered text network analysis platform. InfraNodus promises to provide insights into any discourse, uncover blind spots, and enhance a scholar's perspective by representing text as a network graph with relevant topical clusters and their relations. To understand the tools' effectiveness in analyzing scholarly articles, we used a set of 15 abstracts and 15 full papers. Our findings revealed that InfraNodus could indeed create topical clusters and meaningful patterns from abstracts, but its generated questions and summaries lacked relevance and coherence with the content. A deeper understanding of how the AI operates within the tool would benefit researchers seeking to optimize their literature review processes. [For the full proceedings, see ED636095.]

In this study, the authors have developed a web-based programming exercise system currently implemented in classrooms. This system not only provides students with a web-based programming environment but also tracks the time spent on exercises, logging operations such as program editing, building, execution, and testing. Additionally, it records their results. For educators, the system offers insights into each student's progress, the evolution of their source code, and the instances of errors. While teachers find these functions beneficial, the method of providing feedback to students needs improvement. Immediate feedback is proven to be more effective for student learning. If the final course score could be predicted based on early data (e.g., from the 1st or 2nd week), students could adapt their study strategies accordingly. This paper demonstrates that one can predict the final score using the system's operational logs from the initial phases of the course. Furthermore, the score predictions can be revised weekly based on new class logs. We also explore the potential of offering tailored advice to students to enhance their final score. [For the full proceedings, see ED636095.]

Today more than ever before, we have access to new technologies which provide unforeseen opportunities for educators to pursue new innovations in online education. Pursuing innovation is a complex process! It starts with an idea, but that needs to be coupled with the right team of experts willing to take big risks and put in the hard work to build something new. An instructional design team was empowered to reimagine an Introduction to Sociology university course as a Game-Based Learning (GBL) experience utilizing cutting edge Virtual Reality (VR) technology. The result was a innovate collaborative process that resulted in a brand-new type of learning based in Game theory, Method of Loci, and VR Immersion Simulations to promote deeper retention of core concepts. The team deconstructed the way that university courses operated, in order to rebuild the educational process in a whole new, learner-centric manner. In addition to a review of the build process, this paper will explore the results of in-course surveys completed by student participants. [For the full proceedings, see ED656038.]

The unprecedented growth in the use of AI and its related technologies will put a tremendous stress on US institutions to produce the required number of technologically prepared workers to fill critically important job openings. In the US, low-income and URM students participate less vigorously in STEM-related fields; the problem is even more serious in post-baccalaureate level degrees. To address the future needs of the nation, we must increase the number of low-income students in STEM, with special attention to AI related technologies, to fill the millions of technology job openings. This paper will report on the impact of a NSF SSTEM project in which we combined (a) a mentorship model for talented, low-income students to develop a sense of self-efficacy and belongingness along with (b) a model of curricular and co-curricular supports (e.g., including engagement with AI technologies and research) and (c) limited financial assistance, all of which have increased the low-income student success in completing both their BS degree in engineering and their MS degree in AI, and addressing a national need. [For the full proceedings, see ED656038.]

Modern language learning applications have become 'smarter' and 'intelligent' by including Artificial Intelligence (AI) and Machine Learning (ML) technologies to collect different kinds of data. This data can be used for analysis on a microscopic and/or macroscopic level to provide granulation of knowledge. We analyzed 1,213 French language learner data over a 30-day period, publicly available from Duolingo, to compare the progression of individual learners (microscopic granularity) and large groups of learners (macroscopic granularity). Using network modeling, we compared patterns of individual learners against one another and that of a learning community and determined what groups of learners typically practice across communities. Preliminary results suggest how applications for L2 learning can be designed to create an optimal path for learning. [For the complete volume, "Intelligent CALL, Granular Systems and Learner Data: Short Papers from EUROCALL 2022 (30th, Reykjavik, Iceland, August 17-19, 2022)," see ED624779.]

