This is a linkpost for http://squirrelinhell.blogspot.com.es/2017/10/you-too-can-see-suffering.html
The best tradeoff is when you are well calibrated, just like with everything else.
"Well calibrated" isn't a simple thing, though. It's always a conscious decision of how willing you are to tolerate false positives vs false negatives.
Anyway, I'm not trying to shoot you down here; I really did like your article, and I think you made a good point. Just saying that it's possible to have a great insight and still overshoot or over-correct for a previous mistake you've made, and if you think that almost everyone you see is suffering, you may be doing just that.
"Well calibrated" isn't a simple thing, though. It's always a conscious decision of how willing you are to tolerate false positives vs false negatives.
I beg to differ; being well calibrated has a mathematically precise definition. E.g. if you are thinking of a binary suffering/not suffering classification (oversimplified but it's just to make a point), then I want my perception to assign such probabilities, that if you compare with true answers, cross-entropy is minimized. That's pretty much what I care about when I'm fixing my perception.
Of c...