Patterns

In a recent article in Science, several authors present an algorithm designed to determine which of several predefined, but quite expressive, structures best represents a set of data. The article, How to Grow a Mind explicitly states the goal of reverse engineering the human cognitive process, in particular our ability to draw causal connections based on only a few pieces of data. It's a fascinating capability, one that computers have been poor at emulating.

Pause.

That's right: we want to help computers jump to conclusions better. It's ironic to me that we would spend effort to that end because I find myself so distinctly caught between extremes of promoting and fighting that characteristic in humans.

Fighting because as a professional analyst, the ability to carefully weigh evidence and avoid the key biases--hindsight, anchoring, base rate, immediacy, vibrance of memory--that undermine an effective judgment of a scenario is a prized possession. Much of my professional training revolves around helping to identify how your mind works in those ways in order to preempt the preemption and allow a more rational approach than the many cognitive shortcuts that we have available.

But I promote that same short-circuiting of patterns: in one of my two-year-old daughter's favorite books, Superbabies, comes a fantastic scene where she attempts to intuit the pattern on the babies recently torn blankets. The pattern is red, blue, red, blue, red, ???. What is the right answer? Blue, of course. It's so simple. The superbabies don't have it worst: in many of her books, the authors encourage her to see patterns on the basis of the thinnest of evidence. I'm torn: I want her to have that normal cognitive skill (though I suspect she'd probably develop it whether it's in the book or not), but about half the time I'm reading I think to myself: this could be so many things!! Two instances does not a pattern make!! Oh well.