Abstract for AISB 2000: How to Design a Functioning Mind
AUTHOR: Matthias Scheutz
Department of Computer Science and Engineering
University of Notre Dame
TITLE:Behavioral States: Linking Functional and Physical
Descriptions
ABSTRACT:
Today's most commonly held view on the "mind-body" problem is that
mental states are functional states, which somehow "supervene" on
physical states. An approach to understanding mind that attempts to
relate functional states directly to physical states will in my view,
however, not very likely reveal any interesting results about the
implementing system, since the level of functional states is too high a
level of description. For the same reason it does not seem very helpful
in suggesting ways to build a mind (not to mention all the problems
connected with the involved notion of "implementation" or "realization"
that seem to be largely ignored by the philosophical community). To be
of any practical importance in designing a mind at all, a level of
description has to retain at least some of the relevant physical
properties. I will suggest such an intermediary level, which I call the
level of behavioral states.
A behavioral state is a state an individual is in if it performs a
particular behavior (e.g., such as wincing or looking out for prey).
Note that behavioral states are not restricted to "motor actions", but
include sensory actions as well as more abstract proprioceptive and
reflective actions (such as monitoring inner physiological states or
generating images, producing plans, etc.). Behavioral states are
organized in hierarchies in the sense of the behavioral descriptions
advanced by the classical ethologists (Lorenz and Leyhausen). Within
each level there are clusters of mutually exclusive behavioral states,
which inhibit each other to various degrees, while entertaining
excitatory connections to lower and upper level states (and possibly to
some behavioral states of other clusters at the same level as well).
The hierarchy reflects in part the complexity of each behavior
associated with these states, the lowest level corresponding to simple
reflexes (note that behavior-based robotics has explored this level of
behavioral states in great detail). Behavioral states are temporally
extended as timing plays a crucial role in their interplay. Some of
them are directly connected to sensory or motor organs, others are
indirectly connected via so-called "information channels" (i.e.,
filtering mechanisms that select parts of one or more sensory inputs or
motor outputs), again others might only be connected to other behavioral
states. At any given time, individuals are in multiple behavioral
states and their state space determines the individual's behavior. Note
that these states are not limited to observable behavior.
The behavioral level of description is intermediary, because it
specifies states that could be realized in many different physical ways
(in neural architectures, but possibly also in digital ones, and
others), yet retains at least one crucial physical and causal aspect not
retained in functional states: time! By explicitly incorporating time
and thus allowing for modeling the temporally extended interactions
between different states, this level might not only prove useful for
constructing systems that exhibit complex causal interactions, but also
for explaining how functional states are related to physical states by
viewing them as (not necessarily disjoint) collections of behavioral
states.
=======================================================================
SHORT CV
Joint Ph.D. in Cognitive Science and Computer Science
Indiana University (Bloomington, IN, USA), 1999
Ph.D., Philosophy
University of Vienna (Vienna), 1995
Visiting Assistant Professor in the Department of Computer
Science and Engineering, University of Notre Dame
Lecturer in the Department of Philosophy of Science and Social
Studies of Science, University of Vienna
Organizer of the conference and workshop NTCS'99 on "Computationalismus
- The Next Generation" (see http://www.univie.ac.at/cognition/conf/ntcs99/)
Co-founder of the Austrian Society for Cognitive Science
(see http://www.univie.ac.at/cognition/)
Recent Edited Books & Jourals
Scheutz, Matthias (Ed.) (2000). Computationalism: New Directions. MIT
Press: Cambridge, MA (forthcoming).
Scheutz, Matthias (Ed.) (2000) "Computationalism-The Next Generation?"
Special edition of the Conceptus series. (forthcoming)
Recent Refereed Journal Papers
Scheutz, Matthias (2000) "Computationalism Again?". In Special edition
of the Conceptus series (forthcoming)
Scheutz, Matthias (2000) "Cognitive Science-The Status Quo". Journal of
the OEGAI. (forthcoming)
Scheutz, Matthias (1999) "When Physical Systems Realize Functions...".
Minds and Machines 9,2, 161-196.
Recent Contributions to Edited Works
Scheutz, Matthias (2000) "Computationalism's Next Generation". In M.
Scheutz (Ed.) Computationalism: New Directions. MIT Press: Cambridge.
(forthcoming)
Scheutz, Matthias (1999) "The Ontological Status of Representations". In
A. Riegler, M. Peschl & A. von Stein (eds.) Understanding Representation
in the Cognitive Sciences. Plenum Academic / Kluwer Publishers: Holland,
33-38.
Recent Review Articles
Scheutz, Matthias (1999) Book review of The Philosophical Computer by
Patrick Grim, Gary Mar, and Paul St. Denis. Philosophical Psychology 12, 2,
237-323.
Scheutz, Matthias (1999) Book review of Eine Elementare Einfhrung in die
Theorie der Turing-Maschinen by Oswald Wiener, Manuel Bonik und Robert
Hoedike. Eureka.
Recent Conference Proceedings Refereed
Scheutz, Matthias (1999) "A New Computationalism?". In Proceedings of
NTCS'99 (Matthias Scheutz, ed.): University of Vienna, 1-3. Also online at MIT
Cognet (Marney Smith, ed.): http://cognet.mit.edu.
Recent Submissions
Scheutz, Matthias. "Causal Complexity = Computational Complexity?"
(Submitted to Analysis)
Scheutz, Matthias. "The Computational Power of Switch Systems" (To be
submitted to Synthese)
Scheutz, Matthias. " Surviving in a Hostile Multi-Agent Environment: How
Simple Affective States Can Aid in the Competition for Resources "
(Submitted to AI 2000.)
Scheutz, Matthias. "Evolving Neural Robot Controllers for Simple Reflex
Behavior" (To be submitted to IJCNN2000)
Recent Invited Lectures
Truth Models
Notre Dame Logic Colloquium (University of Notre Dame, Spring 2000)
Computationalism and Dynamicism
Austrian Society for Cognitive Science (University of Vienna, January
12, 2000)
What it Means to Implement a Computation
Computer Science Colloquium (University of Notre Dame, December 7, 1999)
Behavior Based Robotics: The Role of Robots in Artificial Intelligence
and Cognitive Science, Colloquium Series of the Department of Mechanical
Engineering (University of Notre Dame, November 30, 1999)
Research interests:
Implementation - The Missing Link (book project on the foundations of
cognitive science and computer science explicating the central,
foundational role that the notions of implementation and physical
realization play in both disciplines)
Behavior Based Robotics (modeling emotional states in simple simulated
as well as real agents, such as the TRILOBOT; combining multimodal sensor
information in ethologically plausible ways; using genetic algorithms
to evolve neural network controllers for autonomous agents)