The agent should be able to be wrong about its expected valence. I think this is hugely important, actually, and explains a lot about human behavior.
Does this fall out of imperfect information? (Say, instead of green dots seeming nice, and the straight line to them being considered, with red dots seeming bad, and agents not taking paths that go straight to green that pass by red, there could also be limited visibility. Like, the other green/s are on the other side of the mountain and can’t be seen.)
Does this fall out of imperfect information? (Say, instead of green dots seeming nice, and the straight line to them being considered, with red dots seeming bad, and agents not taking paths that go straight to green that pass by red, there could also be limited visibility. Like, the other green/s are on the other side of the mountain and can’t be seen.)