2012 : WHAT IS YOUR FAVORITE DEEP, ELEGANT, OR BEAUTIFUL EXPLANATION?

[ print ]

Psychologist; Director of the Center for Adaptive Behavior and Cognition at the Max Planck Institute for Human Development in Berlin; Author, Gut Feelings

Unconscious Inferences

Optical illusions are a pleasure to look at, puzzling, and robust. Even if you know better, you still are caught in the illusion. Why do they exist? Are they merely mental quirks? The physicist and physiologist Hermann von Helmholtz (182-1894) provided us with a beautiful explanation of the nature of perception, and how it generates perceptual illusions of depth, space, and other properties. Perception requires smart bets called "unconscious inferences."

In Volume III of his Physiological Optics, Helmholtz recalled a childhood experience:

"I can recall when I was a boy going past the garrison chapel in Potsdam, where some people were standing in the belfry. I mistook them for dolls and asked my mother to reach up and get them for me, which I thought she could do. The circumstances were impressed on my memory, because it was by this mistake that I learned to understand the law of foreshortening in perspective."

This childhood experience taught Helmholtz that information available from the retina and other sensory organs is not sufficient to reconstruct the world. Size, distance, and other properties need to be inferred from uncertain cues, which in turn have to be learned by experience. Based on this experience, the brain draws unconscious inferences about what a sensation means. In other words, perception is a kind of bet about what's really out there. But how exactly does this inference work? Helmholtz drew an analogy with probabilistic syllogisms. The major premise is a collection of experiences that are long out of consciousness; the minor premise is the present sensory impression. Consider the "dots illusion" based on V. S.Ramachandran and colleagues.

The dots in the left picture appear concave, receding into the surface away from the observer, while those on the right side appear convex, curved towards the observer. If you turn the page around, the inward dots will pop out and vice versa. In fact, the two pictures are identical, except for being rotated 180 degrees. The illusion of concave and convex dots occurs because our brain makes unconscious inferences.

Major premise: A shade on the upper part of a dot is nearly always associated with a concave shape.
Minor premise: The shade is in the upper part
Unconscious inference: The shape of the dot is concave.

Our brains assumes a three-dimensional world, and the major premise guesses the third dimension from two ecological structures:

1. Light comes from above, and 
2. There is only one source of light.

These two structures dominated most of human and mammalian history, in which the sun and the moon were the only sources of light, and the first also holds approximately for artificial light today. Helmholtz would have favored the view that the major premise is learned from individual experience, others have favored evolutionary learning. In both cases, visual illusions are seen as the product of unconscious inferences based on evidence that is usually reliable, but can be fooled in special circumstances.

The concept of unconscious inference can also explain phenomena from other sensory modalities. A remarkable instance where a major premise suddenly becomes incorrect is the case of a person whose leg has been amputated. Although the major premise ("A stimulation of certain nerves is associated with that toe") no longer holds, patients nevertheless feel pain in toes that are no longer there. The "phantom limb" also illustrates our inability to correct unconscious inferences in spite of better knowledge. Helmholtz's concept of unconscious inferences has given us a new perspective on perception in particular and cognition in general.]

1.Cognition is inductive inference. Today, the probabilistic syllogism has been replaced by statistical and heuristic models of inference, inspired by Thomas Bayes and Herbert Simon, respectively. 
2.Rational inferences need not be conscious. Gut feelings and intuition work with the same inductive inferences as conscious intelligence. 
3.Illusions are a necessary consequence of intelligence. 

Cognition requires going beyond the information given, to make bets and therefore to risk errors. Would we better off without visual illusions? We would in fact be worse off—like a person who never says anything to avoid making any mistakes. A system that makes no errors is not intelligent.