Open Learner Models in the Age of Generative AI

Three invited speakers each deliver a tight 5-minute position statement followed by 5 minutes of immediate reaction from the room. Statements are deliberately in tension: one arguing LLMs finally make OLMs scalable; one arguing LLM-mediated explanations threaten the core principle of inspectability; one asking whether any of this serves the people being modeled. The goal is not to resolve but to surface the tension sharply.

A workshop organizer presents foundational work in Learner Modeling and OLMs — empirical studies, system designs, theoretical frameworks — and approaches at the OLM–LLM intersection, to ground the audience in what is already being attempted and where gaps remain. Participants contribute short presentations of their own work.

Small groups (3–4 participants, mixed across career stage and background) receive a shared scenario: a learner interaction trace from a realistic educational context (programming task, collaborative discussion, or writing exercise). Each group designs how an LLM-enhanced OLM would explain the learner's state to three audiences — the learner, a teacher, and a peer — while explicitly addressing a constraint card of core OLM principles: accuracy, inspectability, learner agency, contestability. Groups then rotate to review each other's designs.

Organizers and participants map findings onto a shared board organized around three questions: What must an OLM preserve even when LLMs are involved? Where does LLM integration genuinely add value? What research does not yet exist? The board becomes a rough but collectively owned research agenda.

Each group nominates one "non-negotiable" principle for LLM-enhanced OLMs. The workshop closes by identifying two or three concrete follow-on actions: a position paper to co-author, a special issue to propose, or a follow-up workshop at a subsequent venue. Participants sign up for future actions before leaving.