I think that if the rate of genetic change increases then acceleration is an fair word.

If one defined the prevalence of each gene in the population you would get a vector were you could possibly 'change direction'. None of this requires any metric of 'goodness' to be defined if that is what you are worried about.