## A short survey of automated reasoning

Citations: | 2 - 0 self |

### BibTeX

@MISC{Harrison_ashort,

author = {John Harrison},

title = {A short survey of automated reasoning},

year = {}

}

### OpenURL

### Abstract

Abstract. This paper surveys the field of automated reasoning, giving some historical background and outlining a few of the main current research themes. We particularly emphasize the points of contact and the contrasts with computer algebra. We finish with a discussion of the main applications so far. 1 Historical introduction The idea of reducing reasoning to mechanical calculation is an old dream [75]. Hobbes [55] made explicit the analogy in the slogan ‘Reason [...] is nothing but Reckoning’. This parallel was developed by Leibniz, who envisaged a ‘characteristica universalis’ (universal language) and a ‘calculus ratiocinator ’ (calculus of reasoning). His idea was that disputes of all kinds, not merely mathematical ones, could be settled if the parties translated their dispute into the characteristica and then simply calculated. Leibniz even made some steps towards realizing this lofty goal, but his work was largely forgotten. The characteristica universalis The dream of a truly universal language in Leibniz’s sense remains unrealized and probably unrealizable. But over the last few centuries a language that is at least adequate for

### Citations

3101 | Graph-based algorithms for Boolean function manipulation
- Bryant
- 1986
(Show Context)
Citation Context ...is usually a more amenable target for highly automated techniques. Such techniques include SAT (propositional satisfiability testing), using new algorithms or high-quality implementations of old ones =-=[14, 100, 92, 79, 41]-=-, sophisticated forms of symbolic simulation [15, 91], and temporal logic model checking [24, 86, 25]. Also, hardware verification is particularly attractive because fixing errors is often invasive an... |

2600 | Model Checking
- Clarke, Grumberg, et al.
- 1999
(Show Context)
Citation Context ...satisfiability testing), using new algorithms or high-quality implementations of old ones [14, 100, 92, 79, 41], sophisticated forms of symbolic simulation [15, 91], and temporal logic model checking =-=[24, 86, 25]-=-. Also, hardware verification is particularly attractive because fixing errors is often invasive and potentially expensive. For example, in response to an error in the FDIV (floating-point division) i... |

1107 |
A computing procedure for quantification theory
- Davis, Putnam
- 1960
(Show Context)
Citation Context ...that could prove facts in Euclidean geometry using human-style diagrams to direct or restrict the proofs. A quite different approach was taken by other pioneers such as Gilmore [39], Davis and Putnam =-=[31]-=-, and Prawitz [84]. They attempted to implement proof search algorithms inspired by results from logic (e.g. the completeness of Gentzen’s cut-free sequent calculus), often quite remote from the way h... |

982 |
Term Rewriting and All That
- Baader, Nipkow
- 1998
(Show Context)
Citation Context ...ialized algorithms. In particular, there has always been strong interest in effective algorithms for purely equational reasoning. Knuth-Bendix completion [63] led to a great deal of fruitful research =-=[3, 4, 56]-=-. Automated proof of geometry problems using purely algebraic methods has also attracted much interest. The first striking success was by Wu [108] using his special triangulation algorithm, and others... |

837 |
Design and synthesis of synchronization skeletons using branching-time temporal logic
- Clarke, Emerson
- 1982
(Show Context)
Citation Context ...satisfiability testing), using new algorithms or high-quality implementations of old ones [14, 100, 92, 79, 41], sophisticated forms of symbolic simulation [15, 91], and temporal logic model checking =-=[24, 86, 25]-=-. Also, hardware verification is particularly attractive because fixing errors is often invasive and potentially expensive. For example, in response to an error in the FDIV (floating-point division) i... |

525 |
Introduction to HOL : a theorem proving environment for higher-order logic
- Gordon, Melham
- 1993
(Show Context)
Citation Context ...s. By contrast, theorem provers usually start from a strict and precisely defined logical foundation and build up other mathematical concepts by a sequence of definitions. For example, the HOL system =-=[42]-=- starts with a few very basic axioms for higher-order logic and a couple of set-theoretic axioms, and these are given a rather precise semantics in the documentation. From that foundation, other conce... |

