The predictive mind
January 16th, 2025

Preface

My work on this book was made possible through invaluable research support from the Australian Research Council and from Monash University.

I am grateful to many researchers from around the world for inspiration, fruitful discussion, and generous comments.

My colleagues at Monash have influenced me, worked with me, and trained me in science; thanks in particular to my co-authors Bryan Paton, Colin Palmer, and Peter Enticott, to Steve Miller and Trung Ngo for including me in many projects and discussions, and to Naotsugu Tsuchiya, Anastasia Gorbunova, Mark Symmons, George van Doorn, Andrew Paplinski, Lennart Gustavsson, and Tamas Jantvik. Thanks also to my colleagues in philosophy, many of whom have repeatedly been roped in to do pilot studies, and to our participants and the patients who have endured many hours of rubber-hand illusion tapping in the lab.

Andreas Roepstorff's group in Aarhus are expert navigators in blue ocean research, and in many ways initiated and enabled my interest in this field. In addition to Andreas, Josh Skewes deserves thanks for many hours of discus- sion. Chris and Uta Frith, also sometimes at Aarhus, remain great, generous influences; they are the paradigm of the open-minded academic, especially when data are in the offing.

I am fortunate to have friends in philosophy and neuroscience who are prepared to endure lengthy discussions on predictive coding and the brain. Tim Bayne very early on encouraged me to push on with the book, and he read and commented extensively on the manuscript at various stages; I am ex- tremely grateful for his academic generosity. Thomas Metzinger likewise went well beyond the call of duty and offered generous comments on a draft of the book; I also greatly benefited from many discussions with Thomas' group of colleagues and students while visiting Mainz. During a week at University of Tokyo's Centre for Philosophy I enjoyed very valuable discussions on the book with Yukihiro Nobuhara and his colleagues and students. I have benefited greatly from many stimulating and encouraging discussions and comments on my writings and ideas from Andy Clark. Ned Block offered fruitful and needed resistance to parts of the story. Tim Lane and Yeh Su-Ling and colleagues from Taipei generously discussed many aspects of the book with me. I have had fruitful discussions with Floris de Lange, Sid Kouider, and Lars Muckli. Anonymous reviewers from the Press offered a host of insightful comments and criticisms.

I am especially grateful to Karl Friston whose work in so many ways has inspired the book. On numerous occasions, Karl has patiently offered feedback on my work. He read and commented extensively on every chapter of this book, he has endured long-haul flights to participate in interdisciplinary workshops, and has in many ways contributed to my work and furthered my understanding of the hypothesis-testing brain. It is very encouraging to experience the open-mindedness with which Karl approaches my attempts to translate the framework into philosophy, even as much of the mathematical rigour and detail is lost in translation. I of course remain responsible for any shortcomings.

The book is dedicated to my family: Linda Barclay, for encouraging me to write it, for predicting my errors, and for being with me; and Asker and Lewey, for being terrific rubber-hand guinea pigs and neurodevelopmental inspirations.

                                                      **Introduction** 

A new theory is taking hold in neuroscience. The theory is increasingly being used to interpret and drive experimental and theoretical studies, and it is finding its way into many other domains of research on the mind. It is the theory that the brain is a sophisticated hypothesis-testing mechanism, which is constantly involved in minimizing the error of its predictions of the sensory input it receives from the world. This mechanism is meant to explain percep- tion and action and everything mental in between. It is an attractive theory because powerful theoretical arguments support it. It is also attractive because more and more empirical evidence is beginning to point in its favour. It has enormous unifying power and yet it can explain in detail too.

This book explores this theory. It explains how the theory works and how it applies; it sets out why the theory is attractive; and it shows why and how the central ideas behind the theory profoundly change how we should conceive of perception, action, attention, and other central aspects of the mind.

                                                     **THE ARGUMENT** 

I am interested in the mind and its ability to perceive the world. I want to know how we manage to make sense of the manifold of sensory input that hits the senses, what happens when we get it wrong, what shapes our phenomenology, and what this tells us about the nature of the mind. It is these questions I seek to answer by appeal to the idea that the brain minimizes its prediction error.

