Generative AI
Generative artificial intelligence has emerged as a transformative tool in education, offering scalable individualized learning. However, there is a lack of theoretically informed and methodologically rigorous research on how GenAI can effectively augment learning. The studies in this research area seeks to illuminate the advantages of adopting a human-centered perspective, particularly when the goal is to foster deeper, self-reflective learning among students.
This research involves examining how a human centered generative AI functions within collaborative learning contexts. A key question is whether embedding GenAI in group-based settings amplifies the benefits of generative processing or whether it introduces coordination costs that may limit its effectiveness. To explore this, the study compares the impact of individual versus collaborative use of a pedagogically informed GenAI tutor. Through a pre-registered randomized controlled experiment conducted in high schools, we examine how interacting with the GenAI tutor, either individually or in small groups of 3-4 students, affects learners’ conceptual understanding and self-efficacy. By situating these interventions within real classroom environments, the research aims to generate insights into how human-AI collaboration can be structured to enhance, rather than hinder, essential cognitive and motivational processes in learning.
In the age of GenAI, we aim to inspire efforts to design learner-centred systems and ways of implementation that promote independent thinking and cognitive growth. In this project, we explored how theory-driven approaches can guide GenAI design to support both the metacognitive and socio-emotional dimensions of learning. Specifically, we developed and tested an emotionally intelligent GenAI chatbot grounded in theories of self-regulated learning (SRL) and affect. The chatbot prompted student engagement in SRL behaviours such as goal setting, strategy use, and self-reflection, while also tailoring emotional support to learners’ affective states. In a randomized controlled trial, we found that incorporating emotional scaffolding alongside SRL support transformed GenAI from a generic assistant into a tool that meaningfully supports self-regulation and fosters positive emotions in ways that ChatGPT alone cannot.
This research is a collaboration between the Virtual Learning Lab and the Centre for Learning and Living with AI (CELLA), an international team committed to preparing young learners for an AI-driven society. The Virtual Learning Lab contributes to CELLA's ongoing research project, which focuses on supporting secondary school students’ self-regulated learning (SRL) skills through analytics-based and generative AI–enabled scaffolding, conducted under the auspices of CELLA. The significance of this research project lies in investigating how GenAI can enhance student performance while simultaneously fostering the cognitive and metacognitive engagement necessary for durable learning outcomes. In pursuit of these objectives, the study will include experimental groups to systematically examine how generative AI interventions (varying in form, content, and intensity) influence both immediate performance and longer-term learning outcomes. By exploring different types and levels of AI-based support, we aim to determine how to amplify the benefits of GenAI while mitigating any potential negative impacts on students’ cognitive and metacognitive engagement and learner agency.
This research aimed (1) to extend theory by proposing how GenAI can support generative sense-making and (2) to examine its effects on conceptual knowledge, self-efficacy, and trust immediately after use, as well as on conceptual knowledge, enjoyment, and behavioral intentions in a delayed follow-up. To address this, two between-subjects experiments were conducted. Study 1 was a pre-registered experiment with 175 university students in an authentic cognitive psychology course, comparing ChatTutor with a generic GenAI system (ChatGPT) and a teaching-as-usual control. Study 2 replicated the design with 234 high school students, comparing ChatTutor with ChatGPT and a re-study control.
- Makransky, G., Shiwalia, B.M., Herlau, T. et al. (2025). Beyond the “Wow” Factor: Using Generative AI for Increasing Generative Sense-Making. Educ Psychol Rev 37, 60.
Researchers
Name | Title | |
---|---|---|
Diordiev, Elena-Alexandra | Student FU |
|
Makransky, Guido | Professor |
|
Shiwalia, Ban Mouid | Student FU |
|
Simon, Sebastian Andreas | Postdoc |
|
Stenberdt, Valdemar Aksel | PhD Student |
|
Svendsen, Magnus Holm | IT Officer |
|
Contact
Guido Makransky
Professor, PI
Study period: 2025 -