By: Jon Baron

Jon Baron — Sat, 09 Jul 2005 19:00:00 +0000

You ask whether it matters if you pick an organism at random or
not. I think the answer depends on what you mean by (the
implicit) "or not." If you select one because it is particularly
well designed, then you must take into account the effect of your
choice itself on the quality of the evidence. You would want
something like p(E/H) and p(E/"~H and I picked this because it
looked good"). The latter would of course be higher than
p(E/~H), and p(E/H) would also be higher than for some organism
picked at random.

You also ask whether there is a rational way to revise beliefs.
Of course I think there is, although it is often essentially
impossible to apply the normative model in real cases. This
case, though, is reasonably simple. The normative model is
Bayesian theory. I sketch a justification of it in
my chapter in
the Blackwell Handbook of Judgment and Decision Making.

Of course, people do not follow this model, which keeps people
like me - psychology researchers who study biases - in business.

Comments on: a bayesian analysis of intelligent design

By: Jon Baron