Abstract for AISB 2000: How to Design a Functioning Mind
AUTHOR: Bruce Edmonds,
Centre for Policy Modelling,
Manchester Metropolitan University
Title: Towards Implementing Free Will
ABSTRACT:
It is argued that real free will develops in the brain during
development. It is shown that this avoids some of the aguments
against the existence of free will. The idea is that the brain
has evolved to be an environment where free will can evolve from
infinitessimal beginnings. This is coherent with a functional
picture of free will which sees free will as the mechanism for
automony composed of two parts: (1) the internal coherency of the
individual in terms of its own descision procedures and (2) the
external unpredictability (in general) of the resulting
decisions. Such a picture provides the possible evolutionary
advantages for the evolution of the necessary features of the
brain to implement this secondary evolution of free-will.
This conception of free-will suggests the outline for a process
whereby one could enable our creations to have free-will in the
sense just described. The following elements are necessary:
(a) a source of randomness (either quantum or pseudo)
(b) a structure that encourages the evolution of coherent (even
logical) decision structures that (i) further the goals of the
agent (ii) encourages (post hoc) coherency with other's decision
making processes (iii) are not predictable to others when this is
advantageous
An implementation of this is then outlined for a collection of
agents. Each agent has a primary decision making process which
learns about appropriate decision making including an essentially
random input. In the co-evolutionary setting of the society of
agents with some elements of competition, this will include
considerable advantages to being unpredictable and complementary
to others. Each agent also has a secondary process which is
developed to model its own as well as others actions - this is a
co-evolutionary social process whereby others are used to aid its
self-modelling as well as its self-model aiding to model others.
This ensures that there is a considerable social agreement on the
post-hoc modelling and hence reporting of its own decision
making.
================================================================
SHORT CV:
Bruce Edmonds is Senior Research Fellow in Logic and Formal Methods at
the Centre for Policy Modelling - an interdisciplinary research group
at the Manchester Metropolitan University. He did his first degree in
Mathematics at Oxford, and his PhD at the University of Manchester in
the Philsophy of Science. He has wide research interests, including:
complexity, evolutionary processes, modelling methofology, cognitive
science and computer languages to support social simulation. He is an
editor of the Journal of Memetics - Evolutionary Models of Information
Transmission, co-chaired a workshop on "socially Situated Intelligence"
and co-edited a special issue of Computational and Mathematical
Organisation Theory on "Social Intelligence". More information about
him and his publications can be found at
http://www.cpm.mmu.ac.uk/~bruce