You are on page 1of 5

Australasian Journal of Philosophy

ISSN: 0004-8402 (Print) 1471-6828 (Online) Journal homepage: http://www.tandfonline.com/loi/rajp20

Surfing Uncertainty: Prediction, Action and the


Embodied Mind, by Andy Clark

Daniel D. Hutto

To cite this article: Daniel D. Hutto (2017): Surfing Uncertainty: Prediction, Action
and the Embodied Mind, by Andy Clark, Australasian Journal of Philosophy, DOI:
10.1080/00048402.2016.1274769

To link to this article: http://dx.doi.org/10.1080/00048402.2016.1274769

Published online: 10 Jan 2017.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=rajp20

Download by: [FU Berlin] Date: 11 January 2017, At: 00:00


AUSTRALASIAN JOURNAL OF PHILOSOPHY, 2017

BOOK REVIEW

Clark, Andy, Surfing Uncertainty: Prediction, Action and the Embodied Mind,
New York: Oxford University Press, 2016, pp. xviii C 401, £19.99 (hardback).

Brains do not sit back and receive information from the world, form truth evaluable
representations of it, and only then work out and implement action plans. Instead, tire-
lessly and proactively, our brains are forever trying to look ahead in order to ensure
that we have an adequate practical grip on the world in the here and now. Focused pri-
marily on action and intervention, their basic work is to make the best possible predic-
tions about what the world is throwing at us. The job of brains is to aid the organisms
that they inhabit, in ways that are sensitive to the regularities of the situations that
those organisms inhabit. Brains achieve this by driving activity that is dynamically
and interactively bound up with, and sensitive to, the causal structure of the world on
multiple spatial and temporal scales. Understanding brains as doing fundamentally pre-
dictive work of this sort—as ‘action oriented engagement machines’ [300]—is perfectly
in tune with the recent trends of conceiving of cognition as embodied, ecologically
situated, extended, and enculturated. These are the main messages of Surfing Uncertainty.
The book’s ten chapters—each sub-dividing into substantial sub-sections (some-
times as many as eighteen)—cover a vast terrain and touch on many diverse topics of
interest to empirically informed philosophy of mind. The topics range from discussions
of the best explanation of binocular rivalry; revising our thinking about motor control;
learning implementation and design lessons in robotics; getting a grip on mental time
travel; rethinking the role of mirror neurons in understanding action; explaining
the signature-response patterns in schizophrenia; and much more. All of these
disparate explorations are part of a single-minded effort to establish the wide-ranging
relevance and power of the predictive processing framework. In some cases, this takes
the form of showing that framework’s empirical superiority to rivals. In others, the
aim is to show its great promise and potential, even if only by touching tantalizingly on
specific topics.
It would be impossible to give detailed review of all of the book’s many rich investi-
gations in a short space, but it is possible to capture and evaluate some of the large-scale
action of the book by focusing on its three parts.
Part I (‘The Power of Prediction’), which comprises the first five chapters, presents
the rudiments of the predictive processing framework for thinking about mind and
cognition. It seeks to demonstrate how that framework offers ‘an attractive “cognitive
package deal” in which perception, understanding, dreaming, memory and imagination
may all emerge as variant expressions of the same underlying mechanistic ploy’ ([107,
137]; see also [xvi]). In seeking to integrate these and other phenomena, Clark is utterly
clear that his aim is not to provide ‘yet another “new science of the mind” but some-
thing potentially rather better’ [10]. That ‘something better’ is a way of understanding
cognition that serves as [ibid.]

a meeting point for the best of many previous approaches, combining elements from work in
connectionism and artificial neural networks, contemporary cognitive and computational
2 BOOK REVIEW

neuroscience, Bayesian approaches to dealing with evidence and uncertainty, robotics, self-orga-
nization, and the study of the embodied, environmentally situated mind.

