Philosophy@Birmingham blog. This month, I am introducing chapter 10 of Jakob Hohwy's The Predictive Mind (OUP, 2013).
Chapter 10 - Perceptual Unity in Action
Presented by Lisa Bortolotti
In this chapter, Hohwy attempts to provide answers to two key questions about perceptual unity by relying on the prediction error minimisation (PEM) model. This is part of the project of showing that the model can account for all the core characteristics of conscious experience, such as having a first-person perspective, being cognitively penetrable, being subject to illusions, etc.
(1) Why do the elements of conscious unity hang together?
(2) Why is there one perceptual state subsuming all other ones?
Tim Bayne (2010, page 75) endorses the "unity thesis", i.e. the idea that conscious states of any subject of experience at any one point in time will occur as the components of a single total phenomenal state, a unitary phenomenal field. Hohwy agrees with Bayne that the unity thesis is true and claims that it is an important (but contingent) feature of consciousness. His task is to explain the unity thesis (as it applies to conscious perception) via the PEM mechanism.
In PEM we switch from seeing a coffee shop and feeling thirsty to making a decision about whether to buy a coffee in the shop or go home and have a glass of water (this is Hohwy's example of a switch from perceptual to active inference as described in previous chapters of the book). The switch involves making many predictions on the basis of different hypotheses and evaluating them in the light of contextually relevant considerations (e.g., money, daily intake of caffeine, time) in order to make a decision and act. What we do as agents who are limited to do one thing at a time is to select one hypothesis as "the best prediction error minimizer". This delivers one perceptual field, and no more.
To answer questions (1) and (2), in Hohwy's words, "perception is unified because it is based on hypotheses in a causally structured hierarchical model". Creatures like us can use only one hypothesis at the time to "sample the world" and make decisions about what to do. Building bridges with the work of Susan Hurley (1998, page 3) on the link between consciousness and action, Hohwy claims that "there would be no need for unity if there were no agency". That said, he rejects Hurley's externalism about perception. Causes of sensory input need to be inferred as they are to some extent hidden.
I found the picture sketched by Hohwy in this chapter both clear and attractive. I would be interested in examples where problems of agency (such as disruptions in the capacity to choose one option over another or in the capacity to act consistently) could be explained as problems of "perceptual disunity". If the transition from perceptual to active inference were somehow compromised and, say, two hypotheses were seen as prediction error minimisers, maybe not simultaneously but in very close temporal succession, then would paralysis of action or inconsistent action patterns result? Could this maybe explain phenomena such as "double-bookkeeping" in delusions?