Agreement Theorems from the Perspective of Dynamic-Epistemic Logic

Transcription

1 Agreement Theorems from the Perspective of Dynamic-Epistemic Logic Olivier Roy & Cedric Dégremont November 10, 2008 Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 1

2 Introduction and preliminaries Overview 1. Introduction to Agreements Theorems (AT) 2. Models of knowledge and beliefs, common priors and common knowledge 3. Three variations on the result: static, kinematic and dynamic Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 2

3 Introduction and preliminaries Overview 1. Introduction to Agreements Theorems (AT) 2. Models of knowledge and beliefs, common priors and common knowledge 3. Three variations on the result: static, kinematic and dynamic Highlights: The received view: ATs undermine the role of private information. The DEL point of view: ATs show the importance of higher-order information. ATs show how static conditioning is different from real belief dynamics. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 2

4 Introduction and preliminaries Agreement Theorems in a nutshell Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 3

5 Introduction and preliminaries Agreement Theorems in a nutshell Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 3

6 Introduction and preliminaries Agreement Theorems in a nutshell Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 3

7 Introduction and preliminaries A short result goes a long way Original result: [Aumann, 1976]. No trade theorems: [Milgrom and Stokey, 1982]. Dynamic versions: [Geanakoplos and Polemarchakis, 1982] Qualitative generalizations: [Cave, 1983], [Bacharach, 1985]. Network structure: [Parikh and Krasucki, 1990] Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 4

8 Introduction and preliminaries A short result goes a long way Original result: [Aumann, 1976]. No trade theorems: [Milgrom and Stokey, 1982]. Dynamic versions: [Geanakoplos and Polemarchakis, 1982] Qualitative generalizations: [Cave, 1983], [Bacharach, 1985]. Network structure: [Parikh and Krasucki, 1990] Good survey: [Bonanno and Nehring, 1997]. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 4

9 Introduction and preliminaries Conclusions, the received view. Theorem If two people have the same priors, and their posteriors for an event A are common knowledge, then these posterior are the same. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 5

10 Introduction and preliminaries Conclusions, the received view. Theorem If two people have the same priors, and their posteriors for an event A are common knowledge, then these posterior are the same. How important is private information? Not quite... How strong is the common knowledge condition? Very... How plausible is the common prior assumption? Debated... Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 5

11 Conclusions, the point of view of DEL. Theorem If two people have the same priors, and their posteriors for an event A are common knowledge, then these posterior are the same. The key is higher-order information. One should distinguish information kinematics vs information dynamics, belief conditioning vs belief update. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 6

12 Definition (Epistemic-Doxastic Model) An epistemic-doxastic model M is a tuple W, I, { i, i } i I such that: Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 7

13 Definition (Epistemic-Doxastic Model) An epistemic-doxastic model M is a tuple W, I, { i, i } i I such that: W is a finite set of states. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 7

14 Definition (Epistemic-Doxastic Model) An epistemic-doxastic model M is a tuple W, I, { i, i } i I such that: W is a finite set of states. I is a finite set of agents. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 7

15 Definition (Epistemic-Doxastic Model) An epistemic-doxastic model M is a tuple W, I, { i, i } i I such that: W is a finite set of states. I is a finite set of agents. i is a reflexive, transitive and connected plausibility ordering on W. There is common priors iff i = j for all i and j in I. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 7

16 Definition (Epistemic-Doxastic Model) An epistemic-doxastic model M is a tuple W, I, { i, i } i I such that: W is a finite set of states. I is a finite set of agents. i is a reflexive, transitive and connected plausibility ordering on W. There is common priors iff i = j for all i and j in I. i is an epistemic accessibility equivalence relation. We write [w] i for {w : w i w }. See: [Board, 2004, Baltag and Smets, 2008, van Benthem, ] Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 7

17 Key notions: Knowledge: w = K i ϕ iff w = ϕ for all w i w. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 8

18 Key notions: Knowledge: w = K i ϕ iff w = ϕ for all w i w. Everybody knows: w = E I ϕ iff w = K 1 ϕ... K n ϕ for 1,...n I. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 8

19 Key notions: Knowledge: w = K i ϕ iff w = ϕ for all w i w. Everybody knows: w = E I ϕ iff w = K 1 ϕ... K n ϕ for 1,...n I. Common knowledge: w = CK I ϕ iff w = E I ϕ and w = E I E I ϕ and... Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 8

20 Key notions: Knowledge: w = K i ϕ iff w = ϕ for all w i w. Everybody knows: w = E I ϕ iff w = K 1 ϕ... K n ϕ for 1,...n I. Common knowledge: w = CK I ϕ iff w = E I ϕ and w = E I E I ϕ and... Beliefs: w = B ψ i ϕ iff w = ϕ for all w in max i ([w] i ψ ). We write B i ϕ for B i ϕ. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 8

