Abstract
This pilot study examines the feasibility of AI-assisted Dynamic Assessment (AI-DA) for Japanese reading comprehension. Multiple-choice (MC) tests are efficient but often reflect test-taking strategies, whereas constructed-response (CR) formats better capture learners’ mental representations at higher scoring cost. Dynamic Assessment (DA) adds graduated mediation to gauge ability and learning potential, but is hard to scale. An AI-DA system that automatically scores CR answers and provides stepwise mediation was piloted with an intermediate Japanese learner across four online sessions. Results suggest MC and CR tap different reading constructs, and AI-DA can balance efficiency with diagnostic depth.
