Nah, you can't choose to un-donate. Whereas you can always make up for lost time. So giving is a case where some mild indecision may be worthwhile.

Obviously the current expected value of your action should be the same as what you expect to think in the future. But getting more info can increase the expected value of your information thus:

Let's say you have a charity budget, and two charities, A and B. Since your budget is a small fraction of the budget of each charity, assume your utility is linear in this decision, so you'll give all your money to the better charity. You think there's a 60% chance that charity A produces 1000 utilons from your donation and charity B produces 100, and a 40% chance that A only produces 10 and B still produces 100. The expected utility of giving to A is 60% * 1000 + 40% * 10 = 604. The expected utility of giving to B is 60% * 100 + 40% * 100 = 100, so you are planning to give to A.

But let's say that by doing some small amount of research (assuming it's costless for simplicity), you can expect to become, correctly, nearly certain (one situation has probability of ~1, the other has probability of ~0). Now if you become certain that A produces 1000 utilons (which you expect to happen 60% of the time), your choice is the same. But if you become certain that A produces only 10 utilons, you give to B instead. So your expected utility is now 60% * 1000 + 40% * 100 = 640, a net gain of 36 expected utilons.