21 Theorem (Static agreement) For any epistemic-doxastic model M with common priors, for all w we have that w = CK I (B i (E) B j (E)) where E W. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 9

22 The key property: Sure-thing principle If, first, you would believe, conditional on the fact that it is cloudy, that it will rain and, Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 10

23 The key property: Sure-thing principle If, first, you would believe, conditional on the fact that it is cloudy, that it will rain and, second, you would believe, conditional on the fact that it is not cloudy, that it will rain, then Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 10

24 The key property: Sure-thing principle If, first, you would believe, conditional on the fact that it is cloudy, that it will rain and, second, you would believe, conditional on the fact that it is not cloudy, that it will rain, then you unconditionally believe that it will rain. [Savage, 1954, Bacharach, 1985] Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 10

25 The key property: Sure-thing principle Agent 1 s IP W P 1 P 2 Agent 2 s IP W Q 1 Q 2 Q 3 Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 11

26 The key property: Sure-thing principle Agent 1 s IP W not B 1 (E/P 1 ) not B 1 (E/P 2 ) Agent 2 s IP W B 2 (E/Q 1 ) B 2 (E/Q 2 ) B 3 (E/Q 3 ) Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 11

27 The key property: Sure-thing principle By the Sure-Thing Principle The CK cell W not B 1 (E) The CK cell W B 2 (E) Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 11

28 The key property: Sure-thing principle The CK cell W Hence no common priors not B 1 (E) The CK cell W = B 2 (E) Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 11

29 Lesson: They might not have the same (first-order) information, but... Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 12

30 Lesson: They might not have the same (first-order) information, but... what is common knowledge is precisely where their (higher-order) information coincide. The cornerstone is higher-order information. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 12

31 Lesson: They might not have the same (first-order) information, but... what is common knowledge is precisely where their (higher-order) information coincide. Question: The cornerstone is higher-order information. How can CK obtain with respect to each others beliefs? Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 12

32 Lesson: They might not have the same (first-order) information, but... what is common knowledge is precisely where their (higher-order) information coincide. Question: The cornerstone is higher-order information. How can CK obtain with respect to each others beliefs? Dialogues or repeated announcements. [Geanakoplos and Polemarchakis, 1982, Bacharach, 1985] Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 12

33 Towards kinematic agreement Definition A kinematic dialogue about A is an epistemic-doxastic model M = W, I, { i, { n,i } n N } i I with Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 13

34 Towards kinematic agreement Definition A kinematic dialogue about A is an epistemic-doxastic model M = W, I, { i, { n,i } n N } i I with, for all i I, the sequence of epistemic accessibility relation { n,i } n N is inductively defined as follows (for 2 agents). Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 13

35 Towards kinematic agreement Definition A kinematic dialogue about A is an epistemic-doxastic model M = W, I, { i, { n,i } n N } i I with, for all i I, the sequence of epistemic accessibility relation { n,i } n N is inductively defined as follows (for 2 agents). 0,i is a given epistemic accessibility relation. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 13

36 Towards kinematic agreement Definition A kinematic dialogue about A is an epistemic-doxastic model M = W, I, { i, { n,i } n N } i I with, for all i I, the sequence of epistemic accessibility relation { n,i } n N is inductively defined as follows (for 2 agents). 0,i is a given epistemic accessibility relation. for all w W : [w] n+1,i = [w] n,i { B n (A) if w = B n,j (A) B n,j (A) otherwise. with B n,j (A) = {w : max j [w ] n,j A}. Intuition: B n+1,i ϕ B B n,j ϕ n,i ϕ. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 13

37 Lemma (Fixed-point) Every kinematic dialogue about A has a fixed-point, i.e. there is a n such that [w] n,i = [w] n +1,i for all w and i. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 14

38 Lemma (Fixed-point) Every kinematic dialogue about A has a fixed-point, i.e. there is a n such that [w] n,i = [w] n +1,i for all w and i. Lemma (Common knowledge) The posteriors beliefs at the fixed-point of a kinematic dialogue are common knowledge. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 14

39 Lemma (Fixed-point) Every kinematic dialogue about A has a fixed-point, i.e. there is a n such that [w] n,i = [w] n +1,i for all w and i. Lemma (Common knowledge) The posteriors beliefs at the fixed-point of a kinematic dialogue are common knowledge. Theorem (Kinematic agreement) For any kinematic dialogue about A, if there is common priors then at the fixed-point either all agents believe that A or they all don t believe that A. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 14

40 Lesson: Common knowledge arise from dialogues... Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 15

41 Lesson: Common knowledge arise from dialogues... Warnings from DEL: This is only a kinematic (i.e. conditioning) dialogue: the truth of A is fixed during the process. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 15

42 Lesson: Common knowledge arise from dialogues... Warnings from DEL: This is only a kinematic (i.e. conditioning) dialogue: the truth of A is fixed during the process. In general, this is not the case. A might be about the agents information. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 15

