Honestly, I think it's pretty straightforward. Just code the AI to maximize happiness for the largest number of people. It's all about the utilitarian calculus, folks. Sure, there will be hiccups, but it's the least worst option we got.