Patterns everywhere
I think it is safe to say that humans, primates even, are pattern seeking creatures. We find patterns everywhere. Faces and animals in the constellations, clouds and foliage. You can understand that last one. Evolution will have selected our ancestors based on an ability to spot a well camouflaged predator in the bushes.
But does this pattern seeking, and I emphasize seeking over mere recognition, serve us well in the analysis of data and finding explanations in science? Even our best technology for analogs of this ability, artificial nueral nets (ANNs) are mindless slaves to a bottom up drive, minimising the cost function, to “get the answer right”. And then we hope this ability generalises to unfamiliar samples. There has been much progress in ANNs since the multi-layer perceptron and now some researchers claim that the hidden layers of the network are in some way uncovering the salient features of the classification task in hand.
I’ll take a look at some of those advancements in a later post, once I’ve understood the details and reproduced the findings myself. But in asking this question I’m more concerned about the human tendency to overgeneralise and seek simple explanations where in fact none exist. Taking a look at this list of cognitive biases is a rather salutary reminder of ways in which we get this wrong.
And a summary of recent findings in the field of network science cast doubt on the prevalence of power law distributions that were only recently thought to be ubiquitous.
The burgeoning field of “data science” which draws from a hefty chunk of statistics and machine learning along with the availability of big data and the requisite data processing power of modern computer networks is a heady mix and these skills are currently in great demand in industry.
But a quick tour of the many distributions that could be fitted to a given data set and the difficult problem of noise and the high degree of skill and time required to do proper statistical analyses should be ringing alarm bells.
Alarm bells that warn of not only of the misuse of data and misapplication of “data science” along with a relentless mechanistic interpretation of the world fed by cognitive bias rather than the science of explanations and reason, but also the ability to scale and embed such flawed decision making so as to robotically and mindlessly distort the fundamental playing field of the human condition. This in the pre AI age could bring about a dystopia so profound as to halt this civilisation in its tracks. Caveat Emptor.