12744606 an epistemology of noise by cecile malaspina ray brassier z lib.org

CECILE jpeg) An Epistemology of Noise Also Available from Bloomsbury Noise Matters, Greg Hainge Epistemology, Archaeology, Ethics, edited by Sebastian Luft and Pol Vandevelde Speculative Realism, Peter Gratton Genealogies of Speculation, edited by Armen Avanessian and Suhail Malik Metanoia, Armen Avanessian and Anke Hennig Introduction to New Realism, Maurizio Ferraris Exceptional Technologies, Dominic Smith Cecile Malaspina Foreword by Ray Brassier BLOOMSBURY ACADEMIC LONDON • NEW YORK • OXFORD • NEW DELHI • SYDNEY BLOOMSBURY ACADEMIC Bloomsbury Publishing Plc 50 Bedford Square, London, WC1B 3DP, UK BLOOMSBURY, BLOOMSBURY ACADEMIC and the Diana logo are trademarks of Bloomsbury Publishing Plc

First published in Great Britain 2018 Copyright © Cecile Malaspina, 2018 Cecile Malaspina has asserted her right under the Copyright, Designs and Patents Act, 1988, to be identified as Author of this work. Cover design by Catherine Wood Cover image: 'inter esse', Berlin 1985–87 © Maria Sewcz All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage or retrieval system, without prior permission in writing from the publishers. Bloomsbury Publishing Plc does not have any control over, or responsibility for, any third-party websites referred to or in this book.

All internet addresses given in this book were correct at the time of going to press. The author and publisher regret any inconvenience caused if addresses have changed or sites have ceased to exist, but can accept no responsibility for any such changes. A catalogue record for this book is available from the British Library. A catalog record for this book is available from the Library of Congress. ISBN: HB: 978-1-3500-1178-6 PB: 978-1-4411-6532-9 ePDF: 978-1-3500-1179-3 ePub: 978-1-3500-1180-9 Typeset by Integra Software Services Pvt. Ltd. To find out more about our authors and books visit com and sign up for our newsletters.

To jpeg) **THEY DO NOT AGREE** Goya Series: They Do Not Agree, 1997 courtesy of John Baldessari Contents Foreword x Acknowledgements xiv Note on Text xvi List of Abbreviations xvii Introduction 1 Part 1 Concepts: Information Entropy, Negentropy, Noise 13 I How to Draw the Line between Information and Noise 15 II Entropy as ‘Freedom of Choice’ 23 III Information Entropy and Physical Entropy 27 IV The Idea of ‘Potential Information’ 29 V Physical Concepts of Information and Informational Concepts of Physics 35 VI Information as Process Rather Than Content 39 VII To Think about Information as a Process of Individuation 43 VIII Redundancy and Necessity 51 IX Logic and Freedom of Choice 57 X Noise as Spurious Uncertainty 61 XI Negentropy 65 XII Complexity on the Basis of Noise 71 XIII The Astigmatism of Intuition 79 XIV The Path of Despair 85

Contents viii Part 2 Empirical Noise 91 I On the Transduction of the Concept of Noise 93 II Accidental Information, Predictable Noise 97 III Ready-Made Information 103 IV Cosmic Background Radiation 109 V Noise in the Gap between Narratives 115 VI Noise in Finance 119 VII Statistics: The Discipline of the Prince 133 VIII The Man without Qualities 139 IX Noise Abatement: The Dawn of Noise 143 X Noise Pollution 149 XI Toxic, Viral, Parasitic 155 Part 3 The ‘Mental State of Noise’ 165 I The Crossroads: Mathematical, Technical, Empirical and Subjective Noise 167 II Internal Chaos, Terror and Confusion 169 III The Vicious Whir of Sensations 179 IV Keat’s Negative Capability 181 V Closure to Noise and the Paradox of the Declining Life 187 VI The Catastophic Reaction to Noise 191 VII Anxiety 195 VIII Order 199 IX Control 203

Contents ix X The Helmsman Metaphor: Kybernetes 207 XI The Helmsman in Plato's Alcibiades Dialogue 213 Bibliography 219 Index 228 Foreword To turn noise from an object of thought into ‘a variable within the process of thought’: this is the goal of Cecile Malaspina’s philosophical investigation of noise – philosophical because it entails transforming noise from an empirical phenomenon into a condition for the possibility of empirical conceptualization. Taking Claude Shannon’s notion of ‘information entropy’ as her starting point, Malaspina shows how the phenomenon of noise harbours a profound philosophical paradox. Information entropy is a measure of the degree of uncertainty or ‘freedom of choice’ about the state of a message.

