Interactive Proof Discovery: An Empirical Study of HOL Users

One commonly cited obstacle to the widespread use of theorem provers is the poor quality of the user interface. Research projects which have addressed this problem have had mixed results. In the majority of cases, the designers of new interfaces to theorem provers have failed to draw upon the most relevant discipline: Human Computer Interaction. The design of these interfaces is carefully considered, but is generally informed by the personal experience of the designer and their intuitions. The ITP project aims to apply the methods of HCI to the problem of designing interfaces to interactive theorem provers. Of particular relevance are task modelling and analysis (Norman, 1988) which can identify the gulfs of execution and evaluation which hinder interaction. Nielsen (1986) presents a linguistic account of interaction by asserting that the different levels at which user-system interaction may be described correspond to the lexical, syntactic and semantic levels of linguistic activity. Such multi-level views of interaction are relevant to our study of user-prover interaction. Theories about the usability and effectiveness of representations (Green, 1989, 1991) and studies of the nature of programming (Wiedenbeck, 1985)(Davis, 1991) are also important.