MinIAtures: Turning AI Guidelines into Daily Habits Through Student-Led Peer Coaching

Beyond the bot

Nicolas BOUZENDORFF, Aleyna DUMAN, Yasmine EL HAOU, Sélim KADIOGLU, Laura LO DICO, Université de liège (HEC Liège)

Training / education / pedagogy

We are implementing ‘MinIAtures’, a multimedia microlearning series where AI acts as a “Study Coach” for first-year students. Inspired by the university charter and developed primarily by students with the guidance/support of faculty members, we train other students to treat AI as a reasoning partner rather than a shortcut for answers. This student-led initiative, integrated into our university’s learning platform, transforms abstract (AI) guidelines into a practical tool that supports active learning and deepens cognitive engagement.

The blind use of AI (“cognitive offloading”) is a direct threat to academic integrity and cognitive depth. Our project addresses the AI dilemma : does the tool train the student, or does the student train the tool? The project, while aligning with the ULiège Charter, moves from high-level “dry” official rules to a two-pronged human centered pedagogical support for students: AI-vigilance videos designed for social media and the set-up of an institutional IA peer-coaching program based on these videos. Crucially, this initiative has a strong potential of joining existing university programs (Programme parrainage étudiant, student corner, mentoring etc). By integrating into these frameworks, we transform traditional mentoring systems into a hybrid model where human mentors use AI driven socratic coaching to help their peers make higher achievements. This ensures the learning process remains indisputably human while scaling high-quality, responsible support to all students during their transition to university.

Beyond the Bot brings together students from engineering, computer science, and international programs, alongside faculty mentors because we believe tackling AI responsibly takes more than one perspective. At the core of MinIAtures is a simple and ambitious bet: students coaching students, supported by faculty, is the most effective way to turn abstract AI guidelines into real daily habits. We intend to track this through concrete KPIs across our microlearning capsules, covering everything from prompting and synthesizing to revising and critical reading. This Challenge is our chance to test the applicability of our work at HEC Liège against the world. We want our model challenged, refined, and scalable. Ultimately, we want to prove that student-led initiatives can ensure learners remain the masters of their own reasoning process.

Our initial contribution goes beyond the development of pedagogical content. We propose a structured framework to govern the integration of generative AI into students’ daily learning practices. First, we introduce a “constrained AI tutor” model, where AI is intentionally designed to guide reasoning rather than provide direct answers. This approach creates what we define as “pedagogical friction”, ensuring that students remain cognitively engaged and actively involved in their learning process. Second, we contribute a measurable evaluation framework to assess the real impact of AI on student learning. Through A/B testing and targeted indicators (such as autonomy, depth of reasoning, and reliance on AI), we aim to move from assumptions about AI use to evidence-based insights. Third, we provide a concrete integration roadmap that embeds our approach into existing university ecosystems. Rather than creating a new tool, we leverage structures such as peer mentoring, LMS platforms, and student support systems to ensure scalability and long-term adoption. In addition, some modules are already available, such as a dynamic video presenting the ULiège charter on the ethical use of AI (link: https://drive.google.com/file/d/1ASKpYc81PKMRIij7fyQ4HjyUv8MsWMzD/view?usp=drivesdk) . Others (summarization, languages, organization) are currently being developed and will be progressively added, while the remaining modules will be completed later. The project is structured into pedagogical modules presented as practical learning videos: • Module 1: learning how to interact effectively with AI (advanced prompting) • Module 2: planning and organizing study work • Module 3: structuring courses and adapting explanations • Module 4: synthesizing and summarizing information • Module 5: understanding mistakes and improving skills • Module 6: practicing with mock exams • Module 7: learning languages with AI Each module is short, practical, and based on active interaction with AI. Finally, our contribution reframes the role of students: from passive users of AI-generated content to “vigilant editors” capable of critically engaging with and refining machine outputs. Required expertise & resources : Learning Science Experts: To validate our “Socratic prompting” approach and ensure its impact on long-term memory and reasoning. Instructional Designers: To structure our microlearning capsules into scalable learning paths compatible with university LMS systems. Advanced LLM Access: To experiment with “constrained tutor” architectures that prioritize reasoning guidance over answer delivery. Pilot Partnerships: Collaboration with universities to conduct A/B testing between traditional peer mentoring and AI-enhanced coaching models, to measure impact on student autonomy and cognitive engagement.