Loading ...
Sorry, an error occurred while loading the content.

[evol-psych] Re: Defending *massive* modularity

Expand Messages
  • robertbiegler
    ... Depends. Neither the set of inputs nor the set of responses has to be small in principle, though I do concede that your description does apply to the
    Message 1 of 28 , Mar 3, 2006
    • 0 Attachment
      --- In evolutionary-psychology@yahoogroups.com, Joao Sousa <j.d.
      sousa@...> wrote:
      > Let me try: A module is a device that, for a small set of inputs,
      > I1.....In, gives a small set of responses, R1....Rm, being n and m
      > small and the underlying algorithms tractable. Do you agree on this?

      Depends. Neither the set of inputs nor the set of responses has to be
      small in principle, though I do concede that your description does
      apply to the concrete examples of modules that I can think of. For
      the unrealistic counterexample, imagine a module for addition. Both
      input and response sets would be infinite. However, the computation
      is specialised and limited.

      But perhaps I misunderstand. Do you mean by small sets not sets with
      small numbers of elements, but sets with strictly limited types of
      input and output? Then I would agree entirely. So the barn owl's
      azimuth computation takes in phase locked action potentitals which
      transmit timing information and returns an angle, and it does nothing
      else whatsoever.

      > It seems that many things in living organisms are modules, for
      > example, a single neuron likely is.

      For the purposes of theoretical neuroscience, people have looked at
      modules as small as three or four neurons in neural netowrk models,
      and I suppose you could even think of single neurons as modules in
      that context, but I don't think that makes sense for psychology. At
      least in vertebrates, I would bet moderate sums of money that there is
      nothing you can see in behaviour which would depend on a single
      neuron.

      > However, it's impossible to imagine a single module for a thing such
      > as Mate Choice. There are too many different inputs

      Depends again on how you look at it. In the barn owl, you have that
      module for determining the azimuth of a sound source from interaural
      time difference. There is another for determining (at high enough
      frequencies) the elevation from interaural intensity differences (the
      barn owl is one of the species with asymmetric ears). I don't
      remember now whether these feed first into an separate auditory map,
      or directly into the optic tectum for combination with visual
      information. In any case, you have to expect not just a hierarchy,
      but a network of modules. Not necessarily a tidy chain of command,
      but something more like low and high level subroutines. A mate choice
      module calling on a number of lower level modules is, at least in
      principle, possible. You would expect increasing integration and
      generality of inputs and outputs as you go up in the hierarchy, and
      there will come a point where people will disagree whether they still
      call something a module or not. West-Eberhard prefers not talk about
      modules at all, but about modularity, as in degrees of.

      > Then, mate choice cannot be executed by a module, either it is
      > executed by many (possibly hundreds or thousands) of different
      > modules - and here we have massive modularity - or by non-modular
      > adaptations.

      In my opinion, that's a false dichotomy. There are other
      possibilities between those extremes.

      Regards

      Robert Biegler
    • Joao Sousa
      ... Yes, that s what I was meaning. Of course, an input may be an analogic signal, in which case it comes in an infinite number of possible values. ... Ok, but
      Message 2 of 28 , Mar 4, 2006
      • 0 Attachment
        At 20:50 03-03-2006 +0000, you wrote:
        >--- In evolutionary-psychology@yahoogroups.com, Joao Sousa <j.d.
        >sousa@...> wrote:
        > > Let me try: A module is a device that, for a small set of inputs,
        > > I1.....In, gives a small set of responses, R1....Rm, being n and m
        > > small and the underlying algorithms tractable. Do you agree on this?

        Robert Biegler:

        >Depends. Neither the set of inputs nor the set of responses has to be
        >small in principle, though I do concede that your description does
        >apply to the concrete examples of modules that I can think of. For
        >the unrealistic counterexample, imagine a module for addition. Both
        >input and response sets would be infinite. However, the computation
        >is specialised and limited.
        >
        >But perhaps I misunderstand. Do you mean by small sets not sets with
        >small numbers of elements, but sets with strictly limited types of
        >input and output? Then I would agree entirely.

        Yes, that's what I was meaning. Of course, an input may be an analogic
        signal, in which case it comes in an infinite number of possible values.


        > So the barn owl's
        >azimuth computation takes in phase locked action potentitals which
        >transmit timing information and returns an angle, and it does nothing
        >else whatsoever.
        >
        > > It seems that many things in living organisms are modules, for
        > > example, a single neuron likely is.
        >
        >For the purposes of theoretical neuroscience, people have looked at
        >modules as small as three or four neurons in neural netowrk models,
        >and I suppose you could even think of single neurons as modules in
        >that context, but I don't think that makes sense for psychology. At
        >least in vertebrates, I would bet moderate sums of money that there is
        >nothing you can see in behaviour which would depend on a single
        >neuron.

        Ok, but I was not suggesting that. I agree that no psychological adaptation
        could be implemented with just 1 neuron, nor probably with just few. Just
        that a neuron seems to be a module, and many other cells and systems in the
        organism likely are too.

        > > However, it's impossible to imagine a single module for a thing such
        > > as Mate Choice. There are too many different inputs
        >
        >Depends again on how you look at it. In the barn owl, you have that
        >module for determining the azimuth of a sound source from interaural
        >time difference. There is another for determining (at high enough
        >frequencies) the elevation from interaural intensity differences (the
        >barn owl is one of the species with asymmetric ears). I don't
        >remember now whether these feed first into an separate auditory map,
        >or directly into the optic tectum for combination with visual
        >information. In any case, you have to expect not just a hierarchy,
        >but a network of modules. Not necessarily a tidy chain of command,
        >but something more like low and high level subroutines. A mate choice
        >module calling on a number of lower level modules is, at least in
        >principle, possible.

        I question that for a thing like mate choice, because it would not comply
        (apparently) with the requirements (with which you and me agreed) that a
        module should have a limited set of types of inputs, a limited and
        tractable set of computations. Mate choice seems to involve thousands of
        different types of inputs.

        If it is implemented through many modules acting in concert, can the whole
        set be called a module?

        Also, if it can be called a module, what are the alternatives? Other than
        modules, in what ways complex adaptations can be implemented?
        Self-organizing systems? Systems such as the "stigmergie" of the social
        insects?

        How can an hypothesis about the modularity of an adaptation be falsified?

        In my view, modularity should be defined in a way that made a claim of
        modularity to be non-trivial, otherwise, we run the risk of declaring that
        any adaptation must be modular, because the small components (neurons, T
        cells, etc) are so, which would make modularity to be the only game in town!
      Your message has been successfully submitted and would be delivered to recipients shortly.