## Reasoning about knowledge in linear logic: modalities and complexity (2004)

Venue: | Logic, Epistemology, and the Unity of Science |

Citations: | 5 - 5 self |

### BibTeX

@INPROCEEDINGS{Marion04reasoningabout,

author = {Mathieu Marion and Mehrnouche Sadrzadeh},

title = {Reasoning about knowledge in linear logic: modalities and complexity},

booktitle = {Logic, Epistemology, and the Unity of Science},

year = {2004},

publisher = {Kluwer}

}

### OpenURL

### Abstract

In a recent paper, Jean-Yves Girard commented that ”it has been a long time since philosophy has stopped intereacting with logic”[17]. Actually, it has not

### Citations

742 |
Situations and attitudes
- Barwise, Perry
- 1983
(Show Context)
Citation Context ...at Fagin and Halpern introduced a logic of ’awareness’ [?], where an agent knows explicitely φ if she is aware that φ and knows implicitely that φ.) Secondly, Levesque uses the situation semantics of =-=[3]-=- to deal with the explicit knowledge of agents. In a given situation, some formulas will have a truth-value assigned to them but it is possible some other formulas can have no truth-value. Levesque al... |

485 |
The calculus of constructions
- Coquand, Huet
- 1988
(Show Context)
Citation Context ...ble to escape exponential complexity, especially in multi-agent systems. In the concluding section of our paper, we shall suggest one way out, namely the encoding of proofs in the proof assistant Coq =-=[7]-=-. The following considerations are of an exploratory and programmatic -in other words: philosophical- nature. There are no new results, except an coding in Coq of two different fragments of modal line... |

451 |
The formulae-as-types notion of construction
- Howard
- 1980
(Show Context)
Citation Context ...recent developments within logic have left philosophers far behind. Prawitz’s timely book Natural Deduction [?], along with a key result obtained at around the same time, the Curry-Howard isomorphism =-=[21]-=-, initiated deep changes within logic; to wit, the development of the substructural logics [?, ?]. Dummett developed within proof-theoretical semantics a mature version of his anti-realist challenge, ... |

347 |
Intuitionistic Type Theory
- Martin-Löf
- 1984
(Show Context)
Citation Context ...ear logic proofs and computer programs and allows us to see cut-elimination as computation. This is a powerful paradigm that has taken over proof theory but, although it has been argued for in, e.g., =-=[28]-=-, it has hardly been noticed by philosophers. The Curry-Howard isomorphism is extended in [1] to classical as well as to the intuitionistic fragment of linear logic. Through this computational interpr... |

228 |
Une extension de l’interpretation de Gödel a l’analyse, et son application a l’elimination des coupure dans l’analyse et la theorie des types
- Girard
- 1971
(Show Context)
Citation Context ...ed tools and one such tool is the proof assistant Coq, which is a higher-order logic based on the calculus of constructions, whose ancestors are to be found in de Bruijn’s Automath, Girard’s system F =-=[13]-=- and Martin-Löf’s intuitionistic type theory [28]. The calculus of constructions enables us constructively to encode other logics in Coq. These logics are treated as object logics vs the metalogic, wh... |

196 |
Knowledge and Belief: An Introduction to the Logic of the Two Notions
- Hintikka
- 1962
(Show Context)
Citation Context ...ty cannot be eluded. The problem of logical omniscience in epistemic logic is a perfect case of an idealization in the above sense. It is usually presented in Hintikka’s original Hilbert-style system =-=[19]-=-. We shall present a Gentzen-style version below. Informally the problem is this: if an agent knows that p and knows that ’p implies q’, deductive closure requires that the agent also knows that q. Th... |

159 | logic: Its syntax and semantics
- Linear
- 1993
(Show Context)
Citation Context ...o the whole argument is seriously to take into account the physical cost of the proof, which is precisely the initial motivation for the development of linear logic, according 3sto Girard (see, e.g., =-=[16]-=-). The optional discharge in the case of the introduction rule for implication has the structural rule of weakening on the left as its counterpart. Relevant logic and linear logic both reject it, as o... |

105 | The complexity of propositional proofs
- Urquhart
- 1995
(Show Context)
Citation Context ...ed. Although the topic was hardly dealt with a mere 25 years ago, computational complexity is by now a well-studied phenomenon, with proof-theoretical measures of complexity, see, e.g., the survey in =-=[33]-=-. One obvious proposal here would be to limit oneself to polynomial time. (This has been suggested, e.g., in [26] for epistemic logic, because of applications to cognitive science.) There are candidat... |

