On Computational Models of Interconnected Perception, Reasoning, and Action — in Dream Worlds

Abstract

We distinguish between real versus unreal worlds, and include in the latter category fictional worlds and — perhaps the hardest type of unreal world to plumb — dream worlds. Dream worlds, from the standpoint of building computational cognitive models, present a number of acute challenges in at least three areas of human mentation that are for us as AI researchers and computational cognitive scientists deeply interconnected; these areas are: perception, reasoning, and action. We are attempting to specifically answer three tough questions, one in each of these three areas; the answers (for us) must be based upon robust computational models that are both theoretically well-founded, and brought to life in implemented simulations. (Such models are to our knowledge in very short supply.) Ultimately, we acknowledge that the research trajectory we summarize herein must arrive at a way to informatively model different modes of consciousness in the human mind, which, remarkably, is capable of sustaining utterly coherent consciousness across waking, imagining, and dreaming.