[admax88q]

> Normal is a pretty shitty heuristic for being good.

A few counter examples doesn't mean that the heuristic is bad. Heuristics will always have counter examples, that's why they're heuristics, not rules.

[shkkmo]

I would argue that inherent discrimination against minority traits is what makes that heuristic bad.

Given a single counter example, the burden of proof should be on the users of the heuristic to show why the use of that heuristic is an improvement over the alternatives (which have no counter examples).

[50656E6973]

How many counter examples do you need before determining if a heuristic is bad, in your opinion?

[SirLuxuryYacht]

IMO, probably when the counterexamples outnumber the examples by which the heuristic is defined.

[coldtea]

And maybe even not then: the negatives from the counterexamples should also outweigh the positives from the heuristic.

If a heuristic has 100 majorly good applications, and 1000s of counter-examples of minor or no real consequence, it will still be good to use...