re: #78 The Ghost of Quesos Past

The other thing to be scared of?

People that want to justify a conclusion they’ve already reached will build software that counterfeits analysis but reaches a foregone result. Done well enough, you’d have to climb into the guts of the algorithm to falsify it.

So, no different than statistics.

There are already groups working on standards for training AI algorithms and determining what data an algorithm is being trained on.