I’ve mentioned Predictive Processing (PP) and the Free Energy Principle (FEP) before in a post, and this was a two-day event gathering philosophers and cognitive scientists (and a few interlopers) to consider the state of play. Although I have set out on the voyage through the poems of John Skelton that I mentioned in my previous post, it seems timely to say a few things about what I took away from it (before I forget).
The basic point of PP is that a great deal of our cognition is based on predictions which are then modified by feedback from the world. Perception, for example, is efficient but sometimes flawed because we are always constructing what we expect to see in advance, and modifying these constructions as we pick up new information. Social situations, for another example, involve constant ongoing predictions about the behaviour of others, and these too are constantly changing when there are prediction errors.
What FEP adds is an underlying pattern. Systems interacting with their environments seek to minimise the amount of ‘free energy’: for a mind in the world, this means aiming for the smallest difference between predictions and evidence, between the internal model and the sensory input. Most pithily, this could be summed up as an effort to minimise surprise.
These two things are becoming very influential in the science and philosophy of the mind (as was demonstrated by the conference). Experimental design and abstract models are both working with PP, using it as a premise and testing its possibilities.
The thing that really struck me overall was the range of perspectives currently engaging with Predictive Processing and Free Energy. At the conference there were papers tackling this as a philosophical matter in the broadest way, and there were others approaching it as a practical means of understanding psychiatric disorders.
In the first of these categories, the last session brought together Karl Friston, the originator of the Free Energy Principle (outstanding website here), and Jakob Hohwy, author of The Predictive Mind (Oxford, 2014; another high-quality website here), which I’d recommend alongside Andy Clark’s book discussed in this post. Hohwy went so far as to say (or so it seemed to me, somewhat light-headed in the rarefied air of philosophy) that we should see the avoidance of surprise as a quality of Things That Exist. Friston’s breakneck paper reconciled many things with many other things by means of equations, and again communicated the apparently limitless reach of this way of thinking.
Friston himself noted the large number of papers rethinking various psychological disorders, spectrums, and syndromes in the light of Predictive Processing. There was a terrific paper from Katherina Schmack (see here) about the ways we might understand psychotic characteristics in the light of the Predictive Processing Model. Is this a matter of faulty predictions, too strong, or too weak? Is it about a failure to deal with noise in the system, evidence from the world that shouldn’t affect predictions? Is it about problems in the re-evaluation, the feedback from input to prediction? Is it about how you deal with surprises, over-rating them or misjudging them? It was a very interesting set of questions, and I’ll definitely be thinking more about them.
I couldn’t attend the whole event, and one paper I missed turned out to make some interesting links with art. (I was kept updated by Micah Allen’s Twitter feed, @neuroconscience, which is generally a good source.) There is an online paper I tracked down by the speaker, Anil Seth (many interesting strings to his bow, see website), which you can read via here. I hope it’s not contraband.
Seth’s point is that there is an overlap between Predictive Processing and ideas in the phenomenological theory of art. In particular, he brings in E.H. Gombrich’s idea of ‘the beholder’s share’, a phrase used in Art and Illusion (Princeton, 1961) to capture the way that the viewer of art does things to form their experience of the picture — and thus to make the picture itself. For Gombrich, then, as for many at the conference, perception is inference.
The essay linked above turns to a number of twentieth-century artworks in the Impressionist, Expressionist, and Cubist traditions, to show how art and cognitive neuroscience ‘can work together to elucidate generative contributions to human perception and phenomenology’. It’s intriguing to see famous paintings set alongside photos of everyday scenes manipulated according to principles (of, say, peripheral perception) predicted by cognitive science. This essay was a real find.
I got a lot out of a nice paper by Sam Wilkinson, about what we mean by ‘explanatory power’. The point was that defining what made a good explanation is not a straightforward matter. It depends on identifying the right question and the appropriate framework in which to answer it (and that means catering to the situation of asking as well as the intellectual domain). It also depends on achieving a useful level of precision, identifying not just why X but why X and not why, etc. Predictive Processing, it was argued, had an interesting sort of explanatory power. I thought for a while about what the explanatory power of literature might be, when it comes to knowing things about your brain. I reckon, generally, that it’s true that a key challenge lies in identifying the questions to be asked and answered, ones that might reach outside the narrow disciplinary world, and that’s something I am not especially good at.
More particularly, I was struck by Wilkinson citing what seems to be an established distinction (though new to me) between the likeliness of an explanation and the loveliness of an explanation: explanatory power, for better or worse, depends on the latter as well as the former. I think there is a possible third term related to things I’ve said on this blog before, liveliness, that is helping me (just a little) think of how to characterise the contribution literary criticism might make to philosophical and cognitive discussions. This ‘liveliness’, as well as being a sweet wordplay on loveliness and likeliness, incorporates the classical principles of mimesis (representing reality in a believable way) and enargeia (creating vivid images), and could be seen as a rather useful characteristic when we’re trying to get across explanations about how the mind works.
This has become a rather long post, and I have more to say. So it will be in the next post that I try to suggest some ways in which literary criticism could pick up on ideas about Predictive Processing, the Free Energy Principle, and all!