415 |
Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme, I. Monatshefte für Mathematik und Physik 38
- Gödel
- 1931
(Show Context)
Citation Context ...lusions reached using some controversial abstract concepts could nevertheless be show still to be valid or even provable without them. The calculus ratiocinator Gödel’s famous incompleteness theorems =-=[40, 98, 37]-=- show that formal systems for deducing mathematics have essential weaknesses. For example, his first theorem is that any given formal system of deduction satisfying a few natural conditions is incompl... |

289 |
Every planar map is four colorable
- Appel, Haken, et al.
- 1977
(Show Context)
Citation Context ...o be faulty. A notable example is the purported proof of the 4-colour theorem by Kempe [62], the flaw only being noticed a decade later [51], and the theorem only being conclusively proved much later =-=[2]-=-. The errors need not be deep mathematical ones, as shown by the following [69]: Professor Offord and I recently committed ourselves to an odd mistake (Annals of Mathematics (2) 49, 923, 1.5). In form... |

288 |
An unsolvable problem of elementary number theory
- Church
- 1936
(Show Context)
Citation Context ...rst of a variety of ‘impossibility’ results that only really become possible when the notion of mathematical proof is formalized. Inspired by techniques used in Gödel’s incompleteness results, Church =-=[23]-=- and Turing [106] proposed definitions of ‘mechanical computability’ and showed that one famous logical decision question, Hilbert’s Entscheidungsproblem (decision problem for first-order logic) was u... |

248 | BerkMin: a fast and robust SAT-solver
- Goldberg, Novikov
- 2002
(Show Context)
Citation Context ...is usually a more amenable target for highly automated techniques. Such techniques include SAT (propositional satisfiability testing), using new algorithms or high-quality implementations of old ones =-=[14, 100, 92, 79, 41]-=-, sophisticated forms of symbolic simulation [15, 91], and temporal logic model checking [24, 86, 25]. Also, hardware verification is particularly attractive because fixing errors is often invasive an... |

217 |
Ein Algorithmus zum Auffinden der Basiselemente des Restklassringes nach einem nulldimensionalen Polynomideal
- Buchberger
- 1965
(Show Context)
Citation Context ...ecial triangulation algorithm, and others have further refined and applied this approach [22] as well as trying other methods such as resultants and Gröbner bases [61, 89]. Incidentally Gröbner bases =-=[16, 17]-=- are more usually considered a part of computer algebra, but as a tool for testing ideal membership they give a powerful algorithm for solving various logical decision problems [95, 60].s4 Application... |

154 | Thorough static analysis of device drivers
- Ball, Bounimova, et al.
- 2006
(Show Context)
Citation Context ...ghly automated tools. There has also been renewed interest in applications to software, particularly partial verification or sophisticated static checking rather than complete functional verification =-=[5]-=-. And for certain applications, especially implementations of mathematically sophisticated algorithms, more general and interactive theoremsproving is needed. A particularly popular and successful tar... |

125 | Completion without failure
- Bachmair, Dershowitz, et al.
- 1989
(Show Context)
Citation Context ...ialized algorithms. In particular, there has always been strong interest in effective algorithms for purely equational reasoning. Knuth-Bendix completion [63] led to a great deal of fruitful research =-=[3, 4, 56]-=-. Automated proof of geometry problems using purely algebraic methods has also attracted much interest. The first striking success was by Wu [108] using his special triangulation algorithm, and others... |

89 |
C.P.: Edinburgh LCF: A Mechanised Logic
- Gordon, Milner, et al.
- 1979
(Show Context)
Citation Context ...resultsnow called “SAM’s Lemma” and the mathematician recognized that it easily yielded a proof of the open conjecture. Not long after the SAM project, the AUTOMATH [32, 33], Mizar [104, 105] and LCF =-=[43]-=- proof checkers appeared, and each of them in its way has been profoundly influential. Many of the most successful interactive theorem provers around today are directly descended from one of these. No... |