By aligning information with unpredictability, Shannon aligns it with uncertainty. But uncertainty implies ignorance. Thus the concept of information entropy entails this vexing consequence: if uncertainty indexes information then certainty indexes noise. But how could certainty, the apex of cognitive aspiration, be a symptom of lack of information? In contrast to Shannon’s twinning of information with entropy, Norbert Wiener’s characterization of information as the negation of entropy or negentropy seems intuitively plausible. Wiener sidesteps the troubling affinity between information and disorder by confirming our spontaneous identification of noise with disorder. Yet the curious reversibility between information and noise remains unaddressed.

Focusing on this reversibility, Malaspina shows how from its inception the concept of noise as the obverse of information rests upon an equivocation between order and disorder. This is not merely an equivocation but rather an essential ambiguity, one that is symptomatic of a latent contradiction in the concept of noise. Rather than seeking to expose it as a flaw, Malaspina sees in this contradiction the clue to a deeper truth about noise. Her approach is dialectical, and the contradictoriness of noise as a concept is the key to its reality as a phenomenon. Working through this contradiction, Malaspina patiently unravels the superficial oppositions of order and disorder, certainty and uncertainty, knowledge and ignorance in all the theoretical contexts where the distinction between noise and information has been deployed.

Her demonstration traverses information theory, cybernetics, thermodynamics, biology, psychiatry and sociology, drawing upon such diverse thinkers as Claude Shannon, Norbert Wiener, Gilbert Simondon, Michel Foucault Foreword xi and George Canguilhem. By engaging these sources, Malaspina achieves a philosophical optic that is genuinely transdisciplinary. (Her achievement in this regard is perfectly complemented by that of Inigo Wilkins, whose work resonates beautifully with her 1) This does not mean fashioning conceptual hybrids from disparate theoretical discourses in an eclectic and ultimately opportunistic fashion. Malaspina constructs a philosophical concept of noise by pinpointing the decisive fault lines in the workings of the various theoretical concepts whose functioning she carefully delineates.

But the dialectical cast of her thinking renders her stance constructive rather than deconstructive. It allows her to integrate concepts from the mathematical and natural sciences alongside those from social and cultural theory. This transdisciplinary remit makes it possible in turn to articulate the epistemology of noise with its ontology, which is to say that it encompasses both what noise is and how it relates to knowing. What is important, from Malaspina’s perspective, is that this does not so much subvert as politicize the conventional distinction between epistemology and ontology: The conceptualization of noise is thus no longer limited to the classical philosophical problem of determining what we can understand of the reality of noise ‘in itself ’ or even ‘for us’.

It is irreversibly contaminated by a political problem, which is the possibility of deliberate or accidental distortion also of our critical faculties through noise. 162) From this follows what is perhaps Malaspina’s most striking insight: Noise, beyond the reference to unwanted sound, thus reveals itself to be conceptually polymorphous because it has never been about types, classes or measures of phenomena that qualify noise as a particular type of disturbance, but about the relation between contingency and control. 203) For Malaspina, ‘noise’ is not just the name for the force scrambling the recognizable outlines of phenomena; it designates the anomaly from whence the distinction between sense and senselessness originates.

It is not merely a natural phenomenon or kind because its co-articulation with information is the consequence of an act of judgement, rather than the registration of a fact. Thus noise is a normative rather than a natural category, which is to say that it is made not given. The empirical discrimination of noise presupposes the normative establishment of its difference from information within a given disciplinary framework. But this difference – between control and contingency, determination and indetermination – follows from what Malaspina calls ‘a suspension in indecision’ or ‘unthinkable freedom of choice’ that is not of the order of structure or destruction.

Precisely because it Foreword xii cannot be naturalized, objectified, catalogued or classified, this anomalousness, from which judgement proceeds, can only be acknowledged by shifting from a perspective that treats noise as object to one that treats it as subject, or in Malaspina’s words, ‘the noise of cognition constituting itself, against the always looming crisis of its dissolution’ 173). Malaspina distinguishes two ‘subjects’ of noise here: the subject of cognition constituting itself and the experiential subject undergoing the legislated difference between information and noise. The subject of cognition enforces conditions of disciplinary regulation through which information is first identified.

The experiential subject is the locus for the psychological, cognitive and affective ramifications of the experience of noise, which Malaspina explores through a particularly inventive reading of Steven Sands and John Ratey’s seminal 1986 article, ‘The Concept of Noise’. Ultimately, Malaspina’s book is an epistemic intervention. Knowledge is ‘of ’ uncertainty in both the subjective and objective senses of the genitive. It is uncertainty that knows. The subject of cognition exercising the power of judgement requires that the normative preconditions of cognitive judgement determine the constitution of empirical facts. This is the grounding power of judgement.

