Finding Patterns Everywhere

[resqgeek]

The human ability to recognize patterns is incredibly robust.  I have been observing the developments in the field of computer vision since 1990, and while the progress that has been made in programming computers to detect and recognize patterns has been impressive, the capabilities of the human brain in this area continues to elude the developers in this technology almost entirely.  Not only do humans have powerful, innate abilities to detect patterns with almost unbelievable speed, but our recognition accuracy is remarkably high.

However, it isn’t perfect.  One of the weaknesses we have in this area is the tendency to find patterns where none exist (the false positive result).  The relatively high false positive rate is a product of our evolution...a false negative (failing to see a pattern where one does exist) can have fatal consequences (such as not seeing a predator hiding in wait), whereas running from a danger that doesn’t actually exist is much less likely to kill someone.  As a result, over time, natural selection minimized our tendency towards false negatives, but did not particularly suppress any tendency to find false positives.

This tendency to see patterns even when no meaningful pattern really exists is easy to demonstrate.  We use it to amuse ourselves when we look for shapes in the clouds, and it is the reason we find the patterns we call constellations among the stars.  It is also probably a major factor in why so many people believe in conspiracy theories.  They find patterns in events, giving meaning to coincidences, even when there is, in reality, no underlying relationship between the events.  It seems that we have evolved to dislike randomness, with a strong preference for patterns, so much so that we will go to great lengths to find patterns everywhere.

This is why scientific inquiry requires experimentation.  When we think we’ve found a pattern, we use the alleged pattern to make a prediction, and then conduct an experiment to see if the prediction is correct.  If it isn’t, we may need to re-evaluate the data to see if the pattern actually exists or if it is just a product of our imagination.