Invited Lectures

Engineering of logics for the content-based representation of information

Franz Baader - Theoretical Computer Science, TU Dresden

Abstract:

Storage and transfer of information as well as interfaces for accessing this information have undergone a remarkable evolution. Nevertheless, information systems are still not "intelligent" in the sense that they "understand" the information they store, manipulate, and present to their users. The content-based representation of information, which tries to overcome this deficit, requires representation formalisms with a well-defined formal semantics. This semantics can elegantly be provided by a translation into an appropriate logic or the use of a logic-based formalism in the first place. However, in this setting there is a fundamental tradeoff between the expressivity of the representation formalism and the efficiency of reasoning with this formalism.

This motivates the "engineering of logics", i.e., the design of logical formalisms that are tailored to specific representation tasks. The talk will illustrate this approach with the example of so-called Description Logics and their application for databases and as ontology languages for the semantic web.

Formal Methods in Robotics

Bernhard Nebel - Albert-Ludwigs-Universität Freiburg

Abstract:

AI research in robotics started out with the hypothesis that logical modelling and reasoning plays a key role. This assumption was seriously questioned by behaviour-based and "Nouvelle AI" approaches. The credo by this school of thinking is that explicit modelling of the environment and reasoning about it is too brittle and computationally too expensive. Instead a purely reactive approach is favoured.
With the increase of computing power we have seen over the last two decades, the argument about the computational costs is not really convincing any more. Furthermore, also the brittleness argument ceases to be convincing, once we start to incorporate probabilities and utilities. I will argue that it is indeed feasible to use computation intensive approaches based on explicit models of the environments to control a robot -- and achieve competitive performance.
Most of the time one has to go beyond purely logical approaches, though, because it is necessary to be better than an opponent. For this reason, decision theory and game theory become important ingredients. However, purely logical approaches can have its place if we want to guarantee worst-case properties. I will demonstrate these claims using examples from our robotic soccer team, our foosball robot and our simulated rescue agent team.
 

Representing and Reasoning with Preferences

Francesca Rossi - Department of Mathematics, University of Padova

Abstract:

Many problems in AI require us to represent and reason about preferences. You may, for example, prefer to schedule all your meetings after 10am. Or you may prefer to buy a faster computer than one with a larger disk. In this talk, I will describe various formalisms proposed for representing preferences. I will discuss how we can reason about preferences, possibly in the presence of hard and soft constraints. I will also consider how we can aggregate the preferences of multiple agents and show, using an extension of Arrow's theorem, why this can never be done fairly.

[Joint work with Toby Walsh and Brent Venable]