But this grounding power is itself based on ungrounding as condition of normative grounding. This ungrounding is a consequence of the ineliminable role played by contingency in the exercise of determining judgement. This is why, for Malaspina, ‘noise is, like disorder, an inconceivable freedom of choice’ 187). Knowledge is not the reduction of uncertainty because it is constituted by it. Thus knowledge is neither solely predictive nor exhaustively fallible, no more than it is either verifiable or falsifiable. It is ‘of ’ uncertainty because it is rooted in this ‘inconceivable freedom of choice’. Malaspina uncovers the productive, form-generating powers of epistemic noise as the occluded source of cognition’s predictive capacities.

Cognitive invention is not rooted in the ‘negentropic’ negation of contingency; it proceeds from the ‘negation of the negation of contingency’ 183). Regulation is not the precondition for innovation; innovation gives rise to regulation through the lawless collapse of regularity. Invention, whether cognitive, aesthetic or political, is not the negation of disorder but the negation of its negation, which is to say, the negation of order. Malaspina achieves a properly dialectical resolution of the tension between the negative physical characterization of noise as form- destroying entropy and the positive aesthetic valorization of noise as form- generating novelty: radical transformation (whether cognitive, aesthetic or political) can only arise through the unconditioned judgement that affirms the irreconcilable tension between the destruction and generation of form.

This Foreword xiii judgement, constituting cognition against the backdrop of its dissolution, is a function of the noise that enables the process of thought. 1. Inigo Wilkins, *Irreversible Noise*, Falmouth: Urbanomic, forthcoming. My first expression of gratitude goes to Prof Alain Leplège and Dr Iain Hamilton Grant E.) for the freedom they granted me and for their unwavering trust in supervising the doctoral thesis on which this book is based. Prof Ray Brassier, Prof Emmanuel Picavet and Dr Matthieu Saladin are warmly thanked for their questions and insightful comments. Frankie Mace at Bloomsbury, as well as Deepakraj Murugaiyan have my heartfelt thanks for their support and inexhaustible attention.

I thank also the photographer Maria Sewcz for generously putting her work ‘inter esse’ at our disposition for the cover of this book and John Baldessari, for letting us use his ‘They do not agree’ as the frontispiece. Catherine Wood paid particular attention to the artists wishes in designing the cover. The Reverberations conference organized by Dr Benjamin Halligan, Dr Paul Hegarty and Dr Michael Goddard at Salford University in 2010 has been determining for the transdisciplinary perspective of this book. I am grateful for their editorial support going into the publication of my contribution to Reverberations, the Philosophy, Aesthetics and Politics of Noise with Continuum in 2012.

The intense exchange about noise and art with Dr Michael Schwab and all collaborators in the Data-Rush symposium organized in Vienna in 2016, especially Prof Mauricio Suarez, and also Dr Paulo de Assis and Tiziano Manca at the Orpheus Institute in Ghent have been immensely enriching. I am also especially grateful for a host of new references and ideas I owe to the generous suggestions of Prof Christian Walter and Prof Emmanuel Picavet at the Chaire Ethique et Finance, College d’Études Mondiales, Fondation Maison des Sciences de l’Homme, Institut des Sciences Juridiques et Philosophiques de la Sorbonne (UMR 8103).

I am deeply grateful also to Dr Anne Lefebvre, for years of collaboration and dialogue, for her invitation to speak at the Van Eyck Academy in Maastricht in 2012 and for her initiative, to which I owe the opportunity of testing early ideas at the transdisciplinary seminar at the Ecole Normale Superieure in Paris in 2010, whose organizers Prof Claude Debru, Prof Jean-Charles Darmon and Prof Frédéric Worms are also warmly thanked. Also the European Meeting for Research in Systems and Cybernetics (EMCSR) in Vienna in 2012, 2014 and Acknowledgements Acknowledgements xv 2016, as well as the Schelling Grundlagen Seminars at the Institute for Design Science, Munich, and have been important milestones.

Prof Rainer Zimmermann is warmly thanked, alongside Stefan Blachfellner, director of the Bertalanffy Center for the Study of Systems Science, as well as Dr Jose Maria Diaz Nafria and the BITrum consortium. Dr David Rousseau, editor at Systema: Connecting Matter, Life, Culture and Technology, has my gratitude for his editorial advice on the publication of my article on epistemological noise, which has enabled me to articulate one of the core ideas going into this book. Not least do I thank all those not mentioned here and not directly cited, whose thought has illuminated the questions I could tackle, but also those questions that motivated me and that remain in the undergrowth.

