There are two conflicting principles at work in our cognitive design: I call them the Bayesian and the Miser. The Bayesian wants to understand the world. It seeks to integrate every piece of information, giving each its proper weight, and constantly updating its beliefs. When the posterior is significantly different from the prior, it experiences surprise; which drives the subject to pay attention, investigate and seek more evidence.
The Miser is the Bayesian's budget manager. It aspires to drive the cost of all cognitive computations to zero. It has all sorts of tricks for doing that — heuristics, cached thoughts, stereotypes, dogmas.
Both these principles are necessary.
If the Bayesian was left unchecked, we would be constantly enticed by surprising stimuli, and would have a hard time managing our attention. The smallest decisions would be painfully long and difficult. The cognitive effort required by the constant updating would soon drive us to exhaustion (some speculate a link between Bayesianism and autism).
If the Miser dominated, we would fail to experience surprise in the presence of new information. Data which conflicted with our priors would be mostly ignored or discarded. Our calcified beliefs and cached ideas would dominate over new evidence, and we would have a hard time changing our minds.
Which of these scenarios seem most familiar?
No comments:
Post a Comment