How an educational chatbot supports thinking
A correct answer is a starting point. Useful hints and student reasoning need deliberate design and evaluation too.
Understand the learner's current thinking
A tutoring conversation needs to understand the learner's attempt. When asked for an answer, it can offer a focused hint or clarifying question and ask for a next step rather than replace the exercise with a complete solution.
Vispo includes Socratic response guidance of this kind. Policies and prompts still need evaluation against actual conversations; they do not guarantee perfect answer withholding or learning gains.
Evaluating pedagogy with lessons from LearnLM
LearnLM studies pedagogical instruction following and evaluates responses in learning contexts. It suggests evaluating not only correctness but whether a hint addresses the difficulty and leaves room for the learner's thinking.
Human–AI roles in Tutor CoPilot
Tutor CoPilot studies real-time AI suggestions to human tutors. Its focus is how educators use that support, rather than a fully autonomous chatbot. It informs a design that leaves teachers room to review and adapt advice.
These are related research references, not claims that Vispo uses their fine-tuned models, training data or has reproduced their results.