Invited Speakers
Speaker: Prof. Yukari Shirota
Affiliations: Gakushuin University, Japan
Prof. Yukari Shirota
Biography: Prof. Yukari Shirota earned her D.Sc. in Computer Science from the University of Tokyo in 1998. After 13 years of industry research, she joined the Faculty of Economics at Gakushuin University in 2001 and was promoted to full professor in 2002. She was an academic visitor at the University of Oxford from 2006 to 2007. She is a Fellow of the Information Processing Society of Japan. She serves on the boards of the Japan Society of Management Mathematics. Her research focuses on industry analysis using AI, web-based data visualization, and visual learning in business mathematics. Recently, she has been analyzing how Maruti Udyog (Maruti Suzuki) successfully developed its local supplier network in India. She also explores the educational applications of AI. In her research, she shows how ChatGPT can solve complex financial mathematics word problems and generate educational graphics. By visualizing the deductive reasoning process as a graph, AI-generated teaching aids help students grasp concepts and verify their understanding. This method has already been successfully applied in her classes.
Speech Information
Abstract:Does generative AI solve mathematics problems correctly? Yes. But does it solve them the way a human teacher should teach? To answer this question, this talk introduces the Deductive Reasoning Graph — a tool that automatically visualises ChatGPT's reasoning steps as a directed graph using Python/Graphviz. For the first time, this makes it possible to directly observe what is happening inside the AI's "mind." Through two concrete mathematics problems — an absolute-value quadratic equation and a circle-and-line intersection problem — our analysis consistently shows that while AI always reaches the correct answer, it does so through exhaustive algebraic procedures that are fundamentally different from the visual and intuitive strategies that human learners naturally use. Importing AI-generated solution paths directly into the classroom, without critical examination, risks undermining the development of students' mathematical thinking.
Speaker: Prof. Shinobu Hasegawa
Affiliations: Japan Advanced Institute of Science and Technology, Japan
Prof. Shinobu Hasegawa
Biography: Shinobu Hasegawa is currently a professor and director at the Center for Information Infrastructure, Japan Advanced Institute of Science and Technology (JAIST). He received his B.S., M.S., and Ph.D. degrees in systems science from Osaka University in 1998, 2000, and 2002, respectively. The primary goal of his research is to facilitate “Human Learning and Computer-mediated Interaction” in a distributed environment. His research field is mainly learning informatics which includes support for Learning 5.0, affective learning, Web-based learning, game-based learning, language learning, cognitive skill learning, distance learning system, and community-based learning.
Speech Information
Abstract:In the context of the human-cantered "Society 5.0," the "Learning 5.0" paradigm redefines learning as a co-evolutionary process between humans and generative AI. It shifts the educational focus from traditional knowledge transmission to metacognitive enhancement and creative collaboration. However, integrating AI in education presents critical challenges, including "hallucinations," a lack of explainability, and the risk of inducing "metacognitive laziness," which diminishes learner autonomy. It also complicates evaluation reliability and educational equity. To address these issues, the Learning 5.0 paradigm treats knowledge not as fixed facts, but as a dynamic process generated through human-AI dialogue, shifting evaluation from simple output correctness to the "quality of the learning process.” One of the practical applications of this paradigm include a hybrid metacognitive support platform that prevents AI over-reliance by improving students' metacognitive skills through dialogue. In this talk, we will also present the current status of human-centered learning support systems that foster learner autonomy and support continuous cognitive/metacognitive growth, as envisioned by Learning 5.0.
Speaker: Asst. Prof. Pattaraporn Warintarawej
Affiliations: Prince of Songkla University, Surat Thani Campus, Thailand
Asst. Prof. Pattaraporn Warintarawej
Biography: Asst. Prof. Dr. Pattaraporn Warintarawej is a faculty member at the Faculty of Science and Industrial Technology, Prince of Songkla University (PSU), Surat Thani Campus, where she has served since 2006. She holds a Ph.D. in Computer Science from the University of Montpellier 2, France (2013). Her research and teaching interests span Web Programming, Data Science for Business, Digital Marketing, and Machine Learning.
