Provided by Cognitive Sciences ePrint Archive
Behaviour-based Knowledge Systems:
An Epigenetic Path from Behaviour to Knowledge
Carlos Gershenson
School of Cognitive and Computer Sciences
University of Sussex
Brighton, BN1 9QN, U. K.
http://www.cogs.sussex.ac.uk/users/carlos
Abstract
In this paper we expose the theoretical
background underlying our current research. This
consists in the development of behaviour-based
knowledge systems, for closing the gaps between
behaviour-based and knowledge-based systems,
and also between the understandings of the
phenomena they model. We expose the
requirements and stages for developing
behaviour-based knowledge systems and discuss
their limits. We believe that these are necessary
conditions for the development of higher order
cognitive capacities, in artificial and natural
cognitive systems.
1. Introduction
In the field of artificial intelligence, knowledge-based
systems (KBS) and behaviour-based systems (BBS) have
modelled and simulated exhibitions of intelligence of
different types, which we could call “cognitive”
intelligence and “adaptive” intelligence, respectively.
Broadly, and independently oftheir methodologies, BBS
have modelled intelligenceexhibitedbyanimals adapting
to their environment, while KBS have modelled “higher”
cognition: reasoning, planning, and problem solving.
Trying to understand how this higher cognition could
evolve from adaptive behaviour, we propose the
development of behaviour-based knowledge systems
(BBKS). They are systems where an artificial creature is
able to abstract and develop through its behaviour
knowledge from its environment, and exploit this
knowledge for having a favourable performance in its
environment. BBKS relate the exhibitions of intelligence
modelled by BBS and KBS, closing the gaps between
them.
In order to develop these ideas, in the next section we
expose abstraction levels (Gershenson, 2002a) in animal
behaviour, which are useful for illustrating our goals. In
Section 3 we present the steps we believe should be
followed in order to develop and exhibit knowledge
parting from adaptive behaviour. In Section 4 we note
limits of BBKS, which are related to the limits of
Epigenetic Robotics and Artificial Intelligence. We also
briefly describe our current work, which consists in the
implementation of a BBKS.
2. Abstraction Levels in Animal
Behaviour
Abstraction levels (Gershenson, 2002a) represent
simplicities and regularities in nature. Phenomena are
easier to represent in our minds when they are simple.
We can have an almost clear concept of them, and then
we can try to understand complex phenomena in terms
of our simple representations. We can recognize
abstraction levels in atoms, molecules, cells, organisms,
societies, ecosystems, planets, planetary systems, galaxies.
An element of an abstraction level has a simple and
regular behaviour, and it is because of this that can be
easily observed and described. At least easier than the
complexities that emerge from the interactions of several
elements.
We can identify abstraction levels in animal behaviour
(Gershenson, 2001, pp. 2-3), taking the definition of
behaviour developed by Maturana and Varela:
“behaviour is a description an observer makes of the
changes in a system with respect to an environment with
which the system interacts” (Maturana and Varela, 1987,
p. 163). Our proposal is not a final categorization, but it
is quite convenient for orienting our work, even when the
borders between levels are fuzzy. The most elemental
type of behaviour is vegetative, which can be seen as
behaviours “by default” (such as breathing, metabolism,
etc.). We can also distinguish reflex behaviours. These
are action-response-based behaviours (such as reactions
to pain). Stepping-up in complexity, we can identify