Amelie Mourgue-D’Algue, Sissi Taseva, my parents and brothers have indefatigably supported, if not freed me to work on this book. To my sons Federico, Paolo and Olivier I owe the greatest debt of gratitude, for enduring the period of research and writing and for encouraging me at critical times. To Andrea I owe the insouciance of beginning this work and the courage of finishing it. All quotations referencing texts with German or French titles are translated by myself. ILFI Simondon, Gilbert. 2005. L’individuation à La Lumière Des Notions de Forme et d’information. Grenoble: Editions me Millon.

METO Mode of Existence of Technical Objects, Gilbert Simondon, trans. Cecile Malaspina and John Rogove, Univocal Publishing, 2017 MTC Mathematical Theory of Communication, Claude Shannon and Warren Weaver, University of Illinois Press, 1964 NP The Normal and the Pathological, Georges Canguilhem, trans. Carolyn R. Fawcett and Robert S. Cohen, Zone Books, 1991 WHO World Health Organization, int/about/en List of Abbreviations. It has become commonplace to use the word noise, almost with inverted comas, in a host of contexts unrelated to sound, often in opposition to information. It is thus not the din of the trading floor that interests us when we talk about noise in finance, but the uncertainty related to random variations in the stock exchange.

Noise has become a concept intrinsic to the statistical analysis of the variability of data in almost every domain of empirical enquiry. Even acoustics can be argued to have fully emerged only during the 1950s, when noise could be represented as graphs of the frequencies and amplitudes of transitory signal changes over time (Castellengo 1994). That these two dimensions of the conceptualizations of noise, as sound and as random variation, speak to each other without being reducible to one another is what this book is about. This new, statistical meaning of noise is first and foremost the expression of one of the most profound methodological transformations of the modern sciences.

Predating cybernetics and information theory, the source of today’s understanding of noise as inextricably linked to variation or even error goes back to the origin of the calculus of probability in games of chance, notably in the work of Pascal and Bernoulli’s law of large numbers – also called the ‘law of possibilities’. Interpreted by Laplace as an adequate representation of ‘errors in measurement in astronomy’, the ‘law of possibilities’ has subsequently been called the ‘law of errors’ (Desrosieres 2006). However, the definition of statistical noise we inherit from the ‘law of errors’ must not obscure the fact that in modern statistics precision itself has become a question intimately linked to noise: Precision is a measure of random noise.

(S. Smith 2002, 34) The special sense of the word noise thus implies both a methodological transformation and a new scientific status of the notions of uncertainty, probability, and error in relation to statistical averages. Thus enriched, the subsequent definition of noise in cybernetics and information theory has come Introduction An Epistemology of Noise 2 to retrospectively encompass also the concept of physical entropy, and more generally of uncertainty, statistical variation and error. No longer considered only as a factor of disturbance, detrimental to information like ‘static noise’ in the channel of communication, the evolving concept of noise also becomes constitutive of new forms of knowledge and of new ways of understanding organization.

This new connotation of noise has entered ordinary language as a side effect of the ‘information paradigm’ we have inherited from information theory and cybernetics (Malaspina 2012a; Morange 2006). Although Raymond Ruyer noted already in 1954 that cybernetics had failed to impose itself as a transdisciplinary scientific paradigm, the notion of noise has nevertheless steadily gained prominence over the past decades as a notion bridging disciplines: not only in relation to computer science and even complexity theory, but across the natural and human sciences and the arts, ideas of ‘order from noise’ or ‘complexity on the basis of noise’ are steadily rising to greater prominence (Atlan 1979; Mersch 2013; Nunes 2012; Ruyer 1954).

The term inforgs has since been coined to emphasize the idea that we no longer inhabit only an ecosphere, but also an infosphere (Floridi 2002). Yet recent developments appear to suggest that, far from the dawn of an Information Enlightenment, we seem to sleepwalk into an era of noise: levels of stress and depression are rising as we pine under both noise pollution and information overload (Bawden and Robinson 2009; Berglund and Lindvall 1995). Even intelligence services suffer from too much information. Big data means that noise can be harvested through data mining, but vast amounts of data once more become noise as soon as we lack pertinent criteria to transform them into information (Watkins 2011, 31).

A curious reversibility of information and noise thus becomes apparent: too much information, and also the repetition of the same information ad nauseam, becomes noise, whereas information that is radically new falls on deaf ears when context and criteria of pertinence are lacking to adequately distinguish information from noise. Despite the ever more apparent complexity of the relation between information and noise, the latter is often taken for granted as the mere opposite of information, based on the intuitive analogy with acoustic noise disrupting communication. What risks being overlooked in this simplistic opposition between information and noise is a palimpsest, a rich layering of intuitive notions of the still and the perturbed, the clear and the turbid (from Latin turba: crowd), opaque or confused.

