We apply a battery of concepts in trying to understand brains. Finding conceptual common ground here is difficult. It seems to me that two accounts of the relation of the brain and body of an organism to its environment may usefully be compared. Each of these accounts tries to describe the relation between the state of the organism and the state of the environment at any given time. They differ in substantial ways, however.
The first set of concepts applied to understanding brain function emphasizes the predictive nature of the brain. Examples include predictive accounts of motor control, as in Wolpert’s work, the entire dopamine literature, and many functional accounts of frontal cortex – though the phrenology of prediction is complex, including primarily frontal and parietal lobes, but ultimately involving pretty much the whole brain. Common to these accounts is the notion that if we do something at t_0, it very much looks later as if the brain were expecting a specific effect of our act upon the environment. Usually, we say that the brain is running a little simulation, or calculating a forward model, so that at a later time, t_1, the actual state of the environment can be compared to the predicted state, thus allowing error correction and fine tuning of movement.
The second set of concepts are dynamical, and they are found in the ecological, enactive, and dynamical approaches to cognition. These approaches model system dynamics, and the system that is so modelled may, and typically does, extend across the organism-environment boundary. Indeed, a strength of these approaches is their concern with delineating the shifting boundaries of autonomy. Consider a person, standing upright in the experimental moving room of Lishman and Lee (1974). The entire room rests upon wheels or rollers. As the subject stands straight in the middle of the room, we can gently move the room back and forwards. This produces a changing pattern of optical stimulation upon the retina that would, under normal circumstances, only be generated by swaying of the torso to and fro. As a result, the subject will exhibit compensatory torso sway, locked in anti-phase with the oscillation of the room. Here, the fullest description of the lawfulness inhering in the situation is to be found in a dynamical description of two yoked sub-systems: the person and the room. They are not distinct, and their autonomy is somewhat lost in the coupling that results.
One of the principle differences between these accounts is in their treatment of causality. Causality is complex. Here, we are faced with two classical kinds of Aristotelian causality: efficient cause in the case of prediction, and formal cause in the dynamical description. The former, efficient cause is characterized by the notion of temporal sequence. An initial action, A occurs earlier in time, and is invariably followed by a distinct event, B at a later time. A caused B. Backward causation is not possible. If A occurs before B, it is not coherent to talk of B causing A.
The latter, formal cause of the dynamicist treats time differently. Here, we seek the fullest description of the behaviour of the state descriptors in time. Time acts as a canvas upon which events are drawn, and from which they are moulded or sculpted. If the system finds itself headed towards an attractor, it is by no means incoherent to consider the attractor as the proximal “reason” for the evolution of the system towards it, but as it is later in time, it can not play the triggering role of the initial event in the A-then-B sequence.
The predictive brain account typically rests implicitly or explicitly upon some notion of executive control. This feature of pretty much all information processing models of cognition is problematic in many ways. It may be a homuncular fallacy, or we might side with Dennett in hoping that each appeal to an executive process demands less and less fluid intelligence, so that at bottom, the regress is not infinite, but runs out in a hoard of mechanisms. Usually, modellers assume it is some else’s problem, as they get their teeth into a part of the problem that does not rely on so chimerical a construct.
But the executive self, appearing in a tableau based on the notion of efficient cause, is crucial to our notion of agency, responsibility, and free will. In the intentional realm, we like to think of our actions having consequences. It has proved useful in our history to think like this, and the normative constraints that ensure that we wear pants to work and do not kill our own kind, are crucially reliant upon an efficient cause notion of action.
The notion of agency and the executive self both serve to further suggest that there is a split between organism and environment. A split that is part of the edifice upon which we have built societies. And which can crumble at any moment. There are lawful biological processes going on between us, that are emergent at the collective level and only available by inference to us down here. They can look a lot like fate, as properties of the constituent layer (that’s us) bring about phenomena whose very existence is not predictable at the constituent layer, but which is manifestly influenced by arbitrary subsets of their properties – those responsible for the emergent properties.
…. or (Nov 2010) ….
Agency + prediction = entrainment. Without the spooky stuff.