We did the Really Getting Bayes game at the Mountain View meetup this week. My impression of it was that the explanation was at first a little unclear, but that once we had gotten the sense of it, it was worthwhile.

One thing that I realized during the game was the degree to which I was using the availability heuristic to provide my likelihood ratios. For instance, one object I was given was either an electrical extension cord or an audio cable. In coming up with RGB likelihoods, I thought, "Electrical extension cords are usually black or white, hence not RGB; whereas audio cables are often RGB" — by using instances I could bring readily to mind to come up with this thought. My partner then told me the item was RGB, and I correctly predicted it as an audio cable. However, the electrical cord in his other hand was in fact orange, not black or white. While my availability heuristic worked for the purpose at hand, it clearly would have been wrong for other purposes.