I would like to begin this section with some general comments about new opportunities for human self-understanding, about extending the explanandum. Academic disciplines are standardly divided into the sciences and the humanities, and some have expressed discomfort about the distance between the two modes of inquiry, or between the two cultures, as Snow (1959) famously put it (also see Brockman 1996). There is an immediate appeal to Metzinger’s assertion that “Epistemic progress in the real world is something that is achieved by all disciplines together” (2003, p. 4). If my claims from the previous section are on the right track, then we have a convergence of results between the two independent modes of inquiry, between the empirical sciences and the humanities. It is tempting to hope that this convergence signals the beginning of a rapprochement between the sciences and the humanities. Perhaps we are at the threshold of a new science of the mind (Rowlands 2010), a science that finds natural and fruitful connections with the world of human experience. In this section, I will explore possible connections with education, public policy, and social interaction.
Clark makes two main claims in the final sections of his article that serve for the basis of my comments here. First, he suggests that PP motivates an understanding of cognitive processing as “maximally context sensitive” (p. 16), which follows from the property of PP systems being highly flexible in setting precision weightings for the incoming prediction errors. Flexibility in weighting precision enables flexibility in the deployment of processing resources. Thus there may be a wide variety of cognitive strategies at our disposal, with a continuous interplay between more costly and less costly strategies. Second, he addresses the challenge of explaining why humans have unique cognitive powers unavailable to non-human animals who have the same fundamental PP architecture. In response to this challenge, Clark suggests that our abilities may be due to our patterns of social interaction as well as our construction of artifacts and “designer environments” (p. 19). Taken together, these two claims can be used to inform practical decisions in a number of ways.
Begin with education. Educational psychology is a broad and important area of research. PP suggests new ways of approaching human learning, ways that might depart from the received views that have guided educational psychology. I cannot begin to engage with this huge issue here, but I would like to offer one quick example. One fairly well-known application of educational psychology is in the concept of scaffolded learning, which is built on work by Lev Vygotsky and Jerome Bruner. As it is used now, scaffolded learning involves providing the student with helpful aids at particular stages of the learning process. These aids could include having a teacher present to give helpful hints, working in small groups, and various artifacts designed with the intention of anticipating stages at which the student will need help, such as visual aids, models, or tools. Clark himself mentions the abacus, which is central example of scaffolded learning (p. 19). More generally, scaffolded learning is a good example of what Richard Menary has called “cognitive practices,” which he defines as “manipulations of an external representation to complete a cognitive task” (2010, p. 238).
If PP is right, then the learning process could be optimized by designing environments in order to provide the cycle of action and perception with precisely controlled feedback (prediction error). With the growing commercial availability of immersive virtual reality equipment, educators could design learning environments (or help students design their own environments) without the messy constraints of the physical world. PP may give us a framework with which to understand—and predict—the detailed bodily movements of subjects as they attempt to minimize their own prediction error. Using this framework, we can design systems that would optimize skill acquisition by efficiently predicting the errors that learners will make. This method could be fruitfully applied in the abstract (mathematics), the concrete (skiing), and in-between (foreign languages). Along these lines, the insights of PP, together with emerging technology, can lead to powerful new educational techniques.
Psychology is also applied in some areas of public policy. Clark mentions that PP challenges Kahneman’s well-known model of human thinking as consisting of a fast automatic system and a slower deliberative system (p. 18). Kahneman’s model has been applied as a basis for influential recommendations about laws and public policy in the United States (Thaler & Sunstein 2008; Sunstein 2014). If PP homes in on a more accurate model of the thinking process, then we ought to use it, rather than (or as a complement to?) the dual systems model as a basis for policy making. Clark’s interpretation of PP suggests that we have a highly flexible range of cognitive systems, not limited to Kahneman’s two.
For example, one application of Kahneman’s model might involve the installation of environmental elements meant to appeal to the fast thinking system, to “nudge” agents towards making decisions in their best interest. If Clark is correct, we might consider even more sophisticated environmental features that have the goal of helping agents to deploy their range of cognitive strategies more efficiently. Clark’s ideas of context sensitivity and designer environments are both relevant here. As a society we may wish somehow to create environments and contexts that take advantage of the large repertoire of cognitive strategies available to us, according to Clark’s version of PP (see Levy 2012, for example).
The final topic I’d like to mention in this section is what is best described in general terms as social interaction. I mean to indicate a number of related topics here, but the main issue is how PP might relate to the well-known philosophical topic of the way in which we understand and explain our behavior to one another. Recall, for instance, Donald Davidson’s (1963) claim that our explanation of our behavior in terms of reasons is a kind of causal explanation—reasons as causes. On his influential view, the connection between reason and actions is a causal connection. In contrast, recall Paul Churchland’s envisioning of the golden age of psychology in which we dispose of folk psychological reason-giving in favor of more precise neurophysiological explanations of behavior (1981). According to Churchland’s radical alternative, the causes of actions are not reasons as expressed using natural language. Instead, our actions are caused by patterns of neurons firing, patterns that can be described using mathematical tools such as a multidimensional state space. In opposition to Churchland’s grand vision, we have Jerry Fodor’s claim that the realization of such a vision would be “the greatest intellectual catastrophe in the history of our species” (1987, p. xii). Is PP the beginning of Churchland’s grand vision coming to pass? Is a great intellectual catastrophe looming?
On one hand, PP seems like an obvious departure from folk psychology: Try explaining your X-ing to someone by claiming that you X-ed in order to minimize prediction error! One big issue here will be the way in which we think about agency itself. It seems mistaken to say that minimizing prediction error is something done by an agent. Such a process seems to be better described as occurring sub-personally. On the other hand, it is not inconceivable that propositional attitudes can capture the dynamics of prediction error minimization on a suitably coarse-grained level, perhaps along the lines suggested using symbolic dynamics (Dale & Spivey 2005; Atmanspacher & beim Graben 2007; Spivey 2007, Ch. 10). I suggest that these fascinating issues warrant further investigation. In particular, further investigation ought to incorporate Clark’s ideas of maximal context sensitivity and the importance of designer environments.
The way in which we understand each other’s behavior is also directly relevant for moral responsibility. Following Peter Strawson’s seminal “Freedom and Resentment” (1962), philosophers have started thinking about moral responsibility in terms of our reactions to one another, reactions that involve holding each other accountable. On one influential view, we hold each other accountable when our actions issue from our own reasons-responsive mechanisms (Fischer & Ravizza 1998). On a more recent proposal, holding each other accountable is best modeled as a kind of conversation (McKenna 2012). These proposals depend, in important ways, on assumptions about human psychology. In particular, they depend on our practice of giving reasons for behavior. As PP suggests a new fundamental underlying principle of behavior, our practices of holding each other accountable may be approached from a new perspective. The new challenge in this area will be to reconcile (if possible) the practice of giving reasons, on one hand, with PP’s account of behavior in terms of error minimization on the other.