Ah, so what we're really talking about here is situations where the world state keeps changing as the memory builds its model.. or even just a situation where the memory has an incomplete subset of the world information. Reading the second article's example, which makes the limitations of the memory explicit, I understand. I'd say the chess example is a bit misleading in this case, as the discrepancies between the memory and world are a big part of the discussion -- and as you said, chess is a perfect-information game.