The exciting conceit of the predictive processing approach is that it conceives of minds
in essentially anticipatory fashion, offering a dramatic reversal of the traditional—
sense-model-act—way of understanding minds. Crucially, on the new view, the brain is
always poised and prepared to act, often readying multiple ways of doing so at once.
Like an impatient New Yorker, ‘the human brain does not wait’ [179, 145]. This goes
against the traditional idea that minds must first collect data from the world and only
then build models and representations of it so as to act. In a dramatic reversal of that
familiar formula, the big idea of predictive processing, as illustrated by a rethinking of
what it is to perceive, is that we are always trying to be cognitively ahead of the curve—
indeed, that ‘we see the world by … guessing the world’ [5].
According to Surfing Uncertainty, the core business of cognition is that of making
delicately balanced, pro-active, probabilistic, Bayesian predictions about the most likely
sensory perturbations; it is not a matter of constructing internal models of the world
that are built upon passively received information furnished by our feature detecting
senses. Throughout the book, Clark exposes the problems, limitations, and empirical
inadequacy of this rival view. Its chief failing, he holds, is its inability to explain how we
can ‘respond fluently to unfolding—and potentially rapidly changing—situations’
[176].
Abandoning the ‘last vestiges of the input-output model’ [139], Clark argues that the
true business of brains—their ceaseless cascade of multi-level, multi-layered, cortical
processing—is all part of a singular effort to predict sensory deliverances [146, 166,
167]. Prediction error occurs when there is a mismatch between what brains predict
and what is supplied to them by the senses. The brain’s aim is to sensitively minimize
the divergences between what it anticipates and actual sensory deliverances. This can
be achieved either by making better guesses or by acting in ways that will generate
more fitting sensory deliverances. Either strategy can be used to reduce prediction error
and uncertainty. Nor is the brain a blunt instrument. Toggling attention, in line with
background knowledge about specific contexts of engagement, is part of a meta-cogni-
tive strategy that can be brought to bear in trying to appropriately adjust the precision
weighting given to either sensory deliverances or top-down expectations, case by case
[62, 64].
Part II (‘Embodying Prediction’) seeks to make good on Clark’s opening claim that
the predictive brain hypothesis is ‘the perfect neuro-computational partner for recent
work on the embodied mind’ [1, 108]; and that it offers a new and explanatorily power-
ful take that helps to ground the ‘enactivist vision’ [125]. That an account of predictive
brains and embodied and enactive approaches to mind complement and connect with
one another should be unsurprising, thinks Clark, because all of these approaches
understand cognition as fundamentally active, world involving, and self-organizing in
nature.
There are special payoffs of locating theorizing about predictive brains within the
space of embodied, enactive, and ecological approaches to mind. Doing so exposes the
flaws in the assumptions that give rise to the infamous ‘dark room’ objection. That
objection holds that if prediction error minimization were the only principle for under-
standing the work of brains, then the optimal strategy for achieving the desired result
would be to locate the most stable environment—viz. a dark room—and stay in it,
AUSTRALASIAN JOURNAL OF PHILOSOPHY 3

engaging the world as little as possible. Yet, as Clark makes clear, while the reduction of
uncertainty is central to the predictive processing activity of brains, it should not be
regarded as ‘the’ foundational, ultimate explanation of all of the adaptive behaviour of
organisms; it is neither the sole nor the most basic factor that drives such behaviour
[263ff.].
Clark also makes the links with embodied, enactive, and ecological cognition when
he defuses the idea the predictive brain hypothesis necessarily entails global scepticism.
He holds that the latter, mistaken view—which takes predictive brains to be epistemi-
cally secluded—is promoted by taking too seriously and literally the idea that the brain
makes ‘predictions’, advances ‘hypotheses’, and trades in ‘inferences’; whereas, read by
Clark’s lights, the predictive processing account firmly opposes ‘intellectualist stories
that … allow the cognizer to throw away the world’ [191]. Over and against this ten-
dency, Clark highlights that on his account the job of brains is not to describe the
world, but instead to find ‘neural activity patterns that most successfully accommodate
current sensory states by means of world engaging action’ [192].
In Part III (‘Scaffolding Predictions’), Clark makes connections with his famous
extended mind thesis and explains how the anticipatory activity of embodied, predic-
tive brains is also shaped over slower timescales, by being situated in culturally crafted,
designer, environments of our own making [138]. Here, again, Clark makes strong con-
nections with central themes in the enactivist literature about how we come to enact
our worlds. Yet, despite recognizing these important links with enactivism, in the penu-
ltimate chapter Clark draws attention to a major issue on which he and other enacti-
vists will continue to disagree. The bone of contention is that Clark’s version of
predictive processing is steadfastly committed to the existence of pragmatic representa-
tions and inner generative models. Everywhere, he speaks of the brain’s activity as
drawing on stored knowledge [6, 21, 79, 166] or, sometimes, the brain’s ‘probabilistic
know-how encoded in complex multi-level generative models’ [296]. Given these lean-
ings, it becomes clear why Clark has difficulty in seeing how one might tell a predictive
processing story about the brain ‘in entirely non-representational terms’ [293].
How seriously should we take the predictive processing framework? Is it our best
current theory of the brain? Clark makes a sustained and convincing jointly theoretical
and empirical case that shows the great promise, and the explanatory and integrative
virtues, of the approach. But even he is careful to hedge his bets. In the book’s closing
chapter, he is especially open about the amount of work that remains to be done. He
emphasizes that predictive processing is a burgeoning research program that is still
very much in its infancy [298]. As a framework, predictive processing is not aiming to
provide, nor should it be evaluated as, a precise and well established mature theory.
Readers who may feel cheated that Clark only touches lightly on many topics must bal-
ance their disappointment with the realization that, at this stage, he is aware that he
often only offers ‘a smattering of hints and suggestions’ [299].
Clark is clear that we must distinguish between two stories that might be told about
predictive processing, and that his primary concern is to defend only the first of those
stories. The first story presents a general, and ‘extremely broad’, vision according to
which the brain’s role in cognition is that of ‘multi-level probabilistic prediction’ [10].
The second, and more specific, story attempts to give more precise details about how
the first story might get told in light of current developments in neuroscience. That
story is still unfolding.
4 BOOK REVIEW

Regardless of whether, in the end, we accept either story, there is one thing about
which we can be certain: in the years to come, Surfing Uncertainty will be a much dis-
cussed and seminal work in the field of the philosophy of cognitive science. Given all
that we already know, it is a safe bet to expect that Clark’s book will inspire and launch
a thousand new research projects, as well as many fierce challenges and fresh intellec-
tual investigations.

Daniel D. Hutto
University of Wollongong

© 2017 Daniel D. Hutto


http://dx.doi.org/10.1080/00048402.2016.1274769

You might also like