43 Lesson: Common knowledge arise from dialogues... Warnings from DEL: This is only a kinematic (i.e. conditioning) dialogue: the truth of A is fixed during the process. In general, this is not the case. A might be about the agents information. Another way to look at it: kinematic agreement = virtual agreement Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 15

44 Towards dynamic agreement Definition A dynamic dialogue about A is sequence of epistemic-doxastic pointed models {(M n, w)} n N such that: M 0 is a given epistemic-doxastic model. M n+1 = W n+1, I, n+1,i, n+1,i with W n+1 = {w W n : w = B n (A n )} with: A n = {w W n : w = A} B n(a n) is B n,i (A n) B n,j (A n) if w = B n,i (A n) B n,j (A n), etc... n+1,i n+1,i are the restrictions of n,i and n,i to W n+1. Intuition: B n+1,i ϕ [B n ϕ!]b n,i ϕ Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 16

45 Lemma (Fixed-point) Every dynamic dialogue about A has a fixed-point, i.e. there is a n such that: M n = M n +1 Lemma (Common knowledge) The posteriors beliefs at the fixed-point of a dynamic dialogue are common knowledge. Theorem (Dynamic agreement) For any dynamic dialogue about A, if there is common priors then at the fixed-point n either all agents believe that A n or they all don t believe that A n. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 17

46 Lemma (Fixed-point) Every dynamic dialogue about A has a fixed-point, i.e. there is a n such that: M n = M n +1 Lemma (Common knowledge) The posteriors beliefs at the fixed-point of a dynamic dialogue are common knowledge. Theorem (Dynamic agreement) For any dynamic dialogue about A, if there is common priors then at the fixed-point n either all agents believe that A n or they all don t believe that A n. But... Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 17

47 Kinematic agreements can be different from dynamic agreements. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 18

48 Kinematic agreements can be different from dynamic agreements. W w 1 w 2 p not p Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 18

49 Kinematic agreements can be different from dynamic agreements. W w 1 w 2 p not p 1 1 Let A = p B 2 p 2 Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 18

50 Kinematic agreements can be different from dynamic agreements. W w 1 w 2 p not p Let A = p B 2 p 1 1 At the fixed points for the kinematic and the dynamic dialogue about A, we have that [w 1 ] n,1 = [w 1 ] n,2 = {w 1 } 2 Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 18

51 Kinematic agreements can be different from dynamic agreements. W w 1 w 2 p not p Let A = p B 2 p 1 1 At the fixed points for the kinematic and the dynamic dialogue about A, we have that [w 1 ] n,1 = [w 1 ] n,2 = {w 1 } Kinematic beliefs: w = B n,i(p B 0,2 p), for i = 1, 2. Dynamic beliefs: w = B n,i(p B n,2p), for i = 1, 2. 2 Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 18

52 Conclusion Agreements theorems: Undermine the role of private information? Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 19

53 Conclusion Agreements theorems: Undermine the role of private information? A better way to look at it ( DEL methodology ): Rest on higher-order information and its role in interactive reasoning. Highlight the difference between belief kinematics and belief dynamics. (Hopefully not so distant) future work : General (countable) case? Announcements of reasons and not only of opinions? Relaxing the common prior assumption? Agreements on everything? Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 19

54 References Aumann, R. (1976). Agreeing to disagree. The Annals of Statistics, 4(6): Bacharach, M. (1985). Some extensions of a claim of aumann in an axiomatic model of knowledge. Journal of Economic Theory, 37(1): Baltag, A. and Smets, S. (2008). A qualitative theory of dynamic interactive belief revision. In Bonanno, G., van der Hoek, W., and Wooldridge, M., editors, Logic and the Foundation of Game and Decision Theory (LOFT7), volume 3 of Texts in Logic and Games, pages Amsterdam University Press. Board, O. (2004). Dynamic Interactive Epistemology. Games and Economic Behavior, 49: Bonanno, G. and Nehring, K. (1997). Agreeing to disagree: a survey. Some of the material in this paper was published in [Bonanno and Nehring, 1999]. Bonanno, G. and Nehring, K. (1999). How to make sense of the common prior assumption under incomplete information. International Journal of Game Theory, 28(03): Cave, J. A. K. (1983). Learning to agree. Economics Letters, 12(2): Geanakoplos, J. and Polemarchakis, H. M. (1982). We can t disagree forever. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 19

55 References Cowles Foundation Discussion Papers 639, Cowles Foundation, Yale University. Milgrom, P. and Stokey, N. (1982). Information, Trade, and Common Knowledge. Journal of Economic Theory, 26: Parikh, R. and Krasucki, P. (1990). Communication, consensus, and knowledge. Journal of Economic Theory, 52(1): Savage, L. (1954). The Foundations of Statistics. Dover Publications, Inc., New York. van Benthem, J. Logical dynamics of information and interaction. Manuscript. Olivier Roy & Cedric Dégremont: Agreement Theorems & Dynamic-Epistemic Logic, 19