Connectionism, Analogicity and Mental Content



Connectionism, Analogicity and Mental Content

coherent fashion. Without these rules, a syntactic engine is utterly blind; with them, it behaves as
if it were a semantic engine, and in so doing is capable of computational work.

So far, then, so good. However, what is not so clear, reading from the literature at least, is
what happens if we decide that Turing’s way is not the mind’s way; that cognition is not, or not
predominantly, rule-governed symbol manipulation. It is here that the failure to distinguish
between the two ideas embodied in Turing’s work is responsible for a good deal of confusion.
For when some theorists in cognitive science come to reject classicism, as many have been doing
recently, there is a tendency for them to reject too much. When they ask “What might cognition
be, if not [classical] computation?” (Van Gelder, 1995), their answer tends towards a radical, non-
representationalist form of dynamical systems theory which discards the whole idea of
semantically coherent processing, and with it the only idea we have about how intelligent
behaviour could arise in the physical world. Thus it’s important to realise that the theoretical
landscape is more extensive than their thinking implies; that one can reject classicism without
rejecting a computational conception of cognition.

4. An Alternative Way: Analog Computation and Connectionism

Everyone knows that computer science routinely distinguishes between digital and analog
computers. So at first glance it would seem that, in analog computation, there is indeed an
alternative means of mechanising computation. But what are analog computers and how do they
differ from digital devices? The standard account of the distinction between digital and analog
computers in the literature centres on the representational vehicles a computational device
deploys. If the physical material that implements each of these vehicles can vary to some extent
without thereby affecting its representational content, the device is said to employ a
discrete
representational medium, and hence is digital. If, on the other hand, any variation in this
physical material constitutes a variation in representational content, the medium is
continuous
and the device is thought to be analog.3

As it stands, however, this way of drawing the distinction is not unproblematic, largely
because it is sometimes unclear whether a particular representational medium is more discrete
or continuous. The legitimacy of the distinction has thus been called into question, with
considered opinion varying from those who think the distinction is merely relative to our
explanatory purposes and interests, to those who argue that the (ultimately discrete) nature of
our world dictates that the class of analog computers is empty. Furthermore, even if this way of
drawing the distinction can be made to work, it isn’t terribly illuminating. For what we really
want to know, and what the standard story doesn’t tell us, is how the
causal operation of analog
computers differs from their digital counterparts.

Fortunately, there is such an account of the distinction between digital and analog
computers in the offing; one, moreover, that because it focuses on causal operation is able to
explain why digital devices deploy discrete representational media while analog machines tend
towards more continuous forms of information coding. The key here, not surprisingly, is
semantically coherent behaviour and how it is achieved. We’ve already seen how digital
computers do it: they have semantic coherence imposed on them by a set of syntactically
applied, symbol transformation rules. Analog devices, as we are about to see, do it differently.

3 One encounters different versions of this distinction digital and analog computers in the literature. For example,
some computer scientists talk in terms of the difference between modelling a physical process with real numbers
(analog computation) and with rational number approximations (digital computation). Others invoke a distinction
between discrete symbol manipulation (digital) and the manipulation of real-valued signals (analog). The version in
the text is closer in spirit to the way this distinction is developed in philosophy, mainly by Nelson Goodman (1969,
chp.4) and David Lewis (1971).



More intriguing information

1. The Role of State Trading Enterprises and Their Impact on Agricultural Development and Economic Growth in Developing Countries
2. Proceedings of the Fourth International Workshop on Epigenetic Robotics
3. How we might be able to understand the brain
4. Implementation of the Ordinal Shapley Value for a three-agent economy
5. Beyond Networks? A brief response to ‘Which networks matter in education governance?’
6. Modelling the health related benefits of environmental policies - a CGE analysis for the eu countries with gem-e3
7. Food Prices and Overweight Patterns in Italy
8. Public infrastructure capital, scale economies and returns to variety
9. Fighting windmills? EU industrial interests and global climate negotiations
10. BODY LANGUAGE IS OF PARTICULAR IMPORTANCE IN LARGE GROUPS
11. The name is absent
12. Who’s afraid of critical race theory in education? a reply to Mike Cole’s ‘The color-line and the class struggle’
13. DISCRIMINATORY APPROACH TO AUDITORY STIMULI IN GUINEA FOWL (NUMIDA MELEAGRIS) AFTER HYPERSTRIATAL∕HIPPOCAMP- AL BRAIN DAMAGE
14. New issues in Indian macro policy.
15. Before and After the Hartz Reforms: The Performance of Active Labour Market Policy in Germany
16. The Provisions on Geographical Indications in the TRIPS Agreement
17. The name is absent
18. Auction Design without Commitment
19. Ein pragmatisierter Kalkul des naturlichen Schlieβens nebst Metatheorie
20. The name is absent