> Though those biases probably came about for good reasons, it could be they've become obsolete and are no longer worth it.
I suspect (but don’t know how to test the hypothesis) that cognitive biases are why human learning can produce good results with dramatically less data than machine learning. More rational, yes, when you get there; but harder to learn at all.
No disagreement there! But, at the risk of cargo-culting, it might be interesting to see the effect of deliberately trying to reproduce such biases in an A.I.
I suspect (but don’t know how to test the hypothesis) that cognitive biases are why human learning can produce good results with dramatically less data than machine learning. More rational, yes, when you get there; but harder to learn at all.