The use of chatbots in language learning has been on the rise. In recent Computer-Assisted Language Learning (CALL) research, there is a consensus that rule-based, scripted voice systems are optimal for language learning. Such systems integrate well into instructed language learning in that interactions with the user are predictable and controlled. Open, AI-based voice systems (such as in personal assistants like Siri) do not provide that degree of task-oriented learning. However, the argument is made here that they have the potential to provide open-ended conversational practice and language development which aligns with an ecological, usage-based perspective on language development. [For the complete volume, "Intelligent CALL, Granular Systems and Learner Data: Short Papers from EUROCALL 2022 (30th, Reykjavik, Iceland, August 17-19, 2022)," see ED624779.]

Images in children's books convey messages about society and the roles that people play in it. Understanding these messages requires systematic measurement of who is represented. Computer vision face detection tools can provide such measurements; however, state-of-the-art face detection models were trained with photographs, and 80\% of images in children's books are illustrated; thus existing methods both misclassify and miss classifying many faces. In this paper, we introduce a new approach to analyze images using AI tools, resulting in data that can assess representation of race, gender, and age in both illustrations and photographs in children's books. We make four primary contributions to the fields of deep learning and social sciences: (1) We curate an original face detection data set (IllusFace 1.0) by manually labeling 5,403 illustrated faces with bounding boxes. (2) We train two AutoML-based face detection models for illustrations: (i) using IllusFace 1.0 (FDAI); (ii) using iCartoon, a publicly available data set (FDAI_iC), each optimized for illustrated images, detecting 2.5 times more faces in our testing data than the established face detector using Google Vision (FDGV). (3) We curate a data set of the race, gender, and age of 980 faces manually labeled by three different raters (CBFeatures 1.0). (4) We train an AutoML feature classification model (FCA) using CBFeatures 1.0. We compare FCA with the performance of another AutoML model that we trained on UTKFace, a public data set (FCA_UTK) and of an established model using FairFace (FCF). Finally, we examine distributions of character identities over the last century across the models. We find that FCA is 34% more accurate than FCF in its race predictions. These contributions provide tools to educators, caregivers, and curriculum developers to assess the representation contained in children's content. [This paper was published in: "Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)," 2022, pp. 462-471.]

The context for this paper is the "Synthetic Training Environment Experiential Learning -- Readiness" (STEEL-R) project [1], which aims to estimate individual and team competence using data collected from synthetic, semi-synthetic, and live scenario-based training exercises. In STEEL-R, the "Generalized Intelligent Framework for Tutoring" (GIFT) orchestrates scenario sessions and reports data as experience API (xAPI)statements. These statements are translated into assertions about individual and team competencies by the "Competency and Skills System" (CaSS). Mathematical models use these assertions to estimate the competency states of trainees. This information is displayed in a dashboard that enables users to explore progression over time and informs decisions concerning advancement to the next training phase and which skills to address. To test, tune, and demo STEEL-R, more data was needed than was available from real-world training exercises. Since the raw data used to estimate competencies are captured in xAPI statements, a component called DATASIM was added. DATASIM simulated training sessions by generating xAPI statements that conformed to a STEEL-R "xAPI Profile." This facilitated testing of STEEL-R and was used to create a demo that highlighted the ability to map data from multiple training systems to a single competency framework and to generate a display that team leaders can use to personalize and optimize training across multiple training modalities. This paper gives an overview of STEEL-R, its architecture, and the features that enabled the use of artificial data. The paper explains how xAPI statements are converted to assertions and how these are used to estimate trainee competency. This is followed by a section on xAPI Profiles and on the xAPI Profile used in STEEL-R. The paper then discusses how artificial data were generated and the challenges of modeling longitudinal development and team in these data. The paper ends with a section on future research. [For the full proceedings, see ED623995.]

With the approach of constant changes and quality assurance in education, we have reached an optimum that no longer justifies all further investments in such changes, as the results of these investments are (and will be) minimal and insufficient. We have reached a stage where we must shift from evolution to revolution, from constant changes in education to its complete transformation. Here, we must point out that we must reverse the flow of systemic changes from the dehumanization of society as that in Industry 4.0 or, in a slightly softer form, the Japanese vision of Society 5.0. This reverse flow offers us the re-humanization of society's development and it can be called Society 6.0 or, historically, also Society 1.1 (back to the past, to the first industrial revolution). [For the full proceedings, see ED629086.]

