What to do Next?
Themamiddag "What to do next: Rational and Social Choice"
Vrijdag 31 oktober 2003. Jaarbeurs Utrecht. Programma begint om 12.30
met een broodjeslunch, u aangeboden door de VvL. Eerste lezing om
13.30. Afsluiting om 17.00 met een borrel. Speciale gast: professor
Rohit Parikh uit New York. Co-sponsor: Onderzoekschool SIKS.
- Robert van Rooij,
Cooperative information exchange:
Exhaustive interpretation and the division of pragmatic labour
In this talk I will discuss whether the principle of exhaustive
interpretation of answers to questions can be derived from more general
(Gricean) principles of rational information exchange in cooperative
dialogues, or whether exhaustive interpretation is rule-governed and
conventional in nature.
- Paul Harrenstein, Game-theoretical Consequence
Logical notions of consequence have frequently been related to
game-theoretical solution concepts.
The correspondence between a formula being classically valid and the
existence of a winning strategy for a player in a related game, has been
most prominent in this context.
We, however, propose a conservative extension of the classical notion of
consequence based on a generalization of the renowned concept of a
Nash-equilibrium. This game-theoretical concept of consequence opens up
a line of theoretical research in which logic, game theory and social
choice theory interact at the same level. We trust this research will
prove to be fruitful in many areas of Artificial Intelligence and the
development of multi-agent systems.
- Mehdi Dastani,
Programming Agent Deliberation
- An important aspect of agent autonomy is the decision making capability
of the agents. Different models of decision making have been proposed,
rooted in different research traditions and objectives. Each proposal
explains the decision making behaviour in terms of different sets of
underlying concepts, e.g. quantitative concepts such as probability and
utility function or qualitative concepts such as belief, goals, and
intentions. I will discuss the decision making behaviour of cognitive
agents from the programming point of view and in terms of agent
deliberation process. Several issues that agents need to deliberate
about to decide which action to perform next will be considered. A
programming language is proposed to implement agent deliberation process.
- Rohit Parikh, What Next in Social Software?
Informal tools for social software have existed for some time, in
Implementation theory, Game Logic, Coalition logic, etc.
Wittgenstein's language games, an idea very far ahead of its time are
an even earlier precursor. However, it seems now that many different
strains are coming together, in part because there are people
acquainted with more than one field who can cross-fertilize.
An important issue is knowledge. We have formal theories of
knowledge, and everyone subscribes to common knowledge as an
important idea, but how relevant is it? How firm are its intuitive
foundations? Chwe in his Rational Ritual subscribes to common
knowledge, but the notion he actually needs is quite a bit weaker.
Fukuyama speaks about trust as an important ingredient in economic
development, and trust is a good substitute for common knowledge in
operational situations. Perhaps these constitute some sort of answer
to the skepticism of Clark and Marshall (and others).
At a more local level, the formal theories of knowledge seem more
useful. There is insufficient entry of knowledge in Deontic logic,
but clearly one can only be responsible when one knows the
circumstances to the relevant extent. A doctor cannot be expected to
treat someone unless he knows that the person is sick. But a
hospital is obligated to find out that a patient has fever. Sometimes
one has an obligation to acquire knowledge, and sometimes not. Can
we have formal theories to get better insight?
Social software exists in two (or more) forms. One is where society
or some larger organization provides us with sub-routines which we
can use in our personal software. Libraries or transportation
systems are a good example. One does not need other people to use a
library. In other situations the software is itself co-ordinated, as
for instance in an orchestra.
There is a rich field to develop and we will give some examples and
some technical material.
Jan van Eijck
Last modified: Fri Feb 13 11:56:40 CET 2004