My overall argument in this book has three strands. The first strand is that this idea explains not just that we perceive but how we perceive: the idea applies directly to key aspects of the phenomenology of perception. Moreover, it is only this idea that is needed to explain these aspects of perception. The second strand in my argument is that this idea is attractive because it combines a compelling theoretical function with a simple mechanical implementation. Moreover, this basic combination is of the utmost simplicity, yet has potential to be applied in very nuanced ways. The third strand of the argument is that we can learn something new from applying this idea to the matters of the mind: we learn something new about the mechanics of perception, and about how different aspects of perception belong together, and we learn something new about our place in nature as perceiving and acting creatures. The overall picture I arrive at from considering the theory is that the mind arises in, and is shaped by, prediction. This translates into a number of interesting, specific aspects of mind:

Perception is more actively engaged in making sense of the world than is commonly thought. And yet it is characterized by curious passivity. Our perceptual relation to the world is robustly guided by the offerings of the sensory input. And yet the relation is indirect and marked by a somewhat disconcerting fragility. The sensory input to the brain does not shape perception directly: sensory input is better and more perplexingly characterized as feedback to the queries issued by the brain.

Our expectations drive what we perceive and how we integrate the perceived aspects of the world, but the world puts limits on what our expectations can get away with. By testing hypotheses we get the world right, but this depends on optimizing a rich tapestry of statistical processes where small deviances seem able to send us into mental disorder. The mind is as much a courtroom as a hypothesis-tester.

Perception, action, and attention are but three different ways of doing the very same thing. All three ways be must be balanced carefully with each other in order to get the world right. The unity of conscious perception, the nature of the self, and our knowledge of our private mental world is at heart grounded in our attempts to optimize predictions about our ongoing sensory input.

More fundamentally still, the content of our perceptual states is ultimately grounded not in what we do or think but in who we are. Our experience of the world and our interactions with it, as well as our experience of ourselves and our actions, is both robustly anchored in the world and precariously hidden behind the veil of sensory input. We are but cogs in a causally structured world, eddies in the flow of information.

The theory promises not only to radically reconceptualize who we are and how aspects of our mental lives fit into the world. It unifies these themes under one idea: we minimize the error between the hypotheses generated on the basis of our model of the world and the sensory deliverances coming from the world. A single type of mechanism, reiterated throughout the brain, manages everything. The mechanism uses an assortment of standard statistical tools to minimize error and in doing so gives rise to perception, action, and attention, and explains puzzling aspects of these phenomena. Though the description of the mechanism is statis- tical it is just a causal neuronal mechanism and the theory therefore sits well with a reductionist, materialist view of the mind.

A theory with this kind of explanatory promise is extremely exciting. This excitement motivates the book. The message is that the theory delivers on the promise, and that it lets us see the mind in new light.

I am confident that many other aspects of this approach to the brain and the mind can and will be explored. This book by no means exhausts the impact of this kind of approach to life and mind. I focus on key issues in perception but largely leave out higher cognitive phenomena such as thought, imagery, language, social cognition, and decision-making. I also mostly leave aside broader issues about the relation of the theory to sociology, biology, evolution- ary theory, ecology, and fundamental physics. This still leaves plenty of work to do in this book.

                                                                  **PLAN** 

The book has three parts. Part I relies on the work of researchers in neurosci- ence and computational theory, in particular that of Karl Friston and his large group of collaborators. In a series of chapters the prediction error minimiza- tion mechanism is motivated, described, and explained. We start with a very simple Bayesian conception of perception and end with a core mechanism that makes Bayesian inference sensitive to statistical estimation of states of the world as well as their precisions, while making room for context-sensitivity and model complexity. The overall view is attractive in part because it appeals to just this one mechanism-this is thus a very ambitious unificatory project.

This area of research is mathematically heavy, and this is indeed part of the reason for its increasing influence: the mathematical equations provide formal rigour and the possibility of quantifiable predictions. However, my exposition is done with a minimum of technical, formal detail. I appeal to and explain very general Bayesian and statistical ideas. This neglects mathematical beauty but will make the discussion accessible and more easy to apply to conceptual and empirical puzzles in cognitive science and philosophy.

My main concern is to bring out the key elements of the prediction error minimization mechanism, in particular, the way prediction error arises and is minimized, how expectations for the precision on prediction error are pro- cessed, how complexity and context-dependence factor in, and how action is an integral part of the mechanism. Furthermore, I set out how this mechanism is re-iterated hierarchically throughout the brain. These are the elements needed to explain everything else and they can be fairly conveyed without too much formal detail. I do provide a brief primer of Bayes' rule at the end of Chapter 1, and describe some rudimentary formal detail in the notes to Chapter 2. I also at times provide some very minimal formal expressions, which mainly serve as reminders for how the more complex points relate to simpler Bayesian expressions; these more formal elements are not essential to

Trang 3

Subscribe to Jim
Receive the latest updates directly to your inbox.
Mint this entry as an NFT to add it to your collection.
Verification
This entry has been permanently stored onchain and signed by its creator.
More from Jim

Skeleton

Skeleton

Skeleton