This opposition is also rich in ideas that Introduction 3 have a proud history at their heels – such as order and disorder, work and futility (the latter indicating a leaking, untrustworthy vessel in medieval alchemy) (Watkins 2011, 31) – and rich also in mathematically formalized concepts, like Ludwig Boltzmann’s formalization of statistical entropy. In this palimpsest of concepts, notions and ideas, noise always appears to occupy the negative place of a dichotomy, be it in that of order and disorder, of physical work and the dispersion of energy in the state of entropy, or of the norm and the abnormal.

In other words, noise is at best associated with the absence of order, of work or of the norm – be it the statistical, moral or aesthetic norm – and at worst, noise is identified as a threat to the norm and subversive of work and order: a perturbation, a loss of energy available for work, a parasite. Noise is thus a word that implicitly plays on the whole register of notion, idea and concept and does so by mobilizing linguistic, historical, sociopolitical and not least of all epistemological registers. If we are to understand the new fortunes of the previously reviled and now revisited idea of noise – from physics to information theory and cybernetics and beyond – then we must not only disentangle these notions, ideas and concepts, but also analyse the subtle ways in which new concepts of information have rewired our conceptions of noise: starting with the concepts of ‘information entropy’ and negentropy, which is what the first part of this book sets out to do, before looking at some cases of empirical noise in Part II (from the discovery of cosmic background radiation to noise pollution and the historical origin of Statistik as the nomenclature of knowledge necessary for the sovereign) and finally at the role of noise in the process of cognition itself, by focusing (in Part III) on the idea of ‘the mental state of noise’, developed in 1986 by S.

Sands and J. Ratey to describe an internally experienced state of crowding and confusion created by a variety of stimuli, the quantity, intensity and unpredictability of which make it difficult for individuals so afflicted to tolerate and organize their experience. Attempts to do so may only add to confusion and psychotic phenomena. (Sands and Ratey 1986) Yet before fanning out the whole spectrum of resonance of today’s notion of noise, it is important to seize the precise moment noise erupts as a key concept in science and technology. Claude Shannon famously devised a mathematical theory of signal transmission that paved the way for the effective elimination of noise from the channel of communication.

According to Claude Shannon’s Mathematical Theory of Communication (MTC), information can be defined, in terms of An Epistemology of Noise 4 its probability, as a measure of ‘information entropy’. Let it suffice to say here that ‘entropy’ indicates what Warren Weaver, in his introduction to MTC, calls ‘freedom of choice’, relative to the unpredictability of a message. What this means is, critically, that a piece of information informs us only, if it is not redundant, in other words, if it contains a margin of unpredictability and hence uncertainty. The status of this physical concept, ‘entropy’ – coined by Rudolf Clausius to describe the loss of available energy in thermodynamic terms – requires careful examination, especially when we consider the inflection that information theory has given to the way we use the word noise in the natural and the human sciences.

Today the dictionary defines entropy as a state of molecular ‘disorder’ (Larousse 2017). Ludwig Boltzmann’s great innovation during the nineteenth century was to devise the statistical formulation of molecular entropy, which became the basis for Shannon’s formalized concepts of both information and noise. Yet despite the high degree of formalization, the intuitive idea of disorder continues to colour our idea of entropy and, consequently, both Shannon’s concepts of information and noise. The persistence of a persuasive and intuitive idea of disorder when we refer not only to noise but also to ‘information entropy’ may help explain why, despite the striking clarity of Shannon’s idea – namely that information must tell us something new, something we could not predict – his definition of information as ‘information entropy’ has failed to impose itself outside the mathematical theory of communication (MTC).

What appears to have dampened the reception of Shannon’s concept of ‘information entropy’ is that the correlation between novelty of information and disorder also threatens the clear-cut conceptual opposition between information and noise. Norbert Wiener’s concept of information as, on the contrary, the negation of entropy has been adopted across the natural and human sciences through the neologism negentropy, coined by Leon Brillouin. A new understanding of information imposed itself as the negation of entropy, and more generally as the negation of disorder, meaning negation of everything contingent or unpredictable. The value of entropy thereby becomes a measure of unwanted variability, imprecision or error – in any case, a value to be eliminated for the sake of efficiency and certainty: entropy henceforth becomes synonymous with noise.

