It's not hard - it's impossible. Even we fully agree on a specific metric, a situation still has to be weighed against possible alternatives and integrated on timescales longer than our lives. A very basic example of this is a company optimizing for "profits", but they're actually optimizing for short term profits at the expense of going out of business two years later. The future is unknowable in the same exact sense that every NP-hard problem is. Heuristics are the only way to tackle this.
> coming up with qualitative points and weighting them in a good-faith-but-ultimately-arbitrary-manner is good enough, and something that everyone does constantly
Of course, which is why I'm referring to "Faceboot" even while recognizing that people get utility out of it. I just think that focusing on arbitrary pairs of specific facets is already headed down the path of madness, which is why I stated my initial critique in terms of basic principles.
Yes. This is why every single comment of mine in this thread recommends a specific heuristic with which to tackle this exact problem.
In practical terms, coming up with qualitative points and weighting them in a good-faith-but-ultimately-arbitrary-manner is good enough, and something that everyone does constantly. For example, I suspect the average HNer weights "wasted human potential from FB" more than "benefit of disaster reporting from FB" (which I also agree with - however "benefit of disaster reporting" almost certainly outweighs "detriment from nebulous analytics firm scandal").