78 |
Ein algorithmisches Kriterium für die Lösbarkeit eines algebraischen Gleichungssystems
- Buchberger
- 1970
(Show Context)
Citation Context ...ecial triangulation algorithm, and others have further refined and applied this approach [22] as well as trying other methods such as resultants and Gröbner bases [61, 89]. Incidentally Gröbner bases =-=[16, 17]-=- are more usually considered a part of computer algebra, but as a tool for testing ideal membership they give a powerful algorithm for solving various logical decision problems [95, 60].s4 Application... |

77 | A science of reasoning
- Bundy
- 1989
(Show Context)
Citation Context ...zation of arithmetic theorems and prove the generalizations by induction. The success of NQTHM, and the contrasting difficulty of fitting its methods into a simple conceptual framework, has led Bundy =-=[20]-=- to reconstruct its methods in a general science of reasoning based on proof planning. Depending on one’s point of view, one can regard the considerable interest in proof planning as representing a su... |

71 | TPS: A theorem proving system for classical type theory
- Andrews, Bishop, et al.
- 1996
(Show Context)
Citation Context ...as well as resolution [88] and its numerous refinements [64, 71, 72, 34, 87, 97]. Despite the general emphasis on pure first-order logic, there has also been research in automating higher-order logic =-=[1]-=-, which allows quantification over sets and functions as part of the logic rather than via additional axioms. However, there have been some successes for more specialized algorithms. In particular, th... |

66 | The Undecidable - Davis, ed - 1965 |

65 | editors. Quantifier Elimination and Cylindrical Algebraic Decomposition - Caviness, Johnson - 1998 |

58 |
de Bruijn. A survey of the project Automath
- G
- 1980
(Show Context)
Citation Context ...he computer automatically proved a resultsnow called “SAM’s Lemma” and the mathematician recognized that it easily yielded a proof of the open conjecture. Not long after the SAM project, the AUTOMATH =-=[32, 33]-=-, Mizar [104, 105] and LCF [43] proof checkers appeared, and each of them in its way has been profoundly influential. Many of the most successful interactive theorem provers around today are directly ... |

57 |
A Proof of Correctness of the VIPER Microprocessor: The First Level
- Cohn
- 1988
(Show Context)
Citation Context ...3], Intel set aside approximately $475M to cover costs. Since the 1980s there has been extensive research in formal verification of microprocessor designs using traditional theorem proving techniques =-=[57, 26, 44, 59, 99]-=-. Generally there has been more emphasis on the highly automated techniques like model checking that lie somewhat apart from the automated reasoning mainstream. However, recently there has been someth... |

56 | Graph Theory 1736-1936 - Biggs, Lloyd, et al. - 1976 |

47 |
A complete proof of correctness of the knuth-bendix completion algorithm. Rapports de Recherche
- Huet
- 1980
(Show Context)
Citation Context ...ialized algorithms. In particular, there has always been strong interest in effective algorithms for purely equational reasoning. Knuth-Bendix completion [63] led to a great deal of fruitful research =-=[3, 4, 56]-=-. Automated proof of geometry problems using purely algebraic methods has also attracted much interest. The first striking success was by Wu [108] using his special triangulation algorithm, and others... |

43 | Integrating Gandalf and HOL
- Hurd
- 1999
(Show Context)
Citation Context ...s [80, 93], has attracted considerable interest. Meanwhile, interactive provers develop better user interfaces and proof languages [48], incorporate ideas from automated provers and even link to them =-=[58]-=-, and develop ever more extensive libraries of formalized mathematics. For a nice survey of some of the major interactive systems, showing a proof of the irrationality of √ 2 in each as an example, se... |

