# Rationis DefensorBest-Path Theorem Proving: Compiling Derivations

Rationis Defensor: Best-Path Theorem Proving: Compiling Derivations [When computers answer our questions in mathematics and logic they need also to be able to supply justification and explanatory insight. Typical theorem provers do not do this. The paper focuses on tableau theorem provers for First Order Predicate Calculus. The paper introduces a general construction and a technique for converting the tableau data structures of these to human friendly linear proofs using any familiar rule set and ‘laws of thought’. The construction uses a type of tableau in which only leaf nodes are extended. To produce insightful proofs, improvements need to be made to the intermediate output. Dependency analysis and refinement, ie compilation of proofs, can produce benefits. To go further, the paper makes other suggestions including a perhaps surprising one: the notion of best proof or insightful proof is an empirical matter. All possible theorems, or all possible proofs, distribute evenly, in some sense or other, among the possible uses of inference steps. However, with the proofs of interest to humans this uniformity of distribution does not hold. Humans favor certain inferences over others, which are structurally very similar. The author’s research has taken many sample questions and proofs from logic texts, scholastic tests, and similar sources, and analyzed the best proofs for them (‘best’ here usually meaning shortest). This empirical research gives rise to some suggestions on heuristic. The general point is: humans are attuned to certain forms inference, empirical research can tell us what those are, and that empirical research can educate as to how tableau theorem provers, and their symbiotic linear counterparts, should run. In sum, tableau theorem provers, coupled with transformations to linear proofs and empirically sourced heuristic, can provide transparent and accessible theorem proving.] http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

# Rationis DefensorBest-Path Theorem Proving: Compiling Derivations

Part of the Studies in History and Philosophy of Science Book Series (volume 28)
Editors: Maclaurin, James
Rationis Defensor — Mar 14, 2012
20 pages

/lp/springer-journals/rationis-defensor-best-path-theorem-proving-compiling-derivations-T0Q9QuvOoY

# References (26)

Publisher
Springer Netherlands
ISBN
978-94-007-3982-6
Pages
255 –274
DOI
10.1007/978-94-007-3983-3_18
Publisher site
See Chapter on Publisher Site

### Abstract

[When computers answer our questions in mathematics and logic they need also to be able to supply justification and explanatory insight. Typical theorem provers do not do this. The paper focuses on tableau theorem provers for First Order Predicate Calculus. The paper introduces a general construction and a technique for converting the tableau data structures of these to human friendly linear proofs using any familiar rule set and ‘laws of thought’. The construction uses a type of tableau in which only leaf nodes are extended. To produce insightful proofs, improvements need to be made to the intermediate output. Dependency analysis and refinement, ie compilation of proofs, can produce benefits. To go further, the paper makes other suggestions including a perhaps surprising one: the notion of best proof or insightful proof is an empirical matter. All possible theorems, or all possible proofs, distribute evenly, in some sense or other, among the possible uses of inference steps. However, with the proofs of interest to humans this uniformity of distribution does not hold. Humans favor certain inferences over others, which are structurally very similar. The author’s research has taken many sample questions and proofs from logic texts, scholastic tests, and similar sources, and analyzed the best proofs for them (‘best’ here usually meaning shortest). This empirical research gives rise to some suggestions on heuristic. The general point is: humans are attuned to certain forms inference, empirical research can tell us what those are, and that empirical research can educate as to how tableau theorem provers, and their symbiotic linear counterparts, should run. In sum, tableau theorem provers, coupled with transformations to linear proofs and empirically sourced heuristic, can provide transparent and accessible theorem proving.]

Published: Mar 14, 2012

Keywords: Leaf Node; Theorem Prover; Natural Deduction; Standard Tableau; Skolem Function