Thinking like a human, a computer program falls for similar illusions

If you designed a computer program to make decisions like a human, wouldn't you expect it to make human mistakes?

At the University College in London, researchers made a computer program that acquired knowledge the same way humans do, from experience, trial and error. The computer studied a series of gray-scale images. Its job was to judge the lightness of the shade at the center of each image. As the computer made its choices, it changed its decision technique depending on whether its previous decision was right or wrong.

Not surprisingly, when looking at certain gray-scale contrasts, the computer program fell for the same optical illusions that trick most of us. For example, when it studied a light object on a darker background, it predicted that the shade of object was lighter than it really was. Dark objects on lighter backgrounds similarly fooled the program. The computer also fell for 'White's Illusion', a pattern in which the same shade of gray is usually predicted by humans to be two different shades (see link for illustration).

The program points out the difference between hard-wired, computer-coded decision making and the more human process of making judgements based on experience. Human minds are not wired for accuracy, they're wired to find what's useful. link