With the rapid advancement of technology, education is undergoing a transformational change. Chatbots have become increasingly popular in recent years and are being utilized as teaching assistants to support teachers and students in various ways. However, little research has been done on the skills required by teachers to prepare curriculum content using chatbots. The research aims to identify the skills teachers need to prepare curriculum content with chatbots. It examines the fundamental and cognitive skills individuals need to interpret content generated by chatbots and explores the difference between self-assessment and evaluator-based assessment. Fifty-eight third-year students, pre-service teachers, in the Elementary education program attempted to write a lesson plan using ChatGPT and completed a questionnaire to assess the skills required. Their communication with the chatbot as well as their prepared lesson plans were reviewed by an evaluator who rated the skills of the participating pre-service teachers. Results indicate that preservice teachers tend to overestimate their skills required to interpret chatbot-generated content compared to the evaluator's ratings. Such discrepancies could lead to inaccurate or incomplete assessments of their skills, which could hinder their potential for growth and development. [For the full proceedings, see ED629086.]

Past research has yielded ample knowledge regarding the design of analytics-based tools for teachers and has found beneficial effects of several tools on teaching and learning. Yet there is relatively little knowledge regarding the design of tools that support teachers when a class of students uses AI-based tutoring software for self-paced learning. To address this challenge, we conducted design-based research with 20 middle school teachers to create a novel real-time dashboard, Tutti, that helps a teacher monitor a class and decide which individual students to help, based on analytics from students' tutoring software. Tutti is fully implemented and has been honed through prototyping and log replay sessions. A partial implementation was piloted in remote classrooms. Key design features are a two-screen design with (1) a class overview screen showing the status of each student as well as notifications of recent events, and (2) a deep dive screen to explore an individual student's work in detail, with both dynamic replay and an interactive annotated solution view. The project yields new insight into effective designs for a real-time analytics-based tool that may guide the design of other tools for K-12 teachers to support students in self-paced learning activities. [This paper was published in: "Proceedings, European Conference on Technology Enhanced Learning, EC-TEL 2022. Lecture Notes in Computer Science, vol 13450," edited by I. Hilliger et al., Cham: Springer, 2022, pp. 16-30.]

Automated scoring of student language is a complex task that requires systems to emulate complex and multi-faceted human evaluation criteria. Summary scoring brings an additional layer of complexity to automated scoring because it involves two texts of differing lengths that must be compared. In this study, we present our approach to automate summary scoring by evaluating a corpus of approximately 5,000 summaries based on 103 source texts, each summary being scored on a 4-point Likert scale for seven different evaluation criteria. We train and evaluate a series of Machine Learning models that use a combination of independent textual complexity indices from the ReaderBench framework and Deep Learning models based on the Transformer architecture in a multitask setup to predict concurrently all criteria. Our models achieve significantly lower errors than previous work using a similar dataset, with MAE ranging from 0.10-0.16 and corresponding R[superscript 2] values of up to 0.64. Our findings indicate that Longformer-based models are adequate for contextualizing longer text sequences and effectively scoring summaries according to a variety of human-defined evaluation criteria using a single Neural Network. [This paper was published in: "AIED 2022, LNCS 13355," edited by M. M. Rodrigo et al., Springer Nature Switzerland, 2022, pp. 756-761.]

Providing timely feedback is crucial in promoting academic achievement and student success. However, for multifarious reasons (e.g., limited teaching resources), feedback often arrives too late for learners to act on the feedback and improve learning. Thus, automated feedback systems have emerged to tackle educational tasks in various domains, including novice programming, short-essay writing, and open-ended questions. However, to the best of our knowledge, no previous study has investigated automated feedback generation on students' project reports. In this paper, we present a novel data-driven system, named "Insta-Reviewer," for automatically generating instant feedback on students' project reports, using state-of-the-art natural language processing (NLP) models. We also propose a framework for manually evaluating system-generated feedback. Experimental results show that feedback generated by Insta-Reviewer on real students' project reports can achieve near-human performance. Our work demonstrates the feasibility of automatic feedback generation for students' project reports while highlighting several prominent challenges for future research. [For the full proceedings, see ED623995.]

