
16 Evolutionary Generation of Faces
422
algorithm, not ideal (Hancock 1997). It seems clear that this should be replaced
by a ranking scheme, in order to remove the scaling problems associated with
participants' erratic use of the rating scale, but we suspect that we may do better
by asking users to pick the best three, and then allowing each of those faces two
reproductive opportunities in the next generation, akin to the (~,2) approach of
evolution strategies. We think this may be an easier task for users.
One reason for the relatively poor performance of our participants here may
be simply that the faces are all so