This is question-begging a number of critical elements, e.g. that the "rafts" cannot be influenced by "passenger" input, and that there is only this one, totalizing crossroad of literal, immediate survival.

We can do it too:

You're in a runaway train accelerating toward a cliff and the break only really stops acceleration, it doesn't decelerate. You can sit in the engine room and hold down the break, and you'll live longer, but you aren't changing the fundamental dynamic of the situation, which ends in your eventual death. Conversely, you can jump off the train, surely injuring yourself, possibly crippling yourself, maybe even killing yourself, but it's the only potential way to change the dynamic of being doomed to fall off the cliff.

Does this prove anything? No, it's just a model of how some people think of the problem, not an argument. It would be really obnoxious and disingenuous to present it as an argument.