JOINING THE DOTS: EXTENDING THE AUTOCATALYTIC PICTURE
TO A CREDIBLE EVOLUTIONARY PROCESS
Commentary on Gabora on Origin-Culture
Bruce Edmonds
Centre for Policy Modelling
Manchester Metropolitan University,
Aytoun Bldg., Aytoun St.
Manchester M1 3GH, UK.
http://www.cpm.mmu.ac.uk/~bruce
b.edmonds who-is-at mmu.ac.uk
ABSTRACT: Gabora's picture of the kick-starting of culture is
reinterpreted and extended in order to make the subsequent internal
evolutionary process credible. This is done by changing the unit of
replication from a concept to a path consisting of a series of
concepts, where each concept in the path has a tendency to cause
the next to be recalled into the stream of thought. Mechanisms of
selection and variation are suggested. The extended picture is
shown to explain how an autocatalytic set of concepts might become
a `world-view' in a more meaningful sense.
KEYWORDS: evolution, meme, path, association, autocatalytic,
world-view, selection, variation
1. In her target article on the origin of culture, Gabora (1998),
presented an innovative account of the co-emergence of world-views and
streams of thought. Regardless of whether her account turns out to be
correct, it is important because it posits credible and specific
mechanisms whereby human culture may have come about. It simultaneously
opens up the possibility of other computational models and provides a
reference point for them. In this commentary I would like to interpret
and extend Gabora's account so as to substantiate the claim that the
resulting process could have an evolutionary nature.
2. Gabora's scenario shows how a sufficient `density' of concepts in an
associative memory structure (e.g. one with the structure of the Sparse
Distributed Memory model) may allow the emergence of an autocatalytic
process consisting of a self-triggering stream of thought. She suggests
that an ability to abstract may have been a critical factor in crossing
this density threshold. Once such a stream is established it then
shapes and is shaped by the `world view' which is composed of the
network of the concepts residing in an individual's memory.
3. The mutual interaction between the stream of thought and this
network is characterised as an internal evolutionary process: the units
are the concepts (which correspond to the points in an SDM
architecture); the fitness of these is the frequency of recollection
into the stream of thought (which is strongly correlated with its
entrenchment in memory); replication is provided by the "correlation
between consecutive memes; "selection is via "associations, drives,
social pressures, affordances and limitations of the environment," and
variation is a result of "sensory novelty, blending, expressive
constraints." This process in turn allows an evolutionary process
between individuals to develop. The internal and external processes are
somewhat conflated in Gabora's account; here I will concentrate only on
the internal process.
4. The characterisation of the internal process as evolutionary does
not quite work since there is no reason why a successful concept (i.e.
one that occurs in a stream of thought) should be more likely to be
reselected via the replication mechanism suggested (the correlation
between this concept and the next). In other words, there would be no
correlation between this fitness and the probability of its
replication. Furthermore, it is unclear exactly how the proposed
mechanisms of variation will act on the concepts.
5. However, if one focuses upon paths consisting of a sequence of
concepts, such that each item in the sequence is likely to invoke the
next, then one can build a more obviously evolutionary picture. In
contrast to a concept, a path has a mechanism by which, if it is
selected as the contents of a stream of thought, it can increase the
chance that it will be reselected - it can form a loop. It can do this
either on its own or in conjunction with other paths. In this
interpretation, each such path would be a meme - the unit upon which
this evolutionary process acted.
6. The set of all such paths would be the memetic population. The
fitness of each path would be the frequency with which it was recalled
(and hence how entrenched it was in memory). There are several methods
by which a path could encourage its own re-selection. First, it could
form a loop so that it directly caused its own recall. Second, it could
form a loop in conjunction with other paths so that its own selection
caused the selection of other paths which had a propensity to cause it
to be reselected. Third, it could be a path that tended to be recalled
with certain stimuli from outside the organism, given that those
stimuli where themselves frequent. These three possibilities are
illustrated in figure 1 below.
ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/Pictures/em1.html
FIGURE 1. SOME WAYS IN WHICH A PATH MAY CAUSE ITS OWN RECALL.
7. Variation could be introduced in a number of ways. First, there
could be a simple addition of new concepts, so that an extra `point' is
added to the path. Second, a `short-cut' could be discovered, cutting
out a point. Third, there could be something similar to genetic
crossover. This could occur when two paths pass through the same small
locality, allowing a `derailment' from one path onto the other. Now the
first section of the first path would lead on to the second part of the
second path. These are illustrated in figure 2.
ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/Pictures/em2.html
FIGURE 2. SOME WAYS IN WHICH VARIATION COULD BE INTRODUCED.
8. In this way we have all the requisites for an evolutionary process
(Calvin 1997), but one where we would expect paths to be evolved rather
than individual concepts. In all other respects it is compatible with
the picture painted by Gabora.
9. The extended picture I have presented seems to have strong parallels
with Adam's (1998) "synaptic Darwinism." They both posit an
evolutionary process occurring to populations of pathways. However, his
is at the neuronal level and is a general mechanism of learning which
(if it turned out to be correct) would underlie learning in animals as
well as humans. My picture represents a further evolutionary process
between pathways of concepts which may (or may not) be implemented by a
process of synaptic Darwinism.
10. My evolutionary picture also throws light on the question of why
the mutual interaction of a dense set of concepts and a continuous
stream of thought might result in a world-view. A world-view implies
something more than just an autocatalytic set; for example, it is
associated with pervasiveness and coherence. Pervasiveness is the
property that whatever stimuli are encountered by the individual, there
will be paths that will be triggered by it; there is no inherent reason
why this would be true of an autocatalytic set. Coherence is the
property that there will not be different sections of the population of
paths that tend to be mutually exclusive; in Gabora's picture it is
possible for two or more separate autocatalytic sets to emerge
(although the probability of this could be a function of the
connectiveness of the concepts). However, given the picture of evolving
paths, one can see that if there were stimuli that did not trigger
paths then one might expect paths to evolve to exploit this fact; and
if there were separate and mutually exclusive parts of the network of
paths, then one might expect one set eventually to win out over the
other. In this way, once an autocatalytic process has occurred so as to
stimulate a continuous stream of thought, an evolutionary process
acting on paths will tend to develop this into a pervasive and coherent
system -- one that might be meaningfully called a `world-view'.
11. Of course, the above picture is simplistic. There are many other
organisational forces at work, including: a continuing autocatalytic
process; a process of endorsing paths as to their adaptive value (i.e.
learning about the environment); inter-individual memetic processes;
the effect of the emotional and chemical state of the brain; and
possibly meta-devices such as a mechanism of boredom to prevent small
loops dominating.
12. Gabora's target article has showed us how functional models of the
emergence of culture are possible, and even profitable to consider.
This commentary has attempted to follow from her lead.
REFERENCES
Calvin, W. H. (1997). The Six Essentials? Minimal Requirements for the
Darwinian Bootstrapping of Quality. Journal of Memetics - Evolutionary
Models of Information Transmission, 1.
http://www.cpm.mmu.ac.uk/jom-emit/1997/vol1/calvin_wh.html
Gabora, L. (1998). Autocatalytic Closure in a Cognitive System.
PSYCOLOQUY 9(67).
ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/
psyc.98.9.67.origin-culture.1.gabora
http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?9.67
Adams, P. (1998). Hebb and Darwin. Journal of Theoretical Biology,
195:419-438.