37 | A methodology for hardware verification based on logic simulation
- Bryant
- 1991
(Show Context)
Citation Context ... Such techniques include SAT (propositional satisfiability testing), using new algorithms or high-quality implementations of old ones [14, 100, 92, 79, 41], sophisticated forms of symbolic simulation =-=[15, 91]-=-, and temporal logic model checking [24, 86, 25]. Also, hardware verification is particularly attractive because fixing errors is often invasive and potentially expensive. For example, in response to ... |

33 |
Form and content in quantification theory
- Hintikka
- 1955
(Show Context)
Citation Context ... special algorithms were subordinated to resolution or similar principles rather than being developed in themselves. There are numerous different algorithms for general proof search, such as tableaux =-=[7, 54]-=-, model elimination [70] as well as resolution [88] and its numerous refinements [64, 71, 72, 34, 87, 97]. Despite the general emphasis on pure first-order logic, there has also been research in autom... |

29 |
A proof method for quantification theory: It’s justification and realization
- Gilmore
- 1960
(Show Context)
Citation Context ...[38] designed a prover that could prove facts in Euclidean geometry using human-style diagrams to direct or restrict the proofs. A quite different approach was taken by other pioneers such as Gilmore =-=[39]-=-, Davis and Putnam [31], and Prawitz [84]. They attempted to implement proof search algorithms inspired by results from logic (e.g. the completeness of Gentzen’s cut-free sequent calculus), often quit... |

28 |
L.: A skeptic’s approach to combining HOL and Maple
- Harrison, Théry
- 1998
(Show Context)
Citation Context ...ization of mathematics. Since the strengths and weaknesses of theorem provers and CASs are almost perfectly complementary, a natural idea is to somehow get the best of both worlds. One promising idea =-=[50]-=- is to use the CAS as an ‘oracle’ to compute results that can then be rigorously checked in the theorem prover. This only works for problems where checking a result is considerably easier than derivin... |

27 |
Semantic Entailment and Formal Derivability”, Mededelingen van de Koninklijke Nederlandse Akademie van Wetenschappen, Afdeling Letterkunde 18
- Beth
- 1955
(Show Context)
Citation Context ... special algorithms were subordinated to resolution or similar principles rather than being developed in themselves. There are numerous different algorithms for general proof search, such as tableaux =-=[7, 54]-=-, model elimination [70] as well as resolution [88] and its numerous refinements [64, 71, 72, 34, 87, 97]. Despite the general emphasis on pure first-order logic, there has also been research in autom... |

27 |
Formal verification and implementation of a microprocessor. VLSI specification, verification and
- Joyce
- 1988
(Show Context)
Citation Context ...3], Intel set aside approximately $475M to cover costs. Since the 1980s there has been extensive research in formal verification of microprocessor designs using traditional theorem proving techniques =-=[57, 26, 44, 59, 99]-=-. Generally there has been more emphasis on the highly automated techniques like model checking that lie somewhat apart from the automated reasoning mainstream. However, recently there has been someth... |

24 |
The SECD Microprocessor, A Verification Case Study
- Graham
- 1992
(Show Context)
Citation Context ...3], Intel set aside approximately $475M to cover costs. Since the 1980s there has been extensive research in formal verification of microprocessor designs using traditional theorem proving techniques =-=[57, 26, 44, 59, 99]-=-. Generally there has been more emphasis on the highly automated techniques like model checking that lie somewhat apart from the automated reasoning mainstream. However, recently there has been someth... |

23 |
Die logischen Grundlagen der Mathematik.Mathematische Annalen 88: 151–165. English translation
- Hilbert
- 1923
(Show Context)
Citation Context ...overy of non-Euclidean geometries. Later still, when the theory of infinite sets began to be pursued for its own sake and generalized, mainly by Cantor, renewed foundational worries appeared. Hilbert =-=[53]-=- suggested an ingenious programme to give mathematics a reliable foundation. In the past, new and apparently problematic ideas such as complex numbers and non-Euclidean geometry had been given a found... |