Doing assignments is a very important part of learning. Students' assignment submission time provides valuable information on study attitudes and habits which strongly correlate with academic performance. However, the number of assignments and their submission deadlines vary among university courses, making it hard to use assignment submission time as a feature to predict academic performance. In this paper, we propose a new method called Relative Assignment Submission Time (RAST) which uses the assignment submission information of peer students to improve the correlation with course grades. Experiments on real-life data of 20 courses show that RAST has a high correlation with students' academic performance. We also build a machine learning model using RAST as a feature to detect students who would suffer from poor grades. Our method outperforms the traditional method by up to 61% on f1-score. We believe that our proposed method can help other studies on assignment submission time to improve the prediction accuracy on academic performance and detecting at-risk students. [For the full proceedings, see ED623995.]

Automatic short answer grading is an important research direction in the exploration of how to use artificial intelligence (AI)-based tools to improve education. Current state-of-the-art approaches use neural language models to create vectorized representations of students responses, followed by classifiers to predict the score. However, these approaches have several key limitations, including i) they use pre-trained language models that are not well-adapted to educational subject domains and/or student-generated text and ii) they almost always train one model per question, ignoring the linkage across question and result in a significant model storage problem due to the size of advanced language models. In this paper, we study the problem of automatic short answer grading for students' responses to math questions and propose a novel framework for this task. First, we use MathBERT, a variant of the popular language model BERT adapted to mathematical content, as our base model and fine-tune it on the downstream task of student response grading. Second, we use an in-context learning approach that provides scoring examples as input to the language model to provide additional context information and promote generalization to previously unseen questions. We evaluate our framework on a real-world dataset of student responses to open-ended math questions and show that our framework (often significantly) outperform existing approaches, especially for new questions that are not seen during training. [For the full proceedings, see ED623995.]

Digital Study Assistants (DSA) are an emerging type of software that combines web-based software architectures, various data sources, and algorithms from artificial intelligence (AI) to assist learners in improving their learning-related behaviors. In this paper, we summarize the implementation and results of a field study with a DSA for hierarchical goal-setting (HGS) at the Bremen, Hannover, and Osnabrück universities from November 2021 to April 2022. The results show that 70% of students in the sample chose to get digital assistance for educational goal-setting, which is the highest interest rate among the nine assistance functions available. Of the 290 students who chose to use the assistant, only 10 completed the full assistive intervention, which equals only 3.4%. We conclude that we should improve the usability and user experience and reduce the interaction costs of the intervention.

Digital resources offer a vast assortment of educational opportunities for students in higher education. From 2018 to 2022, a digital study assistant (DSA), named SIDDATA, was developed at three German universities and consequently field-tested. One of the DSA's features is an AI-driven natural language interface for educational resource recommendation. This paper performs an analysis of the effectiveness of recommendations, by analyzing data generated over the course of two years of DSA usage. We find that although initial user interest is high, only a small percentage of users engage with the recommendation feature. Furthermore, we find that quality of recommendations was perceived as mixed to negative.

Educational chatbots promise many benefits for teaching and learning. Although chatbot use cases in this research field are rapidly growing, most studies focus on individual users rather than on collaborative group settings. To address this issue, this paper investigates how chatbot-mediated learning can be designed to foster middle school students in team-based assignments. Using an educational design research approach, quality indicators of educational chatbots were derived from the literature, which served as a guideline for the development of the chatbot Tubo (meaning tutoring bot). Tubo is part of a web-based team learning environment in which students can chat with each other and collaboratively work on their group assignments. As a team member and tutor of each group, Tubo guides the students through the learning journey by different scaffolding elements and helps with content-related questions the students have. As part of a first design cycle, the chatbot application was tested with a school class of a technical vocational school in Switzerland. The received feedback suggests that the approach of team-based learning with chatbots has a lot of potential from the students' and teachers' point of view. However, the role distribution of the individual group members may have to be further specified to address the different needs of autonomous as well as more control-oriented students.

The use of social robots in education is a growing area of research and the potential future applications are various. However, the conversational models behind current social robots and chatbot systems often rely on rule-based and retrieval-based methods. This limits the social robot to predefined responses and topics, thus hindering it from fluent communication and interaction. Generative language models such as GPT-3 could be beneficial in this context, e.g. for an improved conversation and open-ended question answering. This article presents an approach to utilizing generative language models to enhance interactive learning with educational social robots. The proposed model combines the technological possibilities of generative language models with the educational tasks of a social robot in the role of a tutor and learning partner. The implementation of the model in practice is illustrated by means of a use case consisting of different learning scenarios. The social robot generates explanations, questions, corrections, and answers based on the pre-trained GPT-3 model. By exploring the potential of generative language models for interactive learning with social robots on different levels of abstraction, the paper also aims to contribute to an understanding of the future relevance and possibilities that generative language models bring into education and educational technologies in general.

