The Neuroscience of Racism: Why some People Think “They all look Alike”.

Common Sense tells us that we recognize things by noticing their properties, and then making logical inferences from those properties to put the things into categories. Common Sense also tells us that the world is flat. Astronomy and physics tells us that Common Sense is wrong about the latter, and Computational Neuroscience now tells us that Common Sense is wrong about the former. Our brains actually classify things by a complex statistical process, which formulates no rules, and notices no properties. We now have AI systems that use that process to identify letters of the alphabet, spoken words — and faces. I have written several papers on the philosophical implications of the Neurocomputational software designed to recognize faces. I am not familiar with all the details of the commercial and governmental applications of this software. However, I know enough about its basic principles to clear up some confusion about how it works, and discuss what it tells us about how human brains work.

The first and most famous of these is Garrison Cottrell’s face recognition program. To greatly oversimplify: The program was created by showing the system a “training set” of face images, then making adjustments until the program correctly identified each face it saw. The adjustments were achieved by a complex algorithm called “back propagation” whose details needn’t concern us here. The important point for our purposes is that this algorithm can train a system to output the same signal every time it…