The cognitive agent’s problem

by Neil Rickert

A lot of theorizing about cognition has to do with the such questions as

  • What is thinking, and how is it done?
  • What is experience, and how come we have it?

Then, having selected such issues, theorists then set themselves the problem of how would one design an agent or a robot which can do such things as think and have experience.

When I started thinking about cognition, I took a different approach.  I looked at the problems that a cognitive agent needs to solve.  Most obvious, among those, is the problem of survival as a biological organism.  And, having settled on a problem that agent needs to solve, I set about trying to understand what could evolve to solve that problem.

The reader will perhaps notice an obvious difference there.  The conventional approach is based on design thinking, while I based my approach on evolutionary thinking.  And that may be why I have a rather unconventional view of cognition.

Looking at the problem that I saw, the problem of survival, there is always the possibility that an organism could start its life in nutritionally rich surroundings, and continue to survive for as long as it can absorb nutrients, then die.  But that isn’t the case for most creatures.  Usually, they need to do something to seek food.  Even a plant has to seek food, growing its roots where they can find sufficient nutrients and growing its leaves where they can find sufficient light for photo-synthesis.  The problem is greater for animals, with a need to seek out sources of food.

In order to do any kind of seeking, there is a need for information.  In a simple case, say the roots of a tree, perhaps the information on whether enough nutrients are coming in is all that is required.  And then the tree can grow roots in other directions, if the current roots are not satisfactory for the needs.  So the information need then is mostly information about internal metabolic conditions.  A mobile animal requires rather more information about where to look for food and about how to recognize food.

That leaves, as one of the basic problems for a cognitive agent, the problem of getting information about the world.  And that is the primary problem that has been my concern.  For advanced animals, such as mammals and birds, that amounts to the problem of perception and how it works.

I’ll say more about the problem of information in future posts.  As background, I’ll mention the work of J.J. Gibson and of David Marr.

Marr takes a view of geometric view of vision that is similar to what is usually described.  I’m sure that I was already familiar with at least an elemental version of that view of vision from my teen years, perhaps earlier, though I did not hear of Marr until sometime later.  It is pretty much the kind of view that anyone taking a design perspective is going to consider.  Marr did a lot of work on the theory, and filled in a lot of detail that were not in the elemental version.  But, overall, it is a design approach to the study of vision, presumably aimed at producing the kind of visual experience that we have.

I learned of J.J. Gibson, only after I had been working on my project for a while.  Gibson’s view is very unconventional.  He talks of the objects of perception as affordances, as affording the cognitive agent opportunities, which presumably includes the opportunity of enhancing the likelihood of survival.  So Gibson’s ideas can be seen as based far more on an evolutionary type of thinking.  It looks at the needs of the organism, and how they can be met, whereas the Marr approach looks at the need of human philosophers trying to explain visual perception.

My own view of perception can then be thought of as along the same lines as Gibson’s ideas, but  not at all like the more conventional ideas of Marr.

One Trackback to “The cognitive agent’s problem”

%d bloggers like this: