Abstract

In this talk I explore the idea that consciousness is something that one learns rather than an intrinsic property of certain neural states, and suggest that this perspective offers a way of reconciling Global Workspace Theory with Higher-Order Thought Theory. Starting from the idea that neural activity is inherently unconscious, the question becomes: How does the brain learn to be conscious? I suggest that consciousness arises as a result of the brain’s continuous attempts at predicting not only the consequences of its actions on the world and on other agents, but also the consequences of activity in one cerebral region on activity in other regions. By this account, the brain continuously and unconsciously redescribes its own activity to itself, so developing systems of meta-representations that characterize and qualify their target representations. The main functions of such metarepresentations are (1) to redescribe the target first-order representations in such a way as to explicitly indicate mental attitude, and (2) to subserve prediction-driven control mechanisms. Crucially, (1) such metarepresentations do not need to be conscious themselves (as in HOT ), and (2) they emerge over training and development as a result of unconscious prediction-driven learning and plasticity mechanisms. Metarepresentations form the basis for self-awareness because they enable agents to “know that they know”, that is, to be acquainted with the geography of their own representational systems. In a sense thus, this is the enactive perspective, but turned both inwards and (further) outwards. Consciousness is “signal detection on the mind”; the mind is the brain’s (non-conceptual, embodied implicit) theory about itself. I subtend these ideas by exploring empirical evidence that conscious experience is shaped by learning and through neural network models that simulate the relationships between performance and awareness in the different tasks. Next, I turn to the link between self-awareness and theory of mind. The main argument is that developing infants continuously attempt to predict not only the consequences of their actions on the world, but also the consequences of their actions on other agents. But there is a crucial difference between interactions with the world and interactions with other agents: Understanding the reactions of the latter, unlike the former, requires assuming the existence of hidden, unobservables states. Thus, when one learns to interact with other agents, one also forms mental models of the internal states of those other agents. But this involves the same prediction-driven processes as involved in forming metarepresentations of one own’s mental states. There is thus a direct link between theory of mind and self-awareness, a point that was forcefully argued by Carruthers. Hence we bridge the gap from theory of mind to perceptual awareness through the joint involvement of prediction-driven, learned interactive loops that make it possible for agents to better anticipate the consequences of their actions.