@f0x Looking at how people reacted to azile/eliza, and the (low) complexity of most game AI, I want to hit a couple bullet points

* Yes cute sprites, but they shouldn't all convey the same emotion. The goal here is to start peoples' mirror neurons firing, and in a real setting we usually ease into that by looking at a prominent face

* Some element of randomness - even if it's "the fox is happy and wants to play" vs "the fox is happy and wants to nap on the warm grass", trips a bunch of signals including operant conditioning, where we want to pet the fox more, and a sense that fox has its own internal motivators separate from our own, which spins the mirror neurons up harder to divine its mood

* I'm conflicted about putting the mood up as a word or icon. It might be good for accessibility reasons, but if so, definitely a coarse-grained "happy" or "sad". We're all in this together in observing fox and guessing at their motives, the point is to guess and project.