From Sean Carroll's second response to Sam Harris:
Let’s say that killing a seventy-year-old person (against their will) and transplanting their heart into the body of a twenty-year old patient might add more years to the young person’s life than the older person might be expected to have left. Despite the fact that a naive utility-counting would argue in favor of the operation, most people (not all) would judge that not to be moral. But what if a deadly virus threatened to wipe out all of humanity, and (somehow) the cure required killing an unwilling victim? Most people (not all) would argue that we should reluctantly take that step. (Think of how many people are in favor of involuntary conscription.) Does anyone think that empirical research, in neuroscience or anywhere else, is going to produce a quantitative answer to the question of exactly how much harm would need to be averted to justify sacrificing someone’s life? “I have scientifically proven that if we can save the life of 1,634 people, it’s morally right to sacrifice this one victim; but if it’s only 1,633, we shouldn’t do it.”
At bottom, the issue is this: there exist real moral questions that no amount of empirical research alone will help us solve.