Abstracts Track 2024


Area 1 - Artificial Intelligence in Education

Nr: 183
Title:

Eye-Tracking Comparative Study: ChatGPT versus Traditional Text Reading Behaviours

Authors:

Hayley Weigelt and Omri Kahana

Abstract: The pervasive integration of artificial intelligence (AI) technologies in Education, notably exemplified by ChatGPT, has significantly influenced contemporary human interactions. This study investigates how students in higher education refer and read informative and computing code texts presented as written by ChatGPT versus written by a human expert, using an eye-tracking device and a participant questionnaire. The study was held at the “User Experience Lab” of the faculty of Instructional Technologies, the Holon Institution of Technology. Eye-tracking data was collected using the “Tobii Pro-Lab” with a 120 Hz sampling frequency. The study included 20 students, aged 20-30, from the Institute. Participants were requested to read four texts, equal by subject matter and wording: two informative texts in Hebrew and two programming language texts. Each couple of texts included one text presented as generated by the ChatGPT and the other as generated by a human specialist. The study employs a mixed-methods approach, leveraging both quantitative and qualitative methodologies to comprehensively investigate reading behaviours. Eye-tracking data will be analysed using “Tobii Pro-Lab” software, focusing on Areas Of Interest (AOI) within the textual content. Initial analyses of eye-tracking data from a subset of participants reveal notable differences in fixation patterns between ChatGPT responses and traditional text documents. The preliminary findings of this study, which will be presented fully at the conference, underscore the complex interplay between text format and cognitive engagement, highlighting the need for empirical inquiry into reading while using AI interactions in the field of Education.

Nr: 214
Title:

Leveraging ChatGPT for Intermediate Korean Learners: Exploring Writing Instruction and Ethical Considerations

Authors:

Hye Jin Agnes Ryoo

Abstract: This study presents a novel approach to writing instruction tailored for intermediate Korean learners, employing ChatGPT. Twelve students from N University participated in writing tasks utilizing ChatGPT for intermediate-level writing. Participants were tasked with composing explanatory essays on predetermined topics within a 60-minute time limit, producing texts ranging from one to two pages on A4 paper. The entire writing process, including information exploration using ChatGPT, was recorded and shared via screen-sharing apps. Evaluations were conducted based on predefined criteria derived from TOPIK assessment rubrics and previous research to ensure consistency and fairness. Guidance was provided on leveraging ChatGPT for brainstorming, vocabulary assistance, information summarization, and content revision. Ethical considerations were stressed to discourage excessive reliance on ChatGPT-generated responses and promote independent thinking and language skill development. Evaluation of the participants' work revealed advanced vocabulary usage but occasional awkwardness in sentence structure, often resembling a mechanical or translated style. Despite this, participants demonstrated logical organization in their compositions, suggesting ChatGPT's consistent response patterns. Proficient English speakers were able to refine and adapt ChatGPT-generated content for coherence and quality enhancement. In conclusion, while ChatGPT offers valuable support throughout the writing process, its integration with traditional teaching methods is essential to prevent dependency and ensure enduring learning outcomes. The overarching goal is to nurture learners' creativity, critical thinking, and language proficiency through ChatGPT-integrated writing instruction.

Nr: 219
Title:

CleverFox: A Writing Tools that Exploits Generative AI

Authors:

Yung Chu Chiang, Tzu Chien Lin, Yi Jen Hwang, Cheng En Hsu, Pin Jie Huang, Zi Xian Tang, Yi Ching Luo, Jo Chi Hsiao, Hai Lun Tu, Jason S. Chang and Yasunari Harada

Abstract: In recent years, the government in Taiwan has introduced a so-call Bilingual Country 2030 policy to to improve English education in universities and high-school to better prepare students to compete in a globalized world. Out of the four language skills, teaching writing is most challenge for English teachers with a heavy work load for essay grading and feedback. Some estimated that it takes an English teacher with three classes about 45 hours to correct one English essay assignment. Students do not have sufficient opportunity to practice writing, because teachers could not give many essay assignments and provide meaningful feedback. This paper describe the development of an English writing assistant CleverFox (https://cleverfox.nlplab.cc/) for English teachers and students, aiming at help students to improve their English writing skills. The system provides the following functions: 1. Intelligent Feedback and Assessment: CleverFox is equipped with automatic grammar correction and word choice improvement capabilities to help students to fix writing errors, to level-up word choices, and to review the linking phrases and structure of their essays. In addition, teachers can use the "Learning Box" to can see what students have written and system feedback, and then give additional feedback. 2. Intelligent Agents in CleverFox provides the ability to ask questions and receive feedback. Students can get real-time assessment and suggestions to help them improve their writing skills. 3. Teaching Assistant in CleverFox provides a chatbot to act as a teaching assistant. Teachers can set a number of parameters of the chatbot, such as the content of the lesson, lession preview questions, answers, tone of voice, number of words in the reply, so that the chatbot can better meet the needs of teachers and provide personalized learning support. 4. With Learning Box, Teachers can track student learning through the Learning Box and provide individualized feedback. CleverFox combines artificial intelligence and natural language processing technologies to provide teachers and students with a comprehensive and customized English writing tool. We leverage the knowledge capacity of ChatGPT with 45TB of training data, its powerful conversation functionality. To use ChatGPT more effectively, we employ many principles of Prompt Engineering, a field that has recently gained attention in the realm of generative AI. By combining knowledge of English writing process and Prompt Engineering techniques, we developed CleverFox to assist teachers in reviewing students' English essays. This enables teachers to quickly identify students' writing weaknesses, and also allows students to receive additional and immediate feedback on their English writing. CleverFox has three types of error correction features, namely Grammar Error Correction (GEC), Word Choice Level Up and Transitional Word Analysis. In addition, we also used ChatGPT to design questions for the Teacher Customization section of the quiz and a simulated Chatbot Teaching Assistant. This is a combination of Prompt-Tuning and Few-Shot Learning to create content that meets the needs of most teachers.

