Innate probabilistic inference
It’s been commonly assumed (by statisticians and developmental psychologists) that statistical reasoning is a completely learned, non-innate trait. Pop evolutionary psychology explanations focus on the immediate benefits to early hunters from being able to deal with problems such as predicting the intersecting arcs of thrown objects (a complex mathematical problem only relatively recently solved algorithmically in WW2 for anti-aircraft defenses). This is contrasted to the seeming inability of adult humans to reason consistently, especially in situations requiring a Bayesian analysis, about probabilities. Now doubt has been shed on many of the assumptions. Work by Fei Xu and Vashti Garcia (U.B.C, Vancouver) demonstrates that 8-month old infants are able to draw basic inferences about populations based upon samples[1]. Babies are shown boxes containing different proportions of red and white balls. Subsequently a fresh box is presented and a sample of balls is drawn, the box is then opened to allow the baby to see the contents. If the sample differs largely in proportion to the population then the babies stare longer — a trait which has elsewhere been observed to be a reaction to unexpected events. This still is very simple inferential reasoning, but it’s a demonstration of an unexpected ability. 1. http://www.pnas.org/cgi/content/abstract/105/13/5012
4 years ago