Could anyone kindly address Ypbor's question? I'm stuck with the same puzzle: while I know that the ratio of $\alpha$ and $1-\gamma$, by being different from 1, is supposed to give a preference for early or late resolution, I am unable to prove that with a simple example (for example, the classic Kreps-Porteus example where there are two consumption streams, (5,10) or (5,0) with equal probability 0.5, but uncertainty may resolve early or late. If it resolves early, you know both period payoffs; if it resolves late, you know the second period payoff only later.