Due to the advances of artificial intelligence (AI) and natural language processing, new kinds of Internet-based writing tools have emerged. Among other things, these AI-powered writing tools can be used by students for text translation, to improve spelling or for rewriting and summarizing texts. On the one hand, they can provide detailed recommendations for the adaptation of text elements within seconds. On the other hand, they also produce inconsistencies and errors, that students might not be aware of. How to deal with these tools in an educational context is a difficult question. Since writing tools are usually used unsupervised and without further instructions, students may need guidance from the teacher in interacting with those tools, to prevent the risk of misapplication. To better understand this underlying issue, the paper at hand uses survey data of 365 freshmen students to describe and analyze student perceptions of AI-powered writing tools. Regarding AI-powered writing tools, different student types were identified by using the k-means clustering method. The results suggest that students have different attitudes towards AI-powered writing tools. Some students may use them in an unreflective, non-sceptical way, which can lead to (un)voluntary plagiarism. Other students may not use writing tools at all, partly because they are sceptic, but also because they may lack meaningful learning strategies in general (course repeaters). Depending on the different student types, individualized teaching strategies might be helpful to promote or urge caution in the use of these tools.

This paper deals with a solution allowing digital universities to extend the functionalities of their distance learning platform to offer a secure solution for the dematerialization of assessments. Currently we are witnessing the rise of digital universities, this is the case in Africa, particularly in Senegal. We are witnessing strong growth in the number of students, in a context of extension and diversification of training offers. This is the case of the Virtual University of Senegal (UVS), the number of students has increased from 2,000 students in 2013, the year of its creation, to 50,000 in 2021. It offers 13 licenses and 30 courses of training. With these large numbers, the organization of assessments in these universities becomes more and more tedious. Taking the example of the UVS with 50,000 students and 30 training courses, we will have to deploy millions of exams copies due to one copy per candidate. These universities have digital campuses (connected campuses) or Open Digital Spaces (ENO) which make it possible to organize face-to-face evaluations on the table. This organization has several disadvantages, on the one hand, the management of the proofs and the correction of the copies require the mobilization of human and financial resources; on the other hand, the risks of errors, reports and authenticity of the notes. In this article, we propose a secure system for managing online assessments in digital universities based on LMS Moodle, SEB and remote monitoring with the JITSI video conferencing system. The solution will allow universities to optimize human and financial resources and make assessment results more reliable.

Despite calls to increase the focus on explainability and interpretability in EDM and, in particular, student success prediction, so that it becomes useful for personalized intervention systems, only few efforts have been undertaken in that direction so far. In this paper, we argue that this is mainly due to the limitations of current Explainable Artificial Intelligence (XAI) approaches regarding interpretability. We further argue that the issue, thus, calls for a a combination of AI and social science methods utilizing the strengths of both. For this, we introduce a step-wise model of interpretability where the first step constitutes of knowing important features, the second step of understanding counterfactuals regarding a particular person's prediction, and the third step of uncovering causal relations relevant for a set of similar students. We show that LIME, a current XAI method, reaches the first but not subsequent steps. To reach step two, we propose an extension to LIME, Minimal Counterfactual-LIME, finding the smallest number of changes necessary to change a prediction. Reaching step three, however, is more involved and additionally requires theoretical and causal reasoning - to this end, we construct an easily applicable framework. Using artificial data, we showcase that our methods can recover connections among features; additionally, we demonstrate its applicability on real-life data. Limitations of our methods are discussed and collaborations with social scientists encouraged. [For the full proceedings, see ED623995.]

