Thursday, March 22, 2007

Bernard Baars' Global Workspace Theory

Baars' global workspace theory is a good example of how computer modeling can help us to understand how consciousness works without helping us to understand what consciousness is.

Brain scans during studies of binocular rivalry (where you show different images to the two eyes and the subject is consciousness of only one at a time) have shown that visual consciousness in the brain involves a series of steps. First, the brain identifies a visual field of pixels, then lines and edges, then motion, then colors, and so on. Only in the last step, when there is object recognition in the lower temporal cortex, does the subject become conscious of what he is seeing.

Baars theorizes that consciousness acts as a global workspace that accumulates information and then feeds it back to other parts of the brain. He has developed computer models of consciousness on this basis.

This does help to explain the function of consciousness and why it evolved. I believe you can see it in the waggle dance of honey bees: they seem to be accumulating information as they watch the dance, and once they have accumulated enough information, they go to the location that the dance pointed at.

It is hard to avoid thinking that the bees are conscious when you look at the waggle dance, just as it is hard to avoid thinking that a dog is conscious when he sniffs around before deciding which way to go. Consciousness involves a delay that lets you collect more information before acting than you can collect when you act on reflex.

But this does not explain the fact that we evolved consciousness rather than evolving the sort of mechanical memory workspace that a computer uses to collect information. Presumably, it was more efficient for the brain to evolve in a way that uses consciousness to hold this information than to evolve in a way that uses computer-like memory to hold this information. But what is this consciousness that the brain evolved to use? That is still a mystery.