Logic and Artificial Intelligence Lecture 18

Transcription

1 Logic and Artificial Intelligence Lecture 18 Eric Pacuit Currently Visiting the Center for Formal Epistemology, CMU Center for Logic and Philosophy of Science Tilburg University ai.stanford.edu/ epacuit November 8, 2011 Logic and Artificial Intelligence 1/30

2 Ingredients of a Logical Analysis of Rational Agency informational attitudes (eg., knowledge, belief, certainty) time, actions and ability motivational attitudes (eg., preferences) group notions (eg., common knowledge and coalitional ability) normative attitudes (eg., obligations) Logic and Artificial Intelligence 2/30

3 Ingredients of a Logical Analysis of Rational Agency informational attitudes (eg., knowledge, belief, certainty) time, actions and ability motivational attitudes (eg., preferences) group notions (eg., common knowledge and coalitional ability) normative attitudes (eg., obligations) Logic and Artificial Intelligence 2/30

4 Time One of the most successful applications of modal logic is in the logic of time. Logic and Artificial Intelligence 3/30

5 Time One of the most successful applications of modal logic is in the logic of time. Many variations discrete or continuous branching or linear point based or interval based See, for example, Antony Galton. Temporal Logic. Stanford Encyclopedia of Philosophy: http: //plato.stanford.edu/entries/logic-temporal/. I. Hodkinson and M. Reynolds. Temporal Logic. Handbook of Modal Logic, Logic and Artificial Intelligence 3/30

6 Models of Time T = T, <, V where T is a set of time points (or moments), < T T is the precedence relation: s < t means time point s precedes time point t (or s occurs earlier than t) and V : At (T ) is a valuation function (describing when the atomic propositions are true). Logic and Artificial Intelligence 4/30

7 Models of Time T = T, <, V where T is a set of time points (or moments), < T T is the precedence relation: s < t means time point s precedes time point t (or s occurs earlier than t) and V : At (T ) is a valuation function (describing when the atomic propositions are true). < is typically assumed to be irreflexive and transitive (a strict partial order). Logic and Artificial Intelligence 4/30

8 Models of Time T = T, <, V where T is a set of time points (or moments), < T T is the precedence relation: s < t means time point s precedes time point t (or s occurs earlier than t) and V : At (T ) is a valuation function (describing when the atomic propositions are true). < is typically assumed to be irreflexive and transitive (a strict partial order). Examples: N, <, Z, <, Q, <, R, < Logic and Artificial Intelligence 4/30

9 Other properties of < Linearity: for all s, t T, s < t or s = t of t < s Past-linear: for all s, x, y T, if x < s and y < s, then either x < y or x = y or y < x Denseness for all s, t T, if s < t then there is a z T such that s < z and z < t Discreteness: for all s, t T, if s < t then there is a z such that (s < z and there is no u such that s < u and u < z) Logic and Artificial Intelligence 5/30

10 Priorean Temporal Logic L t be defined by the following grammar p ϕ ϕ ψ Gϕ Hϕ where p At. Logic and Artificial Intelligence 6/30

11 Priorean Temporal Logic L t be defined by the following grammar p ϕ ϕ ψ Gϕ Hϕ where p At. Gϕ: ϕ is going to become true Hϕ: ϕ has been true F ϕ := G ϕ: ϕ is true in the future Pϕ := H ϕ: ϕ was true some time in the past Logic and Artificial Intelligence 6/30

12 M = T, <, V M, t = p iff t V (p) M, t = ϕ iff M, t = ϕ M, t = ϕ ψ iff M, t = ϕ and M, t = ψ M, t = Gϕ iff for all s T, if t < s then M, s = ϕ M, t = Hϕ iff for all s T, if s < t thenm, s = ϕ M, t = F ϕ iff there is s T such that t < s and M, s = ϕ M, t = Pϕ iff there is s T such that s < t and M, s = ϕ Logic and Artificial Intelligence 7/30

