Language, Communication and Rational Agency

Schedule with Abstracts

Saturday, May 30 2009

 9:00 - 9:20:  Coffee  
 9:20-9:30:  Opening  
  Morning session: Agency and Logic  
       9:30 - 10:10        Thomas Icard (with Eric Pacuit and Yoav Shoham)
A Dynamic Logic of Intention and Belief
       10:10 - 10:50        Eric Pacuit
Reasoning with protocols
       11:10 - 11:50        Bill MacCartney
An extended model of natural logic
Abstract We propose a model of natural language inference which identifies valid inferences by their lexical and syntactic features, without full semantic interpretation. We extend past work in _natural logic_, which has focused on semantic containment and monotonicity, by incorporating both semantic exclusion and implicativity. Our model decomposes an inference problem into a sequence of atomic edits linking premise to hypothesis; predicts a lexical semantic relation for each edit; propagates these relations upward through a semantic composition tree according to properties of intermediate nodes; and joins the resulting semantic relations across the edit sequence. A computational implementation of the model achieves 70% accuracy and 89% precision on the FraCaS test suite. Moreover, including this model as a component in an existing system yields significant performance gains on the Recognizing Textual Entailment challenge.

Back to schedule without abstracts
       11:50 - 12:30        Lucas Champollion
Davidsonian events and thematic roles -- are they necessary? A reply to Kratzer and Schein
AbstractWhat do verbs mean? Do they denote relations between individuals (Montague, 1970)? Relations between individuals and events (Davidson, 1967)? Events that are connected to individuals by thematic roles (Parsons, 1990)? Schein (1993) and Kratzer (forthcoming) argue for the presence of both events and thematic roles in logical representations, claiming that they are crucially necessary to represent the meaning of sentences like the following:

(1) Three copy editors caught every mistake.

This sentence has a reading in which "every X" stands in a cumulative relation to another quantifier. The standard interpretation of "every X" as a generalized quantifier only derives distributive readings. So it is not compatible with cumulative interpretations.

Following proposals by Szabolcsi (1997), Beghelli (1997), Matthewson (2001), and others, I argue that "every X" has two components: one denoting the sum of all individuals satisfying X, and another one that universally quantifies over the atomic parts of that sum. The first component can interact cumulatively with other quantifiers. The second component can act distributively. VP pluralizing operators as in Beck and Sauerland (2000) can take scope between the two components. I show that this proposal predicts the relevant reading of (1) without making explicit reference to either events or thematic roles. It also predicts that "every" may interact cumulatively with one quantifier and distribute over another one at the same time. Sentences like the following confirm this prediction.

(2) Three video games taught every quarterback two new plays. (Schein, 1993)

In the relevant reading, the plays may vary with the quarterbacks, but there are only three video games in total. Schein used this reading to argue that both events and thematic roles are necessary. Since my representation makes explicit reference to neither, the argument by Schein and Kratzer is a fallacy: it is based on data that does not bear on the question whether events and thematic roles are present in logical representations.

Back to schedule without abstracts
  Lunch (12:30 - 2:00)  
  Afternoon Session: Formal Epistemology  
       14:00 - 14:40        Krista Lawlor
Scoreboard Semantics for Knowledge Claims
Abstract: When Tom says, "I know that building 90 is somewhere close to here", whether his utterance is true or not depends, in part, on the meaning of "knows." What is the meaning of "knows"? And how does the meaning of "knows" help determine whether what Tom says is true or not? An account of the semantics of knowledge claims answers these questions. Recently, philosophers have produced a number of competing accounts of the proper semantics for knowledge claims. In this talk, drawing on some suggestions of David Lewis' and extending an earlier account of my own, I'll sketch a semantics for knowledge claims that I'll call Scoreboard Semantics. I'll try to show how Scoreboard Semantics does a better job than its competitors in handling key linguistic data, specifically data involving challenges and retractions of knowledge claims. (Background reading: of interest are two papers by David Lewis: "Scorekeeping in a Language Game", "Elusive Knowledge"; and my "Enough is Enough: Pretense, Invariance and the Semantics of 'knows that'.")