Area 2 - Information Technologies Supporting Learning

Nr: 109
Title:

Creating Inclusive Online Courses: Universal Design for Learning

Authors:

Maureen Andrade

Abstract: Students in higher education institutions come from a range of backgrounds with differing home, educational, and life experiences (American Council on Education, 2023). As such, they approach learning in different ways. To accommodate these learning differences and create an inclusive environment in which all students can succeed, instructors must design courses that provide students with opportunities to demonstrate their mastery of concepts in a variety of ways. A guiding framework for inclusive course design is Universal Design for Learning (UDL) (CAST, 2023). The underlying principles of this framework encourage the use of multiple learning pathways to present course concepts and give students assignment and assessment choices (Miller and Humphreys, 2023). These approaches can help mitigate the completion gap for students from diverse socioeconomic, racial, ethnic, educational, and cultural backgrounds (Custer, 2023; National Center for Education Statistics, 2019). The UDL framework is based on the why, what, and how of learning and consists of three categories: engagement, representation, and action and expression (CAST, 2023). Engagement focuses on motivating learners, representation on alternative ways to present content, and action and expression on providing a variety of ways for learners to give evidence of learning. Engagement might entail indicating completion time, offering coping strategy tips, providing assignment rationale statements to make learning purposes more salient, or using a standardized template in the learning management system (Miller and Humphreys, 2023). Providing choice in content delivery such as slides, videos, or article extracts aligns with the principle of representation. Offering students assessment choices such as quizzes, written papers, or graphic representations of knowledge supports the action and expression principle. Technology in online courses can support these UDL elements and offer opportunities for instructional innovations. The current study examined how course design for an online introduction to organizational behavior course was guided by UDL principles to engage students and support learning (CAST, 2023). The design goal for the course was to support a range of learning approaches and provide multiple pathways for students to illustrate their knowledge. Surveys consisting of four open-ended questions were analyzed for insights into students’ learning experiences. 1. What aspects of the course help you learn best? 2. Is anything confusing or unclear? If so, explain. 3. What, if anything, would you like to see change between now and the end of the semester? 4. What advice would you give your instructor to improve your learning in this course? UDL principles encourage multiple ways to engage students, present content, provide choice, and diversify assessment. UDL-facing course design and teaching are good for all students and with appropriate institutional support, can be effectively adopted to improve learning outcomes. With the increasing diversity of students in higher education institutions and gaps in their success rates, creating innovative pedagogies to support academic achievement and completion is crucial. UDL lends itself to responsive and inclusive teaching and learning practice. Examples of assignments for the course involved in the study will be shared as well as findings from the survey analysis.

Nr: 207
Title:

Hybrid Learning Dimensions: First Results of a Systematic Literature Review

Authors:

