Ulrik Juul Christensen, CEO, Area9 Lyceum
Title: How should the learning architecture of the future look for mathematics?
Abstract: In just a few years, we have found ourselves in a complex situation where tensions run high, and the use of artificial intelligence (AI) in education spans from strict bans and paper-and-pencil exams to euphoric visions of a future where no one ever needs to learn that the square root of 49 is 7. Mathematics education has long been a central focus for experts in learning architecture because it is characterized by a high concentration of conceptual learning objectives—more than in most other subjects, from elementary school to higher education. These learning objectives often need to be mastered individually to achieve cohesive competencies.
But what do we really know about how children and young people learn mathematics? What role do AI tools play in both learning and validating whether learning actually occurs? Are paper-and-pencil exams still relevant, or completely bonkers? And what can we learn from the side effects of many early ed-tech tools that have made too many students even more anxious about math—making them feel stupid and incapable of learning it?
This session explores these questions and aims—based on 15 years of global experience with individualized mathematics learning across all age groups—to offer new perspectives on how we can navigate in a time of AI in learning environments and how we can avoid repeating the mistakes of the past.
Tor Ole Bigton Odden, University of Oslo
Title: ChatGPT as improv artist, blurry JPEG, or conceptual blender: Models for LLM “cognition”
Abstract: Over the last two years, generative AI models like ChatGPT have gone from niche research projects in the field of Natural Language Processing to tools that have upended the sectors of education, code development, social science, and more. This rapid development is partially due to the fact that these tools have, through gradual updates, developed capabilities that could not have been predicted in November 2022, like the ability to reason spatially, mathematically, and visually. Many users reasonably wonder how—and more importantly why—they work. In this talk, I will approach this subject from an educational research perspective, in which we use theoretical frameworks and models to try to explain the features of complex systems like human (and now machine) learning. I will use three theoretical frameworks—the improv artist (also called the stochastic parrot), the blurry JPEG, and the conceptual blender—to highlight different ways of understanding the capabilities of large language models like ChatGPT, and contrast their implications for how we approach and use these tools in education.
Alexandros Sopasakis, Lund University
Title: AI in Mathematics – Exploring Current Problem Solving Capabilities of State of the Art Systems
Abstract: Artificial Intelligence has made significant strides in mathematical problem-solving, evolving from basic applications like Photomath to more advanced reasoning models like ChatGPT. This presentation will explore the best, state-of-the-art, AI systems currently available which are capable of tackling complex mathematical questions, including their underlying technologies, strengths, and limitations. We will also examine the progression of these tools, from Optical Character Recognition (OCR) and Natural Language Processing (NLP) to more recent innovations in reasoning models. Through hands-on demonstrations, participants will experience how AI can assist, or fail, in solving a variety of mathematical problems and discover the potential future applications of AI in both academic and professional settings. We will also take a brief look of future trends of the current technology.
Tue Herlau, Technical University of Denmark
Title: Developing an AI tool for formative assessment
Abstract: This talk examines how AI can be used to provide timely formative feedback on written reports within STEM disciplines. Although AI promises to deliver instant continuous formative feedback to students, the feedback it generates may not always be accurate or align with our expectations of human feedback. This presentation will explore one approach to addressing this challenge through the development of an online formative feedback tool, ChatTutor, which is slated for testing in the fall. The talk will cover strategies for incorporating AI feedback into classroom settings and highlight the essential role of human input in validating AI-generated feedback.
Samantha Nordqvist, Sofia Isgar, Leon Börjesson, Arthur Zeuner and Vittorio Fratti, Lund University
Title: Practical experience with AI tools for learning mathematics
Abstract: At the start of the fall semester, students in an introductory calculus course at Lund University were encouraged to use AI tools to support their learning. However, many reported being reluctant, preferring instead to continue to rely on their limited access to instructors and teaching assistants—often citing negative experiences with AI during high school. To address this, the instructor, at the last minute, revised an early hand-in assignment in mathematical communication, asking students to select a challenging problem they felt unable to solve independently, collaborate with an AI tool, and reflect on the process.
In this session, five students will share their insights on the use of AI in mathematics, drawing on their engagement with this assignment and contrasting it to earlier experiences with the use of AI. They will describe how AI helped them navigate complex concepts, provided personalized explanations, and reduced stress by offering constant, on-demand support. The students will also discuss challenges such as handling incorrect AI responses and finding the right balance between AI assistance and independent learning. These reflections are meant to offer fresh perspectives – directly from the students – on how AI can reshape the learning process.
Louise Meier Carlsen, IT University of Copenhagen
Title: How Generative AI, Advances Students’ Modeling Processes
Abstract: This presentation discusses a study from the IT University’s Data Science program aimed at enhancing students’ modeling competence. It focuses on how generative AI, especially ChatGPT, influenced students’ modeling processes compared to static media like webpages and YouTube. We also introduce the question-and-answer maps and Herbartian Scheme from the Anthropological Theory of Didactics to frame the findings.
Md Saifuddin Khalid, Technical University of Denmark
Title: Learning Introductory University Mathematics Using Generative AI Tools: An Exploratory Study
Abstract : Despite the hype and potential benefits of Generative AI (GenAI) tools in learning and teaching mathematics, the research databases still lack empirical studies on the exploration, integration, and impact from the perspectives of educators, students, management and designers. This exploratory study applied a participatory approach to identify, try out and evaluate GenAI mathematics learning tools from the website theresanaiforthat.com by facilitating high school and first-year university students from Denmark. Applying convenience sampling, the participants were divided into groups of three and were facilitated during the workshop to document five topics, two exercise questions from each topic, solve those problems using at least three GenAI tools identified from the platform, and evaluate using problem-tree analysis and net promoter score. Findings interviews and group discussions with educators, teaching assistants, and students on the use of GenAI tools for assessment and feedback in Mathematics are synthesized. The preliminary results will be presented.
Thomas Gjesteland, Universitetet i Agder
Title: Portfolio assessment in mathematics for engineering students with the use of AI.
Abstract: In the first semester, all new engineering students at the University of Agder take a calculus class. This class is 7.5 ECTS and has approximately 400 enrolled students. We have recently changed the assessment form from a final exam to a portfolio assessment. The portfolio consists of 4 digital tests and a written assignment. The written assignment consists of open tasks where the students shall apply and use mathematics in an engineering context. It must be delivered as a typed document written in LaTeX. The students are allowed to use artificial intelligence (AI) tools. However, they must write a method section explaining how they solved the problems, which tools they used, and critically discuss the tools and sources they have used. In the fall semester of 2023, 426 students submitted their portfolio. Our experience is that it was challenging for students to apply mathematics in an engineering context. The students also struggled to communicate mathematics in a typed format. We also found that the method section of the assessment revealed which students had used AI critically.