/ apophenia

My computer doesn't believe in ghosts.

I think in my writing I hint quite a bit at the pervasiveness of pattern recognition in the human ability to "fill in the blanks" when interpreting the world, tonight is going to be a somewhat short post on apophenia -- which is to say the fine art of seeing patterns where there are no patterns at all, we may sneak in some talk on pareidolia a subset of apophenia that involves not only seeing patterns but believing those patterns are significant.

Apophenia is not always a good thing and often paired with mental disorders such as psychosis because once people become perhaps too good at seeing patterns we start to see them everywhere, thus you get conspiracy theorists or those with a firm and unshakable belief in the paranormal. Indeed, it shows something of the catch-22 in programming a computer that can make quick inferences based on patterns in high volumes of information -- at what point do we create a system that sees too many patterns and starts to believe in ghosts?

About a year ago there was a study done that I am still rather obsessed with, in this particular study a group of scientists attempted to recreate the symptoms of schizophrenia in a computer system (DISCERN).[1. http://www.utexas.edu/news/2011/05/05/schizophrenia_discern/] I think I read this paper about 12 times the first week alone. The hypothesis was that in people 'hyperlearning' or the inability to shut out useless 'noise'[2. I'm not sure how clear this is but not literal noise, but like, image noise] information would cause the system fed this much information to begin to get confused and begin to exhibit the jumbled thought patterns we tend to see in schizophrenic patients. This experiment was a success, they got the desired disoriented conversation. Cake.

As I mentioned above, a subset of apophenia is pareidolia which means the interpreter also applies meaning to patterns that aren't really there -- the most common expression of this being the human ability to see faces in just about everything -- from extremely simple geometric shapes to well, the burn patterns in your toast. I just wonder if it is probable or even possible to overwhelm a system as seen in the DISCERN example above with the sort of hyper-pattern recognition that humans seem to have. Most false positives for us when seeing patterns is mostly harmless if occasionally annoying when we don't also identify the false pattern.

More on this at a later date perhaps.