Evaluation of a Semi-Automated Theorem Prover (Part I)

So little evaluation has been carried out on user interfaces to theorem provers that even the question of what should be evaluated appears open. We took a user/task related approach, where subjects in a pilot study were given three theorems to prove, which might or might not require intervention by the user, and logged some of their interactions. In this paper, we describe the facets of semi-automated theorem prover illustrated by these theorems, and how we might expect our system to aid users in making appropriate interventions in proof attempts.