Speech Information
Abstract:This paper presents CodeGuru AI, an LLM-powered Intelligent Tutoring System (ITS) for front-end web programming practice. The system leverages the Google Gemini API (gemini-2.5-flash) to dynamically generate coding exercises across five technology domains---HTML, CSS, JavaScript, TailwindCSS, and ReactJS---at three difficulty levels. Learners receive immediate AI-driven feedback on code submissions, while an AI Tutor Chat provides Socratic-style guidance. An administrative dashboard enables instructors to monitor engagement and export learner data. A formative deployment with 66 undergraduate students yielded a mean score of 16.68/20 (83.4%) for the HTML/CSS/TailwindCSS portfolio, 5.25/10 (52.5%) for the CSS written examination, and 17.73/20 (88.6%) for the React component, with an overall mean of 39.66/50 (79.3%). These findings demonstrate that an LLM-powered ITS represents a practical and scalable approach to supplementing front-end programming education. Future work will focus on activating the session recording module for longitudinal tracking, integrating adaptive difficulty progression through learner modeling, and extending the system to additional programming domains.
Speaker: Dr. Ke-An Cheng
Affiliations: National Taipei University of Education, Taiwan
Dr. Ke-An Cheng
Biography: An elementary school teacher in Taiwan with a master’s degree in information education, focusing on the application of AI in education, especially in K–12 learning environments. Main interests include AI-supported learning, digital literacy, emerging technologies, and game-based learning design, with occasional teaching experience in in-service master’s program courses in information education, particularly in the educational application of RPG Maker.
Speech Information
Abstract: Problem-posing has been recognized as an effective strategy for fostering higher-order thinking and knowledge construction. However, in university-level programming courses, its high cognitive demands often cause novice learners to encounter difficulties during the problem-posing process, accompanied by frustration and negative emotions, which may hinder sustained engagement and overall learning performance. Prior studies have suggested that guided problem-posing strategies and AI-assisted learning can support learners by lowering learning barriers and enhancing performance. Nevertheless, most existing problem-posing learning systems primarily focus on cognitive scaffolding and have limited capability to capture learners’ emotional states in real time, as well as a lack of mechanisms for responding to and regulating learning-related emotions. In high cognitive load learning environments, emotional states are closely associated with learning processes and outcomes; insufficient emotional support may lead learners to disengage when encountering difficulties, reducing persistence in learning. To address this issue, this study proposes an AI-guided problem-posing learning system integrated with emotion recognition. The system detects learners’ emotional states in real time and incorporates learning performance to dynamically provide multi-level guidance, including problem-posing clues, worked examples, and reflective prompts. Such adaptive support is designed to assist learners at critical moments of difficulty, thereby enhancing learning effectiveness and promoting self-efficacy. By integrating cognitive scaffolding with emotional regulation mechanisms, this study aims to develop a programming learning model that supports both cognitive and emotional dimensions. A quasi-experimental design will be employed to examine the effects of the proposed system on learning performance, problem-posing quality, self-efficacy, and cognitive load.
Speaker: Dr Robbie Lee Sabnani
Affiliations: Nanyang Technological University Singapore, Singapore
Dr Robbie Lee Sabnani
Biography: Dr Robbie Lee Sabnani is a Lecturer and Program Leader in the English Language and Literature Academic Group at the National Institute of Education, Nanyang Technological University Singapore. She holds a PhD in Applied Linguistics, an MBA, and a Post Graduate Diploma in Education (Distinction). Her instructional and research interests include metacognitive awareness in language learning, oracy skills and processes and the development of teacher expertise. Robbie chairs and manages the institute’s language enhancement and academic discourse skills courses. She teaches higher degree, pre- and in-service courses on professional English, communication skills, oracy development and language studies. She regularly conducts training on speaking and critical thinking for effective communication in academic presentations, as well as workshops on speaking and interview skills. In recognition of her teaching strengths, Robbie received the distinguished Excellence in Teaching Commendation award. Robbie publishes on pedagogical practice and language education research and has presented her work at several conferences. She has been invited to deliver keynote speeches at international conventions and symposiums on language teaching and learning. Her recent publications include book chapters with Cambridge University Press and Springer as well as articles in international journals such as TESOL Quarterly and several others. Her latest co-edited book by Springer, Second Language Oracy Development for Young Learners: Listening, Speaking and Thinking, is expected to be published in the second half of 2026. For her research excellence and publications of outstanding quality, Robbie was awarded the prestigious Dean’s Commendation for Research.
Speech Information
Abstract: coming soon...