13 Frame Correspondence H PH is valid Logic and Artificial Intelligence 8/30

14 Frame Correspondence H PH is valid iff there is a starting point Logic and Artificial Intelligence 8/30

15 Frame Correspondence H PH is valid iff there is a starting point P is valid Logic and Artificial Intelligence 8/30

16 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point Logic and Artificial Intelligence 8/30

17 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid Logic and Artificial Intelligence 8/30

18 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point Logic and Artificial Intelligence 8/30

19 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid Logic and Artificial Intelligence 8/30

20 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint Logic and Artificial Intelligence 8/30

21 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid Logic and Artificial Intelligence 8/30

22 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching Logic and Artificial Intelligence 8/30

23 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching FPϕ (F ϕ ϕ Pϕ) is valid Logic and Artificial Intelligence 8/30

24 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching FPϕ (F ϕ ϕ Pϕ) is valid iff the past is non-branching Logic and Artificial Intelligence 8/30

25 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching FPϕ (F ϕ ϕ Pϕ) is valid iff the past is non-branching F ϕ FF ϕ is valid Logic and Artificial Intelligence 8/30

26 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching FPϕ (F ϕ ϕ Pϕ) is valid iff the past is non-branching F ϕ FF ϕ is valid iff the flow of time is dense Logic and Artificial Intelligence 8/30

27 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching FPϕ (F ϕ ϕ Pϕ) is valid iff the past is non-branching F ϕ FF ϕ is valid iff the flow of time is dense (F ϕ Hϕ) FHϕ is valid Logic and Artificial Intelligence 8/30

28 Frame Correspondence H PH is valid iff there is a starting point P is valid iff there is no starting point G FG is valid iff there is an end point F is valid iff there is no endpoint PF ϕ (Pϕ ϕ F ϕ) is valid iff the future is not branching FPϕ (F ϕ ϕ Pϕ) is valid iff the past is non-branching F ϕ FF ϕ is valid iff the flow of time is dense (F ϕ Hϕ) FHϕ is valid iff the flow of time is discrete Logic and Artificial Intelligence 8/30

29 Basic Temporal Logic All classical propositional tautologies Distribution G(ϕ ψ) (Gϕ Gψ) G(ϕ ψ) (Gϕ Gψ) Converse ϕ GPϕ ϕ HF ϕ Transitivity: Gϕ GGϕ Modus Ponens: from ϕ and ϕ ψ infer ψ Temporal Generalization: from ϕ infer F ϕ; from ϕ infer Gϕ Logic and Artificial Intelligence 9/30

30 Basic Temporal Logic All classical propositional tautologies Distribution G(ϕ ψ) (Gϕ Gψ) G(ϕ ψ) (Gϕ Gψ) Converse ϕ GPϕ ϕ HF ϕ Transitivity: Gϕ GGϕ Modus Ponens: from ϕ and ϕ ψ infer ψ Temporal Generalization: from ϕ infer F ϕ; from ϕ infer Gϕ Theorem. The above logic is sound and complete with respect to the class of all flows of time Logic and Artificial Intelligence 9/30

31 Logic of Linear Time Theorem. The above logic with the linearity axioms is sound and complete with respect to the class of all linear flows of time PF ϕ (Pϕ ϕ F ϕ) FPϕ (F ϕ ϕ Pϕ) Logic and Artificial Intelligence 10/30

32 Other Languages: Since and Until M, t = ϕuψ iff M, s = ψ for some s such that t < s and M, u = ϕ for all u with t < u < s M, t = ϕsψ iff M, s = ψ for some s such that s < t and M, u = ϕ for all u with s < u < t Logic and Artificial Intelligence 11/30

33 Other Languages: Since and Until M, t = ϕuψ iff M, s = ψ for some s such that t < s and M, u = ϕ for all u with t < u < s M, t = ϕsψ iff M, s = ψ for some s such that s < t and M, u = ϕ for all u with s < u < t Theorem (Kamp). Over the class of linear, continuous orderings, every temporal operator can be defined using the above modalities Logic and Artificial Intelligence 11/30

