Often, we do have a fairly good idea what the ramifications will be if an experiment goes a certain way. But we almost never also have a good idea what the ramifications would be if things go the other way. Without knowledge of what the benefit might be if the outcome surprises you, how can you say that the experiment doesn't have enough potential benefit to be worth the time and money?

There's no way to place an easy upper bound on the payoff of the less-expected outcome, so your cost-benefit analysis can't simply ignore the possibilities that you consider unlikely.