Generative AI promises to support open-ended learning by generating content, personalizing experiences, and modeling learners. But educators, the paper argues, need more than raw capability—they need ways to configure these tools to fit their pedagogical aims. The study, posted on arXiv, investigates how teachers would shape AI behavior in exploratory learning contexts.
Through interviews and design activities, the researchers focused on three dimensions: inference, persistence, and disclosure. Inference concerns how AI interprets learner needs; persistence addresses how the AI maintains context and continuity; disclosure covers how transparent the AI is about its own role and limitations. The paper suggests that educators want deliberate control over these dimensions rather than accepting default settings.
The significance lies in shifting the conversation from what GenAI can do to how educators can direct it. By centering educator perspectives, the study offers a framework for designing AI systems that respect instructional goals while still enabling open-ended exploration. As the abstract notes, the aim is to shape AI capabilities around educational objectives, not the other way around.