34 Branching Time Each moment t T can be decided into the Past(t) = {s T s < t} and the Future(t) = {s T t < s} ( A-series ) Typically, it is assumed that the past is linear, but the future may be branching. Logic and Artificial Intelligence 12/30

35 Branching Time Each moment t T can be decided into the Past(t) = {s T s < t} and the Future(t) = {s T t < s} ( A-series ) Typically, it is assumed that the past is linear, but the future may be branching. F ϕ: it will be the case that ϕ ϕ will be the case in the case in the actual course of events or no matter what course of events Logic and Artificial Intelligence 12/30

36 Branching Time Logics A branch b in T, < is a maximal linearly ordered subset of T s T is on a branch b of T provided s b (we also say b is a branch going through t ). Logic and Artificial Intelligence 13/30

37 Branching Time Logics A branch b in T, < is a maximal linearly ordered subset of T s T is on a branch b of T provided s b (we also say b is a branch going through t ). M, t, b = p iff t V (p) M, t, b = ϕ iff M, t, b = ϕ M, t, b = ϕ ψ iff M, t, b = ϕ and M, t = ψ M, t, b = Gϕ iff for all s T, if s is on b and t < s then M, s, b = ϕ M, t, b = Hϕ iff for all s T, if s is on b and s < t then M, s, b = ϕ M, t, b = ϕ iff M, s, c = ϕ for all branches c through t Logic and Artificial Intelligence 13/30

38 Computational vs. Behavioral Structures q 0 x = 1 q 0 q 0 q 0 q 0 q 1 q 1 x = 2 q 0 q 0 q 0 q 0 q 0 q 1 q 0 q 1 q 0 q 0 q 1 q 1. Logic and Artificial Intelligence 14/30

39 Temporal Logics Logic and Artificial Intelligence 15/30

40 Temporal Logics Linear Time Temporal Logic: Reasoning about computation paths: F ϕ: ϕ is true some time in the future. A. Pnuelli. A Temporal Logic of Programs. in Proc. 18th IEEE Symposium on Foundations of Computer Science (1977). Logic and Artificial Intelligence 15/30

41 Temporal Logics Linear Time Temporal Logic: Reasoning about computation paths: F ϕ: ϕ is true some time in the future. A. Pnuelli. A Temporal Logic of Programs. in Proc. 18th IEEE Symposium on Foundations of Computer Science (1977). Branching Time Temporal Logic: Allows quantification over paths: F ϕ: there is a path in which ϕ is eventually true. E. M. Clarke and E. A. Emerson. Design and Synthesis of Synchronization Skeletons using Branching-time Temproal-logic Specifications. In Proceedings Workshop on Logic of Programs, LNCS (1981). Logic and Artificial Intelligence 15/30

42 Temporal Logics q 0 x = 1 q 0 Fp x=2 q 0 q 0 q 0 q 1 q 1 x = 2 q 0 q 0 q 0 q 0 q 0 q 1 q 0 q 1 q 0 q 0 q 1 q 1. Logic and Artificial Intelligence 16/30

43 Temporal Logics q 0 x = 1 q 0 Fp x=2 q 0 q 0 q 0 q 1 q 1 x = 2 q 0 q 0 q 0 q 0 q 0 q 1 q 0 q 1 q 0 q 0 q 1 q 1. Logic and Artificial Intelligence 16/30

44 Temporal Logics q 0 x = 1 q 0 Fp x=2 q 0 q 0 q 0 q 1 q 1 x = 2 q 0 q 0 q 0 q 0 q 0 q 1 q 0 q 1 q 0 q 0 q 1 q 1. Logic and Artificial Intelligence 16/30

45 Interval Values J. Allen and G. Ferguson. Actions and Events in Interval Temporal Logics. Journal of Logic and Computation, J. Halpern and Y. Shoham. A Propositional Modal Logic of Time Intervals. Journal of the ACM, 38:4, pp , J. van Benthem. Logics of Time. Kluwer, Logic and Artificial Intelligence 17/30

