Incremental Proof Planning by Meta-Rules

We propose a new approach to automated tactical theorem proving and proof planning: By using metarules to control the search for a proof, heuristic knowledge is declaratively collected. This eases the user's understanding of the system's search for a proof thus making user-interactions easier. The system's heuristics can be modiied by simply using diierent sets of metarules. Via metarules contextual preconditions for tactics can be formulated in a transparent way. The metarule interpreter interleaves proof planning , plan execution (tactic application), and reasoning about the tactic's results. In our framework the planner has access to more information, because the metarule interpreter knows about the history of the proof resp. the proof plan by being able to access the (partial) proof-tree.