Through the broad success of Wiener’s cybernetic theory of self- regulating systems with feedback, the concept of negentropy has found its way quite naturally into our thinking of organization and information in general. Indeed, the very notion of a system, any system, can be put in cybernetic terms as a set of organized constraints on contingency, in other words, as Introduction 5 the organized negation of noise. As negation nestles at the very core of the cybernetic concept of information, information comes to reflect the level of organization of any system, insofar as it is apt to negate its spontaneous progression towards entropy.

Noise, in turn, becomes a metaphor for entropy as the chaotic dispersion of energy, as disorder, if not as the entropic ‘death’ of a system. Such an impoverished view of information, impoverished because lacking in the complexity that entropy contributes, of course fails to adequately represent the theoretical wealth of Wiener’s own approach, and of the subsequent development of cybernetics into second-order cybernetics e. the cybernetics of self-observing, self-regulating systems with feedback) as well as of more recent developments in complexity theory. The point is, and will be throughout, that concepts circulate through general discourse and that general discourse in turn leaves its mark on the circulation of concepts: as the metaphor of noise as ‘parasite’ in the channel of communication started thriving, the idea of negentropy disseminated itself across the natural and human sciences and general discourse, often without its mathematical formulation, and frequently without being directly acknowledged.

This early formulation of one of the key concepts of cybernetics has thereby contributed to polarize our epistemic field in its relation with the unpredictable and the improbable. Now widely diffused, what subsists in general discourse of Wiener’s idea of information as negentropy subtly inflects our thinking about organization: from the organism to the ecosphere, from sociopolitical to economic relations, from networks to the idea of globalization. As a result, by emphasizing the negation of contingency our idea of information has become tethered to predictability and consequently antithetical to noise as the unpredictable. And yet our narratives of prediction and self-regulation have failed both spectacularly and catastrophically before the crises that have inaugurated the twenty-first century (Walter and de Pracontal 2009) – heralding a ‘post-truth’ era of politics, catastrophic crises in finance, war and migration.

The time has come, it seems, to re-evaluate the epistemological import of Shannon’s entropic idea of information, and to do so in light of this new protagonist concept: noise. A closer look at both information theory and cybernetics reveals that the opposition between Shannon and Wiener’s mathematical approach to information and noise is far subtler than it seems. But even then, the consequences of this difference in thinking about the relation between information An Epistemology of Noise 6 and noise remain significant. They inform the widely diffused integration of the concepts, both negentropy and noise, into theoretical contexts that are neither fully mathematized nor reducible to the technical idea of ‘noise in the channel of communication’ (Morange 2006).

To understand the conceptual ramifications of noise thus requires a careful evaluation of the relation between physical entropy and ‘information entropy’. Only when the moral and perhaps even ideological connotations of the notions of ‘organization’, ‘work’ and ‘order’ are elucidated in their relation to predictive certainty can we begin to understand how noise, alongside concepts like ‘metastability’ or ‘non-linearity’, could become common parlance in business management and political lobbying alike. Take, for instance, David Cummings reflections on the ‘Vote Leave’ campaign, which he directed leading up to the Brexit referendum in June 2016. Cummings explains that the success of the campaign was driven by new communication technologies and the targeted use of social networks.

His account of the success of the ‘Vote Leave’ campaign is laced with the words ‘non-linear’, ‘interdependent’, ‘unpredictable’, ‘irrational’, ‘complex’ and even ‘noise’: A news broadcast now contains much less information content and much higher noise than reading. The only way to improve this is experimenting with formats in a scientific way. (‘Dominic Cummings: How the Brexit Referendum Was Won’ 2017) The politics of information and noise are thereby elevated to a pseudo-scientific status. However, what this pseudo-scientific smokescreen dissimulates is the oldest trick of the trade: xenophobia, the fear of the other. To acknowledge and analyse the interwoven nature of scientific, technological, moral and political components of the conceptualization of noise is therefore indispensable.

It is all the more important, therefore, to stop and pause before the rapid proliferation of the idea of noise across the natural, the human sciences and public discourse. It is important to pause here for two reasons. As we have seen, the noise metaphor endows discourse with a scientific aura, lending it authority beyond the limits of its rational means. What is perhaps less obvious are the moral and ideological inclinations that can, in turn, also affect scientific discourse – for instance, when noise is associated with irrelevance, abnormality or disturbance. The second reason is more properly philosophical because it alerts us to an aspect of the theory of knowledge that can no longer be sidelined as marginal.

The ambiguity that accompanies the flurry of conceptualizations of noise does Introduction 7 not belong in the margin of error of scientific discourse, if we recognize instead that ambiguity is the space that we must negotiate as part of the moving frame of debate shared by contemporary science and general discourse. It is no longer just the question of how we define empirical noise, but of how the theory of knowledge handles ambiguity. In other words, we are looking at an epistemological problem of noise. The first thing that becomes apparent upon closer inspection is that a shared formal definition of noise is lacking.