46 Interval Temporal Logics Let T = T, < be a frame and I (T ) = {[a, b] a, b T and a b} be the set of intervals over T Models are M = I (T ), {R X }, V where R X I (T ) I (T ) and V : At (I (T )). Logic and Artificial Intelligence 18/30

47 Interval Temporal Logics Let T = T, < be a frame and I (T ) = {[a, b] a, b T and a b} be the set of intervals over T Models are M = I (T ), {R X }, V where R X I (T ) I (T ) and V : At (I (T )). M, [a, b] = p iff [a, b] V (p) M, [a, b] = pt iff a = b M, [a, b] = X ϕ iff there is an interval [c, d] such that [a, b]r X [c, d] and M, [c, d] = ϕ Logic and Artificial Intelligence 18/30

48 a b hai hli hbi hei hdi hoi [a, b]r A [c, d], b = c [a, b]r L [c, d], b<c [a, b]r B [c, d], a = c, d < b [a, b]r E [c, d], b = d, a < c [a, b]r D [c, d], a<c,d<b [a, b]r O [c, d], a<c<b<d c d c c c c d d c d d d Table I Logic and Artificial Intelligence 19/30

49 High Undecidability! D. Bersolin et al.. The dark side of interval temporal logic: sharpening the undecidability border Logic and Artificial Intelligence 20/30

50 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? Logic and Artificial Intelligence 21/30

51 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? 1. Control: The action a should be in the control of agent i or the truth of ϕ should be under i s control. Logic and Artificial Intelligence 21/30

52 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? 1. Control: The action a should be in the control of agent i or the truth of ϕ should be under i s control. 2. Repeatability: Agent i should be able to repeatedly do action a or repeatedly bring about formula ϕ. Logic and Artificial Intelligence 21/30

53 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? 1. Control: The action a should be in the control of agent i or the truth of ϕ should be under i s control. 2. Repeatability: Agent i should be able to repeatedly do action a or repeatedly bring about formula ϕ. 3. Avoidability: Agent i should be able to avoid doing action a. Logic and Artificial Intelligence 21/30

54 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? 1. Control: The action a should be in the control of agent i or the truth of ϕ should be under i s control. 2. Repeatability: Agent i should be able to repeatedly do action a or repeatedly bring about formula ϕ. 3. Avoidability: Agent i should be able to avoid doing action a. 4. Free-will: Agent i should be free to decide whether or not to do action a. That is, the agent should have a choice as to whether or not to do action a or bring about formula ϕ. Logic and Artificial Intelligence 21/30

55 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? 1. Control: The action a should be in the control of agent i or the truth of ϕ should be under i s control. 2. Repeatability: Agent i should be able to repeatedly do action a or repeatedly bring about formula ϕ. 3. Avoidability: Agent i should be able to avoid doing action a. 4. Free-will: Agent i should be free to decide whether or not to do action a. That is, the agent should have a choice as to whether or not to do action a or bring about formula ϕ. 5. Causality: Agent i should cause action a to take place or the formula ϕ to become true. Logic and Artificial Intelligence 21/30

56 Actions and Abilities: Pre-theoretic Intuitions What does it mean for an agent to be able to do some action a or bring about some state of affairs ϕ? 1. Control: The action a should be in the control of agent i or the truth of ϕ should be under i s control. 2. Repeatability: Agent i should be able to repeatedly do action a or repeatedly bring about formula ϕ. 3. Avoidability: Agent i should be able to avoid doing action a. 4. Free-will: Agent i should be free to decide whether or not to do action a. That is, the agent should have a choice as to whether or not to do action a or bring about formula ϕ. 5. Causality: Agent i should cause action a to take place or the formula ϕ to become true. 6. Intentionality: Agent i should intentionally do action a or bring about ϕ. Logic and Artificial Intelligence 21/30

57 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? Logic and Artificial Intelligence 22/30