Back to schedule without abstracts
       14:40 - 15:30        Sheri Rousch
       15:50 - 16:50        Kevin Kelly
A Learning Semantics for Epistemic Modal Logic
Abstract Logic seeks general, algebraic structures in reasoning. Alas, it tends to find what it seeks. Thus, modal logics of knowledge "find" closure under known entailment, perfect knowledge of tautologies, and a blanket entitlement to knowledge that one knows. Outside of modal logic, knowledge is an encomium reserved for the culmination of diligent pursuit of the truth. So here is a straightforward idea: instead of studying static, ideal, accessibility and topological relations one doesn't really know how to interpret in applications, why not model the logical knowledge operator concretely in terms of learning agents who have actually learned the truth and would have learned the truth efficiently had it been otherwise? Then the usual, embarrassing logical omniscience principles disappear and are replaced by plausible logical principles linking knowledge with having learned and logical closure and KK with the diligent employment of inference tickets of appropriate kinds. The proposed semantics links logic, computability, formal epistemology, game theory and the semantics of verbal aspect, so it particularly germane to this workshop.

Back to schedule without abstracts
       16:50 - 17:20        Discussion
  Drinks & Dinner Reception at CSLI (starting at 17:30)  

Sunday, May 31, 2009

 9:15 - 9:30:  Coffee  
  Morning session 1: Logic and Games 
       9:30 - 10:20        Giacomo Bonanno
Revealed preference, iterated belief revision and dynamic games
Abstract We first establish a correspondence between the structures studied in revealed preference theory and the AGM theory of one-shot belief revision. Then we extend these structures to model iterated belief revision and discuss some general principles. Finally we focus on belief formation in games and study the implications of a plausible theory of iterated revision (based on principles proposed by Darwiche and Pearl, Nayak et al and others) on solution concepts for extensive-form games.

One third of the talk is based on a paper which is forthcoming in Artificial Intelligence and the rest is work in progress.

Back to schedule without abstracts
       10:20 - 11:00        Yossi Feinberg
Games with Unawarenes
Abstract We provide a tool to model and solve strategic situations where players reasoning is limited, in the sense that they may only be aware of, or model, some of the aspects of the strategic situations at hand, as well as situations where players realize that other players' perceptions may be limited. We define normal, repeated, incomplete information, and dynamic (extensive) form games with unawareness using a unified methodology. A game with unawareness is defined as a collection of standard games (of the corresponding form). The collection specifies how each player views the game, how she views the other players' perceptions of the game and so on. The modeler's description of perceptions, the players' description of other players' reasoning, etc. are shown to have consistent representations. We extend solution concepts such as rationalizability and Nash equilibrium to these games and study their properties. It is shown that while unawareness in normal form games can be mapped to incomplete information games, the extended Nash equilibrium is not mapped to a known solution concept in incomplete information games, implying that games with unawareness generate novel types of behavior.

You can fin the paper here

Back to schedule without abstracts
  Morning session 2: Information Dynamics  
       11:10 - 11:50        Johan van Benthem
Logics of Information Change
       11:50 - 12:30        Brian Skyrms

  Lunch (12:30 - 2:00)  
  Afternoon Session: Language and Cognition  
       2:00 - 2:40        Adrian Brasoveanu
Exceptional Scope as Scopal Independence
The paper proposes a novel solution to the problem of scope posed by natural language indefinites that captures both the difference in scopal freedom between indefinites and quantifiers -- indefinites have free upwards scope, disregarding not only clausal but also island boundaries -- and the fact that the scopal freedom of indefinites is nonetheless syntactically constrained. Following the main insight of independence-friendly logic, the special scopal properties of indefinites are attributed to the fact that their semantics can be stated in terms of choosing a suitable witness. This is in contrast to "bona fide" quantifiers, the semantics of which crucially involves relations between sets of entities. The syntactic constraints on the interpretation of indefinites follow from the fact that witness choice arises as a natural consequence of the process of (syntax-based) compositional interpretation of sentences and it is not encapsulated into the lexical meaning of indefinites, as choice / Skolem function approaches to exceptional scope would have it. (Handout)

Back to schedule without abstracts
       2:50 - 3:50        Jeroen Groenendijk
Inquisitive Semantics