This lack opens a space for metaphorical reverberation within scientific discourse, and even more so in the straits between the natural and the human sciences, technology and the arts. Inigo Wilkins notably mounts a thorough critique of the metaphorical distortions of the concept of noise in the humanities in his Irreversible Noise (Wilkins forthcoming). Wilkins thereby aims to redress the fetishization of noise as randomness and chance, and its reduction to the unintelligible. By situating the concept of noise within the context of contemporary science, and by re-centring it around mathematical definitions of randomness in the context of complex dynamical systems, Wilkins builds a case for a concept of noise that can act, instead, as an index of intelligible constraint.

Thus re-enforcing the mathematical and scientific parameters of the concept of noise, Wilkins emphasizes both formal (logical) and empirical (scientific) criteria by which to assess the critical potential of conceptualizations of noise in the humanities. The conceptual building site of Wilkin’s approach could be said to coincide with the present approach insofar as both problematize the multiple conceptualizations of noise. However, the problem addressed here differs in perspective. Rather than making metaphorical ambiguity the target of elimination, it is here taken to be a relevant philosophical problem in its own right. The objective here is thus not to eliminate ambiguity in the conceptualization of noise, but rather to make it explicit so that we can get to grips with it as a form of epistemological noise that accompanies the transformation of the epistemological field, especially in the context of today’s increasingly translational and interdisciplinary platforms for research.

Metaphor, of course, remains a dirty word in the context of scientific and even of much philosophical discourse. Yet it is necessary to acknowledge that metaphors are used abundantly, if artlessly, in scientific and philosophical discourse. Being in denial about the critical role of metaphors, especially in the communication of specialist knowledge to non-specialists, and in the context of growing interdisciplinary consortia, is to submit uncritically to their rhetorical power. An Epistemology of Noise 8 Metaphors of noise are not sufficiently subjected to critique, notably because the humanities are kept at bay. The much talked about rapprochement of the natural and the human sciences is currently still confined to making the humanities, especially psychology and sociology, more empirical, weighting analysis towards quantitative, rather than critical discursive, analysis.

Yet what we need is precisely a return to the critical lessons coming from the literary end of the humanities: what we need is to learn how to use metaphors critically, purposefully and artfully. Rather than allowing the metaphor of noise to burrow its way into the cracks of scientific discourse, like a repressed and unacknowledged fear or desire, we must learn to cultivate a critical use of metaphors in public discourse within and on science and technology. We thereby honour the epistemic humility already practised in the sciences but also in the arts, where it has long been established that reality speaks to us through the art of framing, as much as through what is framed.

At stake in this book is the shifting boundary between information and noise and the sprawling of the idea of noise, in the guise of its many technical and non- technical definitions. The epistemological, moral and political implications of its impermanence are the red thread that connects the three parts of this book. It is precisely the difficulty in its conceptual delimitation, especially in light of the idea’s projection across the academic and professional boundaries of diverse theoretical and experimental fields, that makes noise a privileged philosophical problem. This restlessness of the conceptualizations of noise, the shifting boundaries between what we consider to be information and what we discard as noise, requires that we think about this restlessness as a form of epistemological noise.

Meaning quite simply that the communication between diverse theoretical and experimental fields is not only subject to conceptual ‘noise in the channel of communication’, but also generates epistemological noise. The very movement of the idea of noise across disciplinary boundaries, the conceptual distortions provoked by this movement, is a form of epistemological noise accompanying its dissemination and transformations. In other words, the unstable concept of noise is itself an example of epistemological noise in the communication of concepts across theoretical boundaries. The conceptual distortions this sprawling provokes, and the metaphorical and ideological inclinations it reveals, could be said to act as what the French epistemologist Gaston Bachelard called ‘epistemological obstacles’.

Yet rather than chastising interdisciplinary or transdisciplinary discourse for its imprecisions in the use of the term noise, or even its relapses into proto-mythological thinking (for instance, in the rapprochement between noise, disorder and chaos), the Introduction 9 emphasis here is on the epistemological necessity of cross-fertilization between the diverse theoretical and experimental fields, in other words: on the co- constitutive role of noise in the formation of knowledge. This is what will be called epistemological noise. In order to understand the proliferation of ideas and concepts around noise, the focus thus cannot be exclusively technical.