22 |
The unwinding number
- Corless, Jeffrey
- 1996
(Show Context)
Citation Context ...is kind of thing freely. Although systematic approaches to keeping track of branch cuts are possible, most mainstreamssystems don’t use them. For example, using the concept of ‘unwinding number’ u(z) =-=[28]-=-, we can express rigorously simplification rules such as: w �= 0 ∧ z �= 0 ⇒ log(wz) = log(w) + log(z) − 2πiu(log(w) + log(z)) Most users probably find such pedantic details as branch cut identificatio... |

17 |
de Bruijn, The mathematical language AUTOMATH, its usage and some of its extensions, Symposium on automatic demonstration
- G
- 1970
(Show Context)
Citation Context ...he computer automatically proved a resultsnow called “SAM’s Lemma” and the mathematician recognized that it easily yielded a proof of the open conjecture. Not long after the SAM project, the AUTOMATH =-=[32, 33]-=-, Mizar [104, 105] and LCF [43] proof checkers appeared, and each of them in its way has been profoundly influential. Many of the most successful interactive theorem provers around today are directly ... |

16 |
The Calculus of Logic
- Boole
- 1848
(Show Context)
Citation Context ...iversal language in Leibniz’s sense remains unrealized and probably unrealizable. But over the last few centuries a language that is at least adequate for (most) mathematics has been developed. Boole =-=[11]-=- developed the first really successful symbolism for logical and settheoretic reasoning. What’s more, he was one of the first to emphasize the possibility of applying formal calculi to several differe... |

16 |
Semi-automated mathematics
- Guard, Oglesby, et al.
- 1969
(Show Context)
Citation Context ...e of theorem provers in the SAM (Semi-Automated Mathematics) family made their provers interactive, so that the mathematician could work on formalizing a proof with machine assistance. As they put it =-=[46]-=-: Semi-automated mathematics is an approach to theorem-proving which seeks to combine automatic logic routines with ordinary proof procedures in such a manner that the resulting procedure is both effi... |

14 | Proof style
- Harrison
- 1997
(Show Context)
Citation Context ...es), which uses techniques for combining deduction in certain theories [80, 93], has attracted considerable interest. Meanwhile, interactive provers develop better user interfaces and proof languages =-=[48]-=-, incorporate ideas from automated provers and even link to them [58], and develop ever more extensive libraries of formalized mathematics. For a nice survey of some of the major interactive systems, ... |

13 |
A Computer Program for Presburger’s Algorithm. In: Summary of talks presented at the
- Davis
- 1957
(Show Context)
Citation Context ...ure that can in principle verify any logical consequence of those axioms. Moreover many suitably restricted logical problems are decidable. For example, perhaps the very first computer theorem prover =-=[29]-=- could prove formulas involving quantifiers over natural numbers, but with a linearity restriction ensuring decidability [85]. 2 Theorem provers and computer algebra systems Before we proceed to surve... |

12 | Philosophy of mathematics: Selected readings - Benacerraf, Putman - 1964 |

10 |
Some Automatic Proofs in Analysis
- Bledsoe
- 1984
(Show Context)
Citation Context ...ted side, but there have been notable results based on human-oriented approaches. For example Bledsoe attempted to formalize methods often used by humans for proving theorems about limits in analysis =-=[10]-=-. Bledsoe’s student Boyer together with Moore developed the remarkable NQTHM prover [13] which can often perform automatic generalization of arithmetic theorems and prove the generalizations by induct... |

9 |
An introduction to Wu’s method for mechanical theorem proving in geometry
- Chou
- 1988
(Show Context)
Citation Context ...ely algebraic methods has also attracted much interest. The first striking success was by Wu [108] using his special triangulation algorithm, and others have further refined and applied this approach =-=[22]-=- as well as trying other methods such as resultants and Gröbner bases [61, 89]. Incidentally Gröbner bases [16, 17] are more usually considered a part of computer algebra, but as a tool for testing id... |

8 |
Theory of Sets. Elements of Mathematics
- Bourbaki
- 1968
(Show Context)
Citation Context ...90, 82, 49]. The formalization of mathematics The formalizability in principle of mathematical proof is widely accepted among professional mathematicians as the final arbiter of correctness. Bourbaki =-=[12]-=- clearly says that ‘the correctness of a mathematical text is verified by comparing it, more or less explicitly, with the rules of a formalized language’, while Mac Lane [73] is also quite explicit (p... |

