‘The bureaucracy is expanding to meet the needs of the expanding bureaucracy.’ – Oscar Wilde
We, humans, face a similar paradox when it comes to AI and our emotions. A lot of our brainpower goes to compensate and build on top of our emotions, that is because it is not healthy to go against them from a personal perspective. Plus our emotions have a specific social function, so it is highly unlikely for us to have a social life without them. Our emotions make us socially pleasant and predictable, but if AI shows (simulate) such emotions-like behaviour we would think that we have been deceived – the AI does not actually have emotions. I don’t see the point in that, if the behaviour is socially pleasant and predictable, why the underlying mechanic should matter? It’s like asking everybody to have the same history because there is only one way (the right way) to be motivated for good.
Now about the paradox, making AI consider emotion-like behaviour including most of the known cognitive biases is against its intellect but makes it more acceptable for society.
If simplified, the more intellect power is devoted to dumbing down AI in a specific way, the more successful in the eyes of the people it will be. So we will create high intellect in order to play well stupid.