Speaker: Dr Andrew B. Williams
Affiliations: The Citadel, The Military College of South Carolina, Charleston, USA
Dr Andrew B. Williams
Biography: Andrew B. Williams, Ph.D., is Assistant Provost for Innovation and Entrepreneurship and former Dean of Engineering at The Citadel, The Military College of South Carolina, where he founded and directs the Center for AI, Algorithmic Integrity, and Autonomy Innovation (AI3). He earned his B.S. and Ph.D. in electrical engineering, with an emphasis in artificial intelligence, from the University of Kansas, and his M.S. in electrical and computer engineering from Marquette University, where he held the John P. Raynor Distinguished Chair and directed the Humanoid Engineering and Intelligent Robotics (HEIR) Lab. His research spans autonomous agents and multi-agent systems, humanoid robotics, human-robot interaction, and biomedical and educational applications of AI. He led the research and development for international RoboCup robotics teams, the “world cup” of robotics and AI, and has built humanoid robots ranging from teen-sized soccer players to assistive co-robots for workers with intellectual and developmental disabilities. He has authored and presented over 100 publications and invited talks, authored Out of the Box: Building Robots, Transforming Lives, and has secured research funding from the NSF, NIH, NASA, Google, Apple, and Boeing. He has worked in industry at Apple, GE, and Boeing. He is a former IEEE Distinguished Speaker, an inaugural AWS Education Champion, and a Business Insider Cloudverse 100 honoree. He completed the Massachusetts Institute of Technology (MIT) Professional Education programs in Applied Agentic AI and Applied Generative AI and now leads faculty AI capacity-building and agentic-AI initiatives in higher education.
Speech Information
Abstract: Engineering education was already adapting to generative AI when the ground shifted again. We are moving from chatbots that answer questions to agentic AI, meaning autonomous, goal-directed systems that plan, use tools, and coordinate as multi-agent teams. What should engineering programs do to produce graduates who can build, direct, and critically evaluate these systems? Drawing on more than two decades of work in autonomous agents and multi-agent systems, humanoid robotics, and human-robot interaction, including leading research and development for international RoboCup teams and building assistive robots for workers with intellectual and developmental disabilities, Dr. Williams traces a path from chatbots to agents and what it implies for how we teach. He shares practical experience leading faculty AI capacity-building, agentic-AI seminars, and an institution-wide AI action plan, and offers a framework engineering educators can use to evolve courses, curricula, and assessment for an agentic world, without losing sight of responsible and ethical AI.
Speaker: Assoc. Prof. Taha Ertuğrul Kuzu
Affiliations: University of Education Schwäbisch Gmünd, Germany
Assoc. Prof. Taha Ertuğrul Kuzu
Biography: Since October 2023, Dr. Taha Ertuğrul Kuzu has been a Tenure-Track Professor in Educational Sciences (Primary Education) at the University of Education Schwäbisch Gmünd. He studied language and mathematics education at TU Dortmund University (2008–2014) and completed his doctorate in 2018 under the supervision of Prof. Dr. Susanne Prediger, focusing on multilingual learners’ translanguaging processes in mathematics education. From 2018 to 2020, he worked as a primary school teacher and researcher in mathematics and language education. In 2020, he joined the University of Münster as a postdoctoral researcher in the working group of Prof. Dr. Marcus Nührenbörger, focusing on pre-algebraic thinking, translanguaging, and digital media in primary mathematics education (2020–2023). During the winter semester of 2022/23, he served as a visiting professor at the Institute for Mathematics Education (IDM), Bielefeld University. His research focuses on (a) heterogeneity, inclusion, and translanguaging in primary education, (b) the analysis and design of teaching and learning processes, and (c) digital media and AI-supported learning environments. Methodologically, he specializes in qualitative-reconstructive research (Interpretive Interaction Analysis) and design-based research.
