2 ms·
They explained in the article how their decision-making still depends on circumstance, so it's not a rule so much as a bias. A properly calibrated bias will mi
by skybrian 7y ago
They explained in the article how their decision-making still depends on circumstance, so it's not a rule so much as a bias.
A properly calibrated bias will minimize error when you don't have any further information. This is like assuming the probability of rain is low in Arizona. It does rain, but no rain is the way to bet if you don't know how to do better. A more sophisticated weather prediction algorithm should be compared with a simple one to see if it actually does better.
But you do need to think about whether false positives or false negatives are more harmful, and whether uncertainty can be decreased by gathering more information.