AI Agents as Cognitive Partners in Micro-Project-Based Learning: A Professional Development Workshop Study

Abstract

This case study examines how an LLM-based AI agent
supported adult learners in professional development during
a micro-project-based learning (micro-PBL) workshop. In
this activity, learners developed their individually owned
“ideas worth spreading” into mini TED-style talks through
iterative cycles of reflection, critique, and public
presentation. Interviews with 10 participants revealed four
generative themes underlying learners’ interactions with
the AI agent: (1) momentum restoration via cognitive
extension under time pressure to overcome cognitive
impasse; (2) role negotiation, where learners shifted AI
roles from tools to partners; (3) appropriate reliance,
enacted through trust calibration routines; and (4) affect
regulation, providing a psychologically safe space for
experimentation. Building on these themes, we propose four
design principles for AI-supported micro-PBL: ensure
learner ownership, design for momentum, orchestrate
multi-source feedback, and embed calibration routines. The
study offers an explanatory account of how AI becomes a
“cognitive partner” in compressed, ill-structured learning
tasks.