8 | Floating-point verification using theorem proving
- Harrison
- 2006
(Show Context)
Citation Context ...tions of mathematically sophisticated algorithms, more general and interactive theoremsproving is needed. A particularly popular and successful target is the verification of floating-point algorithms =-=[78, 90, 82, 49]-=-. The formalization of mathematics The formalizability in principle of mathematical proof is widely accepted among professional mathematicians as the final arbiter of correctness. Bourbaki [12] clearl... |

8 |
Map-colour theorem. The Quarterly
- Heawood
(Show Context)
Citation Context ...ty of well-documented cases where published results turned out to be faulty. A notable example is the purported proof of the 4-colour theorem by Kempe [62], the flaw only being noticed a decade later =-=[51]-=-, and the theorem only being conclusively proved much later [2]. The errors need not be deep mathematical ones, as shown by the following [69]: Professor Offord and I recently committed ourselves to a... |

7 | Automated geometric reasoning: Dixon resultants, Gröbner bases, and characteristic sets
- Kapur
(Show Context)
Citation Context ... success was by Wu [108] using his special triangulation algorithm, and others have further refined and applied this approach [22] as well as trying other methods such as resultants and Gröbner bases =-=[61, 89]-=-. Incidentally Gröbner bases [16, 17] are more usually considered a part of computer algebra, but as a tool for testing ideal membership they give a powerful algorithm for solving various logical deci... |

6 |
Nivelle. Ordering refinements of resolution
- de
- 1994
(Show Context)
Citation Context ...ing developed in themselves. There are numerous different algorithms for general proof search, such as tableaux [7, 54], model elimination [70] as well as resolution [88] and its numerous refinements =-=[64, 71, 72, 34, 87, 97]-=-. Despite the general emphasis on pure first-order logic, there has also been research in automating higher-order logic [1], which allows quantification over sets and functions as part of the logic ra... |

5 |
An ideal-theoretic approach to word problems and unification problems over finitely presented commutative algebras
- Kandri-Rody, Kapur, et al.
- 1985
(Show Context)
Citation Context ... Gröbner bases [16, 17] are more usually considered a part of computer algebra, but as a tool for testing ideal membership they give a powerful algorithm for solving various logical decision problems =-=[95, 60]-=-.s4 Applications of automated reasoning At present there are two main applications of automated reasoning. Formal verification One promising application of formalization, and a particularly easy one t... |

4 |
it isn’t quite that simple
- Well
- 1992
(Show Context)
Citation Context ...prover might be distinguished by a few features, which we consider in the following sections. Logical expressiveness The typical computer algebra system supports a rather limited style of interaction =-=[27]-=-. The user types in an expression E; the CAS cogitates, usually not for very long, before returning another expression E ′ . The implication is that we should accept the theorem E = E ′ . Occasionally... |

4 |
Realization of a geometry-theorem proving machine
- Gelerntner
- 1959
(Show Context)
Citation Context ...cial Intelligence’ (AI) approach. For example in the 1950s Newell and Simon [81] designed a program that could prove many of the simple logic theorems in Principia Mathematica [109], while Gelerntner =-=[38]-=- designed a prover that could prove facts in Euclidean geometry using human-style diagrams to direct or restrict the proofs. A quite different approach was taken by other pioneers such as Gilmore [39]... |

3 |
Formal techniques and sizeable programs (EWD563
- Dijkstra
- 1976
(Show Context)
Citation Context ...of difficult mathematical theorems automatically. Besides, one might object that even if it were possible, it is hardly desirable to automate proofs that humans are incapable of developing themselves =-=[35]-=-: [...] I consider mathematical proofs as a reflection of my understanding and ‘understanding’ is something we cannot delegate, either to another person or to a machine. A more easily attained goal, a... |