58 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? 1. Abl i ϕ ϕ (or ϕ Abl i ϕ) Logic and Artificial Intelligence 22/30

59 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? 1. Abl i ϕ ϕ (or ϕ Abl i ϕ) 2. Abl i Logic and Artificial Intelligence 22/30

60 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? 1. Abl i ϕ ϕ (or ϕ Abl i ϕ) 2. Abl i 3. (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) Logic and Artificial Intelligence 22/30

61 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? 1. Abl i ϕ ϕ (or ϕ Abl i ϕ) 2. Abl i 3. (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) 4. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) Logic and Artificial Intelligence 22/30

62 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? 1. Abl i ϕ ϕ (or ϕ Abl i ϕ) 2. Abl i 3. (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) 4. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) 5. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) Logic and Artificial Intelligence 22/30

63 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? 1. Abl i ϕ ϕ (or ϕ Abl i ϕ) 2. Abl i 3. (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) 4. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) 5. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) 6. Abl i Abl j ϕ Abl i ϕ, Abl i Abl i ϕ Abl i ϕ Logic and Artificial Intelligence 22/30

64 Games: (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) Logic and Artificial Intelligence 23/30

65 Games: (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) s 1 A s 2 B B s 3 p p, q p, q q Logic and Artificial Intelligence 23/30

66 Games: (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) s 1 A s 2 B B s 3 p p, q p, q q s 1 = Abl A p Logic and Artificial Intelligence 23/30

67 Games: (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) s 1 A s 2 B B s 3 p p, q p, q q s 1 = Abl A p Abl A q Logic and Artificial Intelligence 23/30

68 Games: (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) s 1 A s 2 B B s 3 p p, q p, q q s 1 = Abl A p Abl A q Abl A (p q) Logic and Artificial Intelligence 23/30

69 Games: (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) R. Parikh. The Logic of Games and its Applications. Annals of Discrete Mathematics (1985). M. Pauly and R. Parikh. Game Logic An Overview. Studia Logica (2003). J. van Benthem. Logic and Games. Course notes (2007). Logic and Artificial Intelligence 23/30

70 ϕ Abl i ϕ Suppose an agent (call her Ann) is throwing a dart and she is not a very good dart player, but she just happens to throw a bull s eye. Logic and Artificial Intelligence 24/30

71 ϕ Abl i ϕ Suppose an agent (call her Ann) is throwing a dart and she is not a very good dart player, but she just happens to throw a bull s eye. Intuitively, we do not want to say that Ann has the ability to throw a bull s eye even though it happens to be true. Logic and Artificial Intelligence 24/30

72 Abl i (ϕ ψ) Abl i ϕ Abl i ψ Continuing with this example, suppose that Ann has the ability to hit the dart board, but has no other control over the placement of the dart. Logic and Artificial Intelligence 25/30

73 Abl i (ϕ ψ) Abl i ϕ Abl i ψ Continuing with this example, suppose that Ann has the ability to hit the dart board, but has no other control over the placement of the dart. Now, when she throws the dart, as a matter of fact, it will either hit the top half of the board or the bottom half of the board. Logic and Artificial Intelligence 25/30

74 Abl i (ϕ ψ) Abl i ϕ Abl i ψ Continuing with this example, suppose that Ann has the ability to hit the dart board, but has no other control over the placement of the dart. Now, when she throws the dart, as a matter of fact, it will either hit the top half of the board or the bottom half of the board. Since, Ann has the ability to hit the dart board, she has the ability to either hit the top half of the board or the bottom half of the board. Logic and Artificial Intelligence 25/30

75 Abl i (ϕ ψ) Abl i ϕ Abl i ψ Continuing with this example, suppose that Ann has the ability to hit the dart board, but has no other control over the placement of the dart. Now, when she throws the dart, as a matter of fact, it will either hit the top half of the board or the bottom half of the board. Since, Ann has the ability to hit the dart board, she has the ability to either hit the top half of the board or the bottom half of the board. However, intuitively, it seems true that Ann does not have the ability to hit the top half of the dart board, and also she does not have the ability to hit the bottom half of the dart board. Logic and Artificial Intelligence 25/30

