In this paper we suggest a complexity-based taxonomy of some families
of preferences over lotteries. We define the complexity of a given fam
ily to be of order n if the question whether or not a given preference
belongs to this family can be determined by comparing lotteries who d
iffer on only n outcomes at most. We first show that the complexities
of the expected utility family, the weighted utility family and the be
tweenness family are all of order 3. We then show that the complexity
of singletons is always equal to 3. Finally, we provide examples of fa
milies of preferences with arbitrary orders of complexities. The paper
also provides a better understanding of the large body of experimenta
l results in the theory of decision making under risk.