MODERA (Moderation and Optimization of Discussions with Enhanced Reasoning AI)

Dialogue is the foundation of meaningful learning. When students question, debate, and build on each other’s ideas, they not only acquire knowledge but also cultivate analytical depth, intellectual autonomy, and collaborative reasoning (Stahl et al., 2014; Jeong & Hmelo-Silver, 2016). Yet these conditions are increasingly absent in higher education. Large enrolments, limited instructor capacity, and low participation in online forums leave many students engaging passively rather than critically. The challenge is clear: how can universities restore dialogue to the centre of learning at scale?Recent advances in large language models (LLMs) present a timely opportunity. Unlike static digital tools, LLMs are generative: they can pose questions, highlight connections, and prompt reflection (Dwivedi et al., 2023). While concerns about bias and misinformation remain valid, new multi-agent verification approaches—where models cross-check and critique one another’s outputs—have markedly improved accuracy and transparency (Pavlick, 2023; Zhang et al., 2023; Guo et al., 2025; Weng et al., 2024). These developments make it possible to reconceive AI not as an answer-giving authority that fosters passivity, but as a moderator that sustains rigorous, reflective, and genuinely collaborative dialogue.This project introduces MODERA, an innovative system designed to complement instructor input in undergraduate courses. Implemented on Discord, a platform already familiar to students, MODERA will prompt clarification, reasoning, and connection in weekly discussions while deliberately reducing its presence once dialogue is productive. A verification pipeline will ensure accuracy and ethical standards.The project will address two central questions: (1) whether AI facilitation improves knowledge co-construction and learning outcomes compared to instructor-only moderation, and (2) how students experience AI-facilitated dialogue in terms of trust, engagement, and perceived learning. Using a within-subject design in LT-4261 Applied Psycholinguistics, fifty students will alternate between AI- and instructor-facilitated discussions. Outcomes will be evaluated across engagement, argument quality, learning gains, and student perceptions, using both behavioural measures and established rubrics.Deliverables will include open-access pedagogical guidelines, reusable tools for adoption across institutions, peer-reviewed publications, faculty development workshops, and partnerships with industry to ensure scalability and sustainability.By re-establishing dialogue, collaboration, and active learning at the heart of higher education, this project positions AI as a catalyst for what universities aspire to but struggle to deliver: classrooms where discussion is not incidental but essential, where learning is not passive but active, and where students are empowered to think together toward deeper understanding.