Tetyana Vereshchahina, Bernadette Charlier and Paola Costa Cornejo

Abstract: The notions of “Hybrid Learning” (HL) or “Blended learning” (BL) became umbrella terms for a relevantly new education environment encompassing on-campus and online learning. The scientists admit the gaps in understanding of the phenomenon among the researchers (Peltier & Seguin (2021), Bozkurt (2021)). These scientific gaps lead to the problem of the validity of the research as the terms are used interchangeably, although the contexts of their application differ. To address them, it is necessary to study the research phenomenon based on common and negotiated dimensions, which will help have more consensus on definitions. That is why, the present research aims to approach the problem of hybrid learning dimensions expressed in the definitions of BL and HL. A systematic literature review was conducted in 2022-2024 for the period of the past five years (2018-2023). We searched in WEB OF SCIENCE, SCOPUS, and ERIC. Snowball searching was also completed in GOOGLE SCHOLAR, SieLo, and CAIRN as well. The 3138 blended and hybrid learning-related articles in English, French, and Spanish were identified. We applied a textual narrative synthesis approach to select, group, and interpret results. The present contribution is a part of the entire systematic review which includes more research questions. Here, we raise the following question: What are the common dimensions of hybrid learning in the selected definitions? Among the identified peer-reviewed papers, there were included, extracted, and assessed 39 papers including extended definitions explicitly explaining what HL or BL is. The included definitions were analysed according to the categories or dimensions defined by HY-SUP researchers (Peraya, Charlier et Deschryver, 2014). They define such dimensions for hybrid learning as 1)articulation of present/distant (active participation of students in presence and distance); 2) mediation (the course aims to ensure that students learn to collaborate, learn the process of learning); 3) mediatization (provision of learning support tools; provision of management, communication, and interaction tools; multimedia resources.); 4) guidance (methodological, metacognitive support and students learning guidance); 5) openness (a wide choice of learning activities.; external resources or experts in your course). In the process of categorising, there were also defined two additional dimensions: 7) learning design in the definition (an explicit statement that HL or BL is a designed dispositive not just a combination of different elements); 8) learning evaluation (evaluation component included in the learning system). According to the results, all the definitions (n=39) embraced the online and offline components of HL. More than half of the definitions included (n=27) learning mediatization, and almost one-third (n=14) included the mediation component. The component of teaching guidance was mentioned in (n=9) definitions, openness in (n=17 ) the component of design in (n=7) definitions, and the evaluation was in (n=4) definitions. The results show that the concept of the HL is complex and multidimensional. The researchers have to consider the dimensions they include in the definitions and teaching practices they research. The definitions may differ, but the included dimensions will be precise and make them clear. This alignment of the definitions with the practices analyzed in the research would facilitate the understanding of HL and BL as a research phenomenon.

Nr: 221
Title:

“Campus Changer”: The Unity-Based Serious Game as a Digital Twin Model of the Warsaw University of Technology Main Campus Supporting Geoparticipation Process

Authors:

Agnieszka Wendland, Robert Olszewski, Kornel Samociuk, Anna Szalwa and Urszula Szczepankowska-Bednarek

Abstract: Warsaw University of Technology (WUT) is one of the best technical universities in Poland, with almost two hundred years of tradition and history, and more than ten thousand students. The Main Campus is located in the city center but can be characterized as an unintegrated space dominated by car traffic and car parks. There is a strong need to revitalize the Campus, respecting the tradition and history, but also citizens' opinions. The area needs to provide a sense of community of urban space and ideas. The experiment aims to develop an innovative tool to support the process of public participation using gamification methodology. The developed serious game, using the UNITY game engine and 3D models of a part of the city (Warsaw University of Technology Main Campus), allows the collection of user data. This data relates to the preferences of the academic community regarding the use of open space, buildings, etc., as well as preferences for their revitalization. The users will have the opportunity to test and simulate potential changes in the virtual space before the real/physical space is transformed. A major advantage of the game is the ability to observe proposed changes in real-time. The collected data is then processed, which allows to develop a general model of social preferences regarding spatial development. The indirect purpose of the experiment is to provide a safe, comfortable environment for cooperation, making a “culture” of cooperation important, which can be created in the process. There is a need for cross-generation inputs that will cover different perspectives on the future shape of this important urban space. The use of gamification methodology (serious game), digital twin (based on 3D models of the city) for community data collection, and artificial intelligence for data processing will strengthen the process of community involvement and public participation. Data is collected on the WUT server in a PostgreSQL+PostGIS relational database. Open-source tools are used to process the data: QGIS and the Python programming language and programming libraries dedicated to machine learning, such as PyTorch. After data collection and processing, the final results will be evaluated and discussed with stakeholders from the Warsaw University of Technology and the Municipality of Warsaw. The methodology can be easily generalized to other academic campuses or cities. The conducted methodology will be open to other universities that would like to perform similar analyses on their campuses in the future, but also cities interested in the geoparticipation process, and the possibility of applying the tool in the urban planning process.

Area 3 - Learning/Teaching Methodologies and Assessment

Nr: 222
Title:

Effects of Dynamic Multiple Representations via Simulations on Student Learning in Physics

Authors:

Wei Jia

Abstract: Teaching and learning physics is often challenging, especially when it requires students to mentally visualize abstract or complex concepts. To address this difficulty, physics teachers have attempted to use many visual representations, including graphs and drawings, to assist in students’ physics learning and understanding. However, inappropriate use of such representations might impose an additional cognitive burden on students. In the present study, we use GeoGebra, an educational software tool, to systematically design the simulation with dynamic multiple representations (DMR) to support physics teaching and learning. This research aims to: (a) evaluate the effectiveness of changing visualized representations dynamically during physics learning through GeoGebra, for both teaching and learning; and (b) develop a framework for representing various types of concepts and models in physics. In the main study, we will use a quasi-experimental dose-response mixed research design, involving six secondary physics teachers and their 18 classes of students, to examine the effects of DMR on physics classroom teaching and student learning outcomes. Data in the forms of video recordings, test papers, self-reported questionnaires, and interviews will be collected, analyzed, and synthesized. This investigation has the potential to identify the effectiveness of using DMR for supporting physics teaching and learning. It may also inform the design and innovative use of educational technologies (e.g., GeoGebra) for physics teachers and students.