March 15, 2011
Confidence
Here is an important insight from the webcomic Saturday Morning Breakfast Cereal:
This is why I don't trust confident people and why I am in favour of radical life extension.
From a Bayesian standpoint maybe this kind of behaviour can be explained by the growth of the number possible hypotheses and the accumulation of evidence (beside the obvious irrational biases, of course). At first we have few hypotheses to choose between, and new evidence makes us more confident in one or a few. Then we start seeing the possibility of more hypotheses, and now the accumulating evidence starts to update a much broader field of hypotheses. In the really long run we finally get over the explosion of possibilities and start to zoom in on the best ones again.
Presumably this would be different in different domains. In simple domains there will be few possible alternative hypotheses and confidence will just increase. In domains with little evidence accumulation but easily generated hypotheses confidence will tend to decrease. Examples of such domains are left as an exercise for the reader.
Posted by Anders3 at March 15, 2011 02:15 PM