62 |
Logic and the complexity of reasoning
- Levesque
- 1988
(Show Context)
Citation Context ...ed phenomenon, with proof-theoretical measures of complexity, see, e.g., the survey in [33]. One obvious proposal here would be to limit oneself to polynomial time. (This has been suggested, e.g., in =-=[26]-=- for epistemic logic, because of applications to cognitive science.) There are candidates for this in linear logic, such as the Bounded Linear Logic (BLL) of [18], in which the use of exponentials is ... |

19 | Gentzen method in modal calculi - Ohnishi, Matsumoto - 1959 |

10 |
Making proofs without modus ponens: An introduction to the combinatorics and complexity of cut elimination
- Carbone, Semmes
- 1997
(Show Context)
Citation Context ...tatements that are true that we cannot understand in practice because it would take too long, and that there are statements which we can understand if we permit ourselves to use cuts and not otherwise=-=[6]-=-, p.137 It is for these reasons, which need to be argued for in a more substantial manner (counterarguments easily spring to mind), that we chose to deal with 18sthe problem from a computational viewp... |

8 |
A Logic of Implicit and Explicit
- Levesque
- 1984
(Show Context)
Citation Context ... Further reasons for its inadequacy in the epistemic context will surface below. The only attempt that we know of at developing an epistemic logic by using a substructural logic is by Hector Levesque =-=[24, 25]-=-, who used relevant logic. 6sLevesque’s approach is original in many respects. First, he distinguishes between ’implicit’ and ’explicit’ knowledge. According to Levesque, the possible worlds semantics... |

7 | Cut-elimination for full intuitionistic linear logic
- Braüner, Paiva
- 1996
(Show Context)
Citation Context ... not settle the matter entirely, as one could attempt to re-phrase the puzzle (without any loss of meaning) so that it could be expressed in MMLL and use the intuitionistic version of . introduced in =-=[4]-=-, but this is mere speculation. Therefore, in order to be able to solve the puzzle, we had to work with the full modal linear logic. So we had to add lists to both sides of our sequents: D1 � ‘A ⊢ D2 ... |

4 |
Computational interpretations of linear logic
- Abramski
- 1993
(Show Context)
Citation Context ...ments easily spring to mind), that we chose to deal with 18sthe problem from a computational viewpoint and we chose to have an encoding in Coq. The computational approach to linear logic initiated in =-=[1]-=- links it to functional programming languages and the key here is an extension the Curry-Howard isomorphism [21], which establishes a correspondence between linear logic proofs and computer programs a... |

4 |
Y.Moses, and M.Y.Vardi, Reasoning about Knowledge
- Fagin, Halpern
- 1995
(Show Context)
Citation Context ...L1A L1B ⊢ L1B −◦ L L1A, L1A −◦ L1B ⊢ L1B 3d congruence L1A, L1(A −◦ B) ⊢ L1B One of the standard puzzles for multi-modal epistemic logic is ’wise men’ or ’King, three wise men and 5 hats’ puzzle (see =-=[11]-=-, p. 12): a king has three wise men and 5 hats: 2 green and 3 red. He asks the wise men to close their eyes and puts a hat on the head of each of them. Then asks them to open their eyes and poses a qu... |

4 |
Working with Linear Logic
- Power, Webster, et al.
- 1999
(Show Context)
Citation Context ...encoded in Coq’s higher-order logic; these encodings allows us to state and prove theorems using facilities of this proof assistant. Intuitionistic linear logic has been previously encoded in Coq, in =-=[32]-=-, by associating the constructs of Coq together with linear logic proofs. Modal logic, too, has been previously encoded in Coq in [23]. Our encoding method will be based on that of [12], in which the ... |

3 | Decidability Extracted: Synthesizing “Correct-by-Construction” Decision Procedures from Constuctive Proofs
- Caldwell
- 1998
(Show Context)
Citation Context ...object logics vs the metalogic, which is the higher-order logic of Coq. The key point here is that, once theorem-proving in these logics becomes automated in Coq, then one has constructive programmes =-=[5]-=-. Hence, the encoding of our two modal linear logics in Coq provides us with automated proofs which lead to constructive programs. Proof automation has not been examined in this paper but, as mentione... |

3 | Radical Antirealism and Substructural Logics’, to appear - Dubucs, Marion - 2003 |

