Thoughts on computationalism

by Neil Rickert

Recently, Massimo Pigliucci hosted a discussion of the computational theory of mind on his Rational Speaking podcast, with an accompanying post on his blog:

That blog post has a link to the podcast.  I listened to that podcast this morning, and will comment on it in this post.

I have been clear that I am skeptical of computationalism.  And Pigliucci is equally clear that he, too, is a skeptic.  But I don’t plan to repeat those earlier posts here.

Analog computation

What surprised me about the discussion, was that O’Brien emphasized analog computation.  Perhaps O’Brien is conceding that there might be problems with computationalism in the form of digital computation.

I remember, perhaps around 15 years ago, somebody argued for analog computation rather than digital computation.  This was in a usenet post, and possibly the poster was Stefan Harnad.  I remember, at the time, that my response was something like:

Anything analog can do, digital can do better;

Digital can do anything better than that;

No, it can’t!

Yes it can!

No, it can’t!

Yes it can, yes it can yes it can.

I think you get the picture there, and you probably see the parallel to a well known Broadway song.

Experience has shown that digital computation can always outperform analog computation.  This is because analog computation is limited in precision, while we can always further increase the precision of digital computation.

O’Brien seems to see a problem with digital computation, in that it is done with meaningless symbols.  By contrast, he sees analog computation as dependent on use of meaningful analogies.  So he apparently hopes to solve the semantic problem that way.

I see this as a mistake.  O’Brien might have a point about the semantics.  But in that case, he should not be calling this computation.  The essence of computation, is that it is abstract mechanistic rule following.  If you are doing something that depends on semantics, you are not doing computation.  Of course, we do use computation to solve meaningful problems.  But the meaning comes from outside the computation.  It does not come from the computation itself.

Representation and simulation

O’Brien also described his ideas as involving representation.  I might perhaps describe that as representation and simulation.  Personally, I do hold the view that thinking is something like representation and simulation.  It’s just that I don’t see that it uses much computation, except where we are consciously computing.

In his discussion, O’Brien mentions the idea of pancomputationalism.  That’s the view that everything is computation.  The trouble with that, is that if everything is computation, then the word “computation” loses its meaning as being about very specific types of activity.  If everything is computation, then it does not tell us anything to say that the mind works by computation.  We might just as well limit ourselves to saying “the mind works”, for with pan computationalism, appending “by computation” does not tell us anything.


O’Brien talks a little about information.  He seems to be dissatisfied with the Shannon account of information.  He is looking for what some might call “semantic information”.  That is to say, he wants to see information as something meaningful, that can then be encoded as Shannon information.  But the encoded meaning is paramount.

If you had asked me, say 20 years ago, I might have said something similar.  I have since come to realize that as a misdirection.  The problem is that this alleged semantic information is too elusive.  I doubt that we will ever find it.  As best I can tell, there is no such thing.  We use information in meaningful ways.  But the meaning always comes from us.  The meaning is not carried by the information.  Perhaps I’ll need a separate post on that topic.


To the extent that O’Brien is looking to representation and simulation, he is on the right track.  What I see as mistaken, is the view that he is taking of computation.  And perhaps another mistake is that he expects the semantics to be carried by the representations and simulations, rather than having the semantics come from elsewhere.  And if I am right about that, then he is not getting to the central issues of mind, which have to do with the question “where is that elsewhere, from which the meaning flows?”

%d bloggers like this: