Jon Pretty

false. You need to think globally. There's someone from 2050, so at least some humans survived. So you can learn how they survived, and use that to influence globally the path toward sustainability:https://twitter.com/fanf42/status/1192896905233256448 …

But you can do so very little with just one binary answer...

Imagine how it changes if with probability p they lie.

Well it certainly doesn't make things easier... But I guess that the utility of the answer would become proportional to 2|½-p|.

I think 1-H(p) but with a single question it’s hard to think what this really means for a human.

What's H? If they lie with probability 1, that's as good as telling the truth. And p = ½ gives no information. Ok, so let's assume the person from 2050 is omniscient enough to answer, "was reducing our emissions by X enough to save the planet?" And we have to pick X.

So, should we choose X to be the value for which current science believes there's a 50/50 chance it will be "enough"? Or would it make more sense to pick a smaller X, while expecting the answer to be much more likely "yes" than "no", but *if* we get a "no", we learn much more.

Yep. We want to pick the point where we think we have a 50/50 chance of yes/no to maximize our information.