3 |
Two-level meta-reasoning
- Felty
(Show Context)
Citation Context ... own ideas about ’impossible possible worlds’ [?], based on Rantala’s ’urn’ models [?, ?], to the syntactical solutions of [?, ?], Parikh’s ’knowledge algorithms’ [?, ?], and the logic of ’awareness’ =-=[12]-=-. This is not the place for a critical evaluation of these alternatives. It should be pointed out, however, that number of these solutions can be characterized by the wish to adhere come what may to t... |

3 | degree in signal processing in 1996, both from Lule3 University of Technology, Lule, Sweden. He spent six months as a Researcher with the Division
- unknown authors
- 1992
(Show Context)
Citation Context ...ng system and a general proof strategy. Guidelines for context-handling are mentioned and used succesfully in [32]. The general proof strategy can be found in the linear logic programming approach in =-=[20]-=-. Once proofs are automated, Coq provides a mechanism for construction of programmes out of automated theorem proving. As we just pointed out, automated proofs in Coq are constructed proofs that are c... |

2 |
Syntax and Semantics of Linear Logic
- Avron
- 1988
(Show Context)
Citation Context ...didate for further semantical and philosophical elaboration. There are two main strategies for introducing modalities within linear logic. First, one could interpret exponentials as modalities, as in =-=[2]-=-, the system would be a form of linear S4 with indexed modalities. However, the resulting connectives would both control contraction and weakening and serve as modalities. They would appear in both st... |

2 |
A logic of implicit and explicit belief, Fairchild FLAIR
- Levesque
- 1984
(Show Context)
Citation Context ... Further reasons for its inadequacy in the epistemic context will surface below. The only attempt that we know of at developing an epistemic logic by using a substructural logic is by Hector Levesque =-=[24, 25]-=-, who used relevant logic. 6sLevesque’s approach is original in many respects. First, he distinguishes between ’implicit’ and ’explicit’ knowledge. According to Levesque, the possible worlds semantics... |

2 |
Modal and Fixpoint Linear Logic
- Martin
- 2002
(Show Context)
Citation Context ...ld appear in both structural and modal rules; a very bizarre cocktail. Secondly, one could add modalities to fragments of linear logic. Some such combinations and their semantics have been studied in =-=[27]-=-.In what follows, we shall assess only two of them, namely the multi-modal linear logic or MMLL of [22], which has been inspired by and shown to have applications in reasoning about location-dependent... |

1 |
Feasibility in
- Dubucs
(Show Context)
Citation Context ...gents reasoning about knowledge without conceiving of them as omniscient. There are no reasons, except philosophical prejudice, not to explore the substructural world. We prefer here to follow Dubucs =-=[9]-=- and try and look for a weaker logic relatively to which agents can be said to be omniscient, without this omniscience being problematic. In that paper, Dubucs only discussed intuitionistic logic as a... |

1 |
On the Meaning of Logical Rules I:Syntax vs. Semantics’, http://iml.univ-mrs.fr/ girard/Articles.html
- Girard
- 1998
(Show Context)
Citation Context ...hieu Marion and Mehrnouche Sadrzadeh Université du Québec à Montréal ∗ In a recent paper, Jean-Yves Girard commented that ”it has been a long time since philosophy has stopped intereacting with logic”=-=[17]-=-. Actually, it has not been such a ”long” time since, e.g., Dag Prawitz and Michael Dummett developed philosophical arguments within the paradigm of Gentzen-style systems in favour of the adoption of ... |

1 |
A.Scedrov and P.Scott, ’Bounded Linear Logic, A Modular Approach to Polynomial-Time
- Girard
- 1992
(Show Context)
Citation Context ...e. (This has been suggested, e.g., in [26] for epistemic logic, because of applications to cognitive science.) There are candidates for this in linear logic, such as the Bounded Linear Logic (BLL) of =-=[18]-=-, in which the use of exponentials is bounded in advance, or the more recent Light Linear Logic (LLL) of [17], that has a (locally) polynomial-time cut-elimination. Both BLL and LLL are strong enough ... |

1 |
Epistemic Logic in Higher Order Logic
- Lescanne
- 2001
(Show Context)
Citation Context ...ntuitionistic linear logic has been previously encoded in Coq, in [32], by associating the constructs of Coq together with linear logic proofs. Modal logic, too, has been previously encoded in Coq in =-=[23]-=-. Our encoding method will be based on that of [12], in which the system to be encoded is treated as the object logic and Coq’s Calculus of Constructions (CC) as the metalogic. We are here encoding fo... |