76 Abilities Abl i ϕ: agent i has the ability to bring about (see to it that) ϕ is true What are core logical principles? Depends very much on the intended application and how actions are represented Abl i ϕ ϕ (or ϕ Abl i ϕ) 2. Abl i 3. (Abl i ϕ Abl i ψ) Abl i (ϕ ψ) 4. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) 5. Abl i (ϕ ψ) (Abl i ϕ Abl i ψ) 6. Abl i Abl j ϕ Abl i ϕ, Abl i Abl i ϕ Abl i ϕ Logic and Artificial Intelligence 26/30

77 D. Elgesem. The modal logic of agency. Nordic Journal of Philosophical Logic 2(2), 1-46, G. Governatori and A. Rotolo. On the Axiomatisation of Elgesem s Logic of Agency and Ability. Journal of Philosophical Logic, 34, pgs (2005). Logic and Artificial Intelligence 27/30

78 A Minimal Logic of Abilities Cϕ means the agent is capable of realizing ϕ Eϕ means the agent does bring about ϕ Logic and Artificial Intelligence 28/30

79 A Minimal Logic of Abilities Cϕ means the agent is capable of realizing ϕ Eϕ means the agent does bring about ϕ 1. All propositional tautologies 2. C 3. Eϕ Eψ E(ϕ ψ) 4. Eϕ ϕ 5. Eϕ Cϕ 6. Modus Ponens plus from ϕ ψ infer Eϕ Eψ and from ϕ ψ infer Cϕ Cψ Logic and Artificial Intelligence 28/30

80 Brief digression: weak systems of modal logic Logic and Artificial Intelligence 29/30

81 PC Propositional Calculus E ϕ ϕ M (ϕ ψ) ( ϕ ψ) C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ Logic and Artificial Intelligence 30/30

82 PC Propositional Calculus E ϕ ϕ M (ϕ ψ) ( ϕ ψ) C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ A modal logic L is classical if it contains all instances of E and is closed under RE. Logic and Artificial Intelligence 30/30

83 PC Propositional Calculus E ϕ ϕ M (ϕ ψ) ( ϕ ψ) C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ A modal logic L is classical if it contains all instances of E and is closed under RE. E is the smallest classical modal logic. Logic and Artificial Intelligence 30/30

84 PC Propositional Calculus E ϕ ϕ M (ϕ ψ) ( ϕ ψ) C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. In E, M is equivalent to ϕ ψ (Mon) ϕ ψ Logic and Artificial Intelligence 30/30

85 PC Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon Logic and Artificial Intelligence 30/30

86 PC 6. Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon EC is the logic E + C Logic and Artificial Intelligence 30/30

87 PC Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon EC is the logic E + C EMC is the smallest regular modal logic Logic and Artificial Intelligence 30/30

88 PC Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon EC is the logic E + C EMC is the smallest regular modal logic A logic is normal if it contains all instances of E, C and is closed under Mon and Nec Logic and Artificial Intelligence 30/30

89 PC Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon EC is the logic E + C EMC is the smallest regular modal logic K is the smallest normal modal logic Logic and Artificial Intelligence 30/30

90 PC Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon EC is the logic E + C EMC is the smallest regular modal logic K = EMCN Logic and Artificial Intelligence 30/30

91 PC Propositional Calculus E ϕ ϕ ϕ ψ Mon ϕ ψ C ( ϕ ψ) (ϕ ψ) N K (ϕ ψ) ( ϕ ψ) RE Nec ϕ ψ ϕ ψ ϕ ϕ MP ϕ ϕ ψ ψ E is the smallest classical modal logic. EM is the logic E + Mon EC is the logic E + C EMC is the smallest regular modal logic K = PC(+E) + K + Nec + MP Logic and Artificial Intelligence 30/30