Speech Information
Abstract: The rapid emergence of large language models (LLMs) has sparked worldwide discussions about the future of teaching and learning. In policy debates, curriculum development, and public discourse, primary school children are often approached from a strongly protective perspective. As a result, AI use in primary education is frequently restricted to highly structured and teacher-controlled scenarios, reflecting concerns that younger learners may lack the cognitive, linguistic, or critical capacities required to interact independently with AI systems. Consequently, little is known about whether primary school children are actually capable of engaging productively in prompting practices, critically evaluating AI-generated responses, and using large language models as tools for learning and problem solving. This keynote addresses this question by presenting findings from the German design-based research project KI-KoPrim, which investigates AI-supported multilingual learning environments in Grades 3 and 4. Across multiple iterative design cycles, primary school learners interacted with LLMs in mathematical modelling activities, explanatory tasks, and multilingual learning settings. Drawing on classroom videos, screen recordings, chat logs, and qualitative interaction analyses, the project examines how children integrate AI into their problem-solving and meaning-making processes. The findings challenge the widespread assumption that young learners either use AI uncritically or are not yet capable of engaging with it productively. Instead, the analyses reveal complex processes of questioning, negotiation, evaluation, and prompt refinement. Learners do not merely consume AI-generated answers. Rather, they compare AI responses with their own reasoning, identify inconsistencies, reformulate prompts, and gradually develop a more critical understanding of AI-generated knowledge. Particularly productive learning opportunities emerge when AI responses are incomplete, ambiguous, or even incorrect, requiring learners to reflect on both the content and the quality of their own prompts. At the same time, the findings indicate that such productive interactions do not emerge automatically. Rather, they depend on specific instructional conditions, including scaffolded and age-appropriate prompt-engineering support as well as systematically embedded opportunities for reflection. These forms of didactic guidance appear to be essential for enabling learners to move beyond a receptive use of AI towards a more critical, reflective, and purposeful engagement with AI-generated knowledge. Based on these empirical insights, the keynote derives emerging instructional conditions and design principles for AI-supported learning in primary education. Particular attention is given to scaffolded prompting, teacher-guided reflection, collaborative AI use, multilingual learning opportunities, and the importance of prior independent problem solving before interacting with AI. The presentation argues that the educational challenge is not whether primary school children should use large language models, but how learning environments can be designed so that AI becomes a catalyst for reflection, reasoning, and critical thinking rather than a substitute for learners’ own reasoning and problem-solving processes.
Speaker: Prof. Kento Sasano
Affiliations: Okayama University, Japan
Prof. Kento Sasano
Biography: Kento Sasano is a scholar of educational informatics and Machine Information Behavioral Sciences. His research centers on human-AI symbiosis and the design of AI-driven personalized and collaborative learning experiences, with particular emphasis on the “Teach-to-AI” paradigm and its “T.E.A.C.H.” framework, in which learners deepen their own understanding by teaching a generative-AI agent rather than being taught by it. He is the proposer of “Singularity Studies,” an interdisciplinary field examining the prospect of artificial intelligence surpassing human intelligence and its societal and ethical implications, and he works to bridge theory and practice across academia, education, and industry. Through Japan’s Ministry of Education (MEXT) “DX High School” initiative, he has implemented this approach in public high schools using the Copal AI learning agent, and he founded the national student competition “Generative AI Koshien.” He also founded MIBO (the Machine Information Behavior Observatory), an open initiative for the longitudinal, multi-model observation of how generative-AI systems behave. He graduated from the Faculty of Law at Kyoto University, completed graduate study at the Graduate School of Interdisciplinary Information Studies at the University of Tokyo, and represented Japan at the International Philosophy Olympiad.
Speech Information
Abstract: The prevailing use of generative AI in education casts it as an all-knowing tutor, an “oracle” that answers and explains on demand. Yet that design encourages cognitive offloading and can crowd out real learning. This talk introduces an alternative that reverses the usual direction: rather than being taught by AI, students learn by teaching it. In this “Teach-to-AI” paradigm, explaining a topic to a generative-AI agent is what deepens their understanding. Grounded in the Protégé Effect, knowledge-externalization theory, and Symbiotic Scaffolding Theory, the paradigm is put into practice through the “T.E.A.C.H.” framework: a five-phase cycle of Think (organize what to teach), Explain (articulate it to the AI), Ask back (let the AI question you), Check (evaluate and correct its responses), and Hone (refine the explanation through iteration). A central design principle is “AI-as-Naive-Learner”: the agent is deliberately kept from being too capable, so that students must explain clearly and confront the gaps in their own understanding. Drawing on classroom implementation in Japanese high schools under the Ministry of Education’s “DX High School” initiative, the talk reports early observations, such as students discovering what they do not yet understand the moment they try to teach it. It then connects this pedagogy to the speaker’s Machine Information Behavior Observatory (MIBO), an open project that longitudinally observes how generative-AI systems behave, on the view that understanding how machines behave and rethinking how humans learn are part of the same symbiosis. The talk closes by asking how education can cultivate Homo Symbiosus, learners who grow alongside increasingly capable AI rather than being supplanted by it.