Machine learning is a powerful method for predicting the outcomes of interactions with educational software, such as the grade a student is likely to receive. However, a predicted outcome alone provides little insight regarding how a student's experience should be personalized based on that outcome. In this paper, we explore a generalizable approach for resolving this issue by personalizing learning using explanations of predictions generated via machine learning explainability methods. We tested the approach in a self-guided, self-paced online learning system for college-level introductory statistics topics that provided personalized interventions for encouraging self-regulated learning behaviors. The system used explanations generated by SHAP (SHapley Additive exPlanations) to recommend specific actions for students to take based on features that most negatively influenced predicted learning outcomes; an "expert system" comparison condition provided recommendations based on predefined rules. A randomized controlled trial of 73 participants (37 expert-system condition, 36 explanation condition) revealed similar learning and topic-choosing behavior between conditions, suggesting that XAI-informed interventions facilitated student statistics learning to a similar degree as expert-system interventions. [For the full proceedings, see ED623995.]

How can we test whether state-of-the-art generative models, such as Blender and GPT-3, are good AI teachers, capable of replying to a student in an educational dialogue? Designing an AI teacher test is challenging: although evaluation methods are much-needed, there is no off-the-shelf solution to measuring pedagogical ability. This paper reports on a first attempt at an AI teacher test. We built a solution around the insight that you can run conversational agents in parallel to human teachers in real-world dialogues, simulate how different agents would respond to a student, and compare these counterpart responses in terms of three abilities: speak like a teacher, understand a student, help a student. Our method builds on the reliability of comparative judgments in education and uses a probabilistic model and Bayesian sampling to infer estimates of pedagogical ability. We find that, even though conversational agents (Blender in particular) perform well on conversational uptake, they are quantifiably worse than real teachers on several pedagogical dimensions, especially with regard to helpfulness (Blender: [delta] ability = -0.75; GPT-3: [delta] ability = -0.93). [For the full proceedings, see ED623995.]

Considerable amount of research in educational data mining has focused on developing efficient algorithms for Knowledge Tracing (KT). However, in practice, many real-world learning systems used at scale struggle to implement KT capabilities, especially if they weren't originally designed for it. One key challenge is to accurately label existing items with skills, which often turns out to be a herculean task. In this paper, we investigate whether an increasingly popular approach to knowledge tracing, the use of neural network models, can be a partial solution to this problem. We conducted a case study within a commercial math blended learning system. Using the data collected from middle school students' use of the system over two years, we compare the performance of a neural network-based KT model (DKVMN) in three scenarios: 1) with the original (possibly unreliable) system-provided skill tags, 2) with coarser-grained domain tags based on state standards, and 3) without inputting any mappings between content and skills. Our results suggest that including the system-provided skills in the training of the model leads to the worst performance. The best performance is observed when the skills are entirely disregarded. This supports the possibility of bypassing the laborious step of item-skill tagging in real-world learning systems which were not originally designed to work with KT models, especially if the goal is only to predict the performance of a student on future items. We discuss the implications of our findings for practice and future research. [For the full proceedings, see ED623995.]

Regular expression (regex) coding has advantages for text analysis. Humans are often able to quickly construct intelligible coding rules with high precision. That is, researchers can identify words and word patterns that correctly classify examples of a particular concept. And, it is often easy to identify false positives and improve the regex classifier so that the positive items are accurately captured. However, ensuring that a regex list is complete is a bigger challenge, because the concepts to be identified in data are often sparsely distributed, which makes it difficult to identify examples of "false negatives." For this reason, regex-based classifiers suffer by having low recall. That is, it often misses items that should be classified as positive. In this paper, we provide a neural network solution to this problem by identifying a "negative reversion set," in which false negative items occur much more frequently than in the data set as a whole. Thus, the regex classifier can be more quickly improved by adding missing regexes based on the false negatives found from the negative reversion set. This study used an existing data set collected from a simulation-based learning environment for which researchers had previously defined six codes and developed classifiers with validated regex lists. We randomly constructed incomplete (partial) regex lists and used neural network models to identify negative reversion sets in which the frequency of false negatives increased from a range of 3%-8% in the full data set to a range of 12%-52% in the negative reversion set. Based on this finding, we propose an interactive coding mechanism in which human-developed regex classifiers provide input for training machine learning algorithms and machine learning algorithms "smartly" select highly suspected false negative items for human to more quickly develop regex classifiers. [For the full proceedings, see ED623995.]