[Humanist] 23.390 MacKay's three questions

Humanist Discussion Group willard.mccarty at mccarty.org.uk
Fri Oct 23 09:32:11 CEST 2009

                 Humanist Discussion Group, Vol. 23, No. 390.
         Centre for Computing in the Humanities, King's College London
                Submit to: humanist at lists.digitalhumanities.org

        Date: Fri, 23 Oct 2009 08:27:52 +0100
        From: Willard McCarty <willard.mccarty at mccarty.org.uk>
        Subject: MacKay's three questions

One of the most gifted writers among those involved with computing in 
the early years was Donald MacCrimmon MacKay, a physicist who taught at 
King's College London for a time and then went off to found the 
Department of Communication and Neuroscience at Keele. He gave the 1986 
Gifford Lectures at Glasgow. These are summarized at 
www.asa3.org/ASA/PSCF/1992/PSCF3-92Thorson.html and were published as 
Behind the Eye, edited by Valerie MacKay, Basil Blackwell Ltd., Oxford, 
1991. The book is a treasure.

After spending half the book on the neurophysiology of vision, MacKay 
turns to computing explicitly. He notes that in comparing the brain with 
machines, specifically artificial agents, one can ask three sorts of 
questions (pp. 164-7):

1. How closely the brain resembles existing mechanisms we have built for 
whatever purposes. He finds doing this -- putting the brain on a 
Procrustean bed to stretch or hack it to fit -- quite boring (though I'd 
say it's interesting that people do this -- and they do do this quite a 

2. How far can we go to design a machine to fit the functioning of the 
brain and body? Here we turn to the digital computer, he notes. We get 
an answer that is at first quite exciting, then turns rather dull, 
orbiting Turing's thesis: "it seems to promise unlimited scope for the 
powers of digital symbol processing machines. When you think twice about 
it, though, it's a bit dull", he comments, as you find yourself in an 
unending loop adding ever more bits to the logical process to take care 
of this or that circumstance. And meanwhile, he notes, "the really 
interesting things about human behaviour -- inventiveness; falling in 
love; sensing whether someone is sincere, prejudiced, fair-minded; or 
making judgements whether they are well-informed, generous or wise" -- 
all these run ahead of whatever test we have, likely always in principle 
to escape any test we might write.

3. How far can we go in writing rules for the principles of organization 
of human behaviour? This is the question he finds most interesting, the 
question of theoretical psychology or artificial intelligence. "What is 
aimed at here", he writes, "is not mere external imitation of behaviour 
but internal imitation or replication of the principles on which human 
behaviour is organized." This is an empirical question, which must 
proceed in the sciences "by trying to build hypothetical templates."

All of the above, as noted, are comparisons. That much seems 
inescapable. The second we call "modelling", in which one starts with 
behaviour (of a person, of an artefact) and tries to get close to it, 
learning meanwhile from the failures to do that. The third, I would 
suppose, is best called "simulation", in which one starts with a 
procedural abstraction encoded into a heuristic machine of some sort and 
lets it do what it will (to some degree unpredictably) do. Fair?

It seems to me that we (or at least I) have been stuck at the second 
question for quite a long time, at best. Not surprising, I suppose, 
given that the artefact is our thing, our idol. But what if we took the 
processes of dealing with an artefact, such as a text, devised a 
template from these processes and let it go. What might we learn from this?



Willard McCarty, Professor of Humanities Computing,
King's College London, staff.cch.kcl.ac.uk/~wmccarty/;
Editor, Humanist, www.digitalhumanities.org/humanist;
Interdisciplinary Science Reviews, www.isr-journal.org.

More information about the Humanist mailing list