This would be to overestimate the fidelity of conceptualizations of noise to mathematical formalization, or even to the technical applications the term noise has found in computational logic and in new technologies of communication. However, also literary and artistic evocations of noise are insufficient on their own, if they are limited to a Romantic indulgence in noise as indicative of the incommensurability of Being. If noise becomes the placeholder concept of a philosophical Other, as that which does not submit to reason, then also hopes invested in its revolutionary potential risk petering out in enthusiasm.

What is more, the temptation to indulge in noise as the mere negation of limit (embracing only the Greek apeiron, the unlimited) or of established norms (as the abnormal) squarely inhabits the conservative logic of negation. In short, it fails to subvert the very logic that the idea of negentropy occupies in the cultural and scientific imaginary. The present conceptualization of noise owes much to the attention paid in French epistemology and contemporary French philosophy to the ‘shifting sands of emergent truths’, as Alberto Toscano aptly expresses it in his translator’s introduction to Alain Badiou’s Being and Event: For millennia, philosophy has attempted to ground itself on One Eternal Necessity such as the prime mover, or the dialectic of history.

Here it consciously chooses to ground itself on the shifting sands of emergent truths. (Badiou 2005, xxiii) Readers may recognize echoes of Badiou’s insistence on the question ‘what counts as ’ when we enquire into the question: what counts as information, and what can be discounted as noise? Also a post-Cartesian perspective on the subject, such as it animated a certain generation of French philosophers and theorists, will come into play when we consider the ‘mental state of noise’ in light of John Keats’s definition of the poet’s negative capability. However, the way noise will be problematized here, as a polyvalent and polymorphous concept, will not base itself on equating mathematics with ontology.

Necessarily fanned across a wide range of topics, the problem considered here, e. the moving boundary between information and noise, An Epistemology of Noise 10 will be treated more modestly like a minimal gesture in philosophical terms, a way of reiterating the act of drawing a line from various angles, rather than an epic tableau of noise in the grand genre of systematizing philosophy (Badiou 2005, 15). This minimal gesture of repeatedly drawing the line between information and noise in various contexts, has the advantage of revealing a margin of conceptual indeterminacy between diverse fields of knowledge.

To take stock of the epistemological noise that arises from the inevitable sharing of concepts and from the unavoidable recourse to common language, helps us understand the growing complexity of the field of knowledge as a whole, in analogy to the way in which the philosopher and bio-physicist Henri Atlan speaks of ‘complexity on the basis of noise’ (Atlan 1979). Such an approach to the theory of knowledge, as engaging all fields of knowledge in their plurality, takes in its stride the shift from an ideal of knowledge without noise, indebted to the Cartesian Method, going towards an idea of knowledge that gains in complexity by being exposed to epistemological noise.

To re-evaluate noise as a problem of epistemic complexity is to acknowledge the functional role of uncertainty and ambiguity in the process of concept formation. Michel Foucault’s introductory words to George Canguilhem’s re-edition of The Normal and the Pathological here help to express the motivation for this book: Error is at the root of what makes human thought and its history. (Foucault 1989, 22) This is how noise can be understood, ultimately, as a radical concept, in the sense that Foucault and Canguilhem understand error, as touching the root of human thought and of its historical irreversibility.

At stake is the relation between thought and contingency, which has become emblematic for a certain way of thinking about the philosophy and history of the sciences. Yet, while the swarming interest in noise makes it an imperative to engage with it conceptually, the synthetic view that is called for is, by definition, condemned to fail in making even a dent in any of the individual fields of knowledge and practice that gravitate around the notion of noise: this book will not improve stochastic models of noise, it will not resolve new problems of noise in big data, nor will it improve propositions to tackle noise pollution – least of all will it attempt to tell artists and musicians, or cultural and critical theorists, how to conceptualize noise.

In fact, it cannot even begin to do justice to the extent of diffusion of the notion of noise to other disciplines, which is in a process of active fomentation, expansion and dispersion. Introduction 11 While this book owes everything to the growing wealth of literature on noise, the conceptual movement I am after is not primarily concerned with the knowledge of diverse phenomena understood as noise, but with the idea of noise in the relation between the known, the unknown and the differently known. This is why, unlike Greg Hainge’s journey through the many dimensions of noise in his Noise Matters (Hainge 2012), this book cannot aim at what he calls an ‘ontological taxonomy’ of noise, because this is predominantly an epistemological enquiry, rather than an ontological one.

Rather than aiming at a phenomenology of noise (Cage et al. 2012; Voegelin 2010), and despite benefiting hugely from the vast spectrum of literature on acoustic noise, on the cultural (Bijsterveld 2001; Boutin 2015; Gibson and Biddle 2016; Schafer 1994; M. M. Smith 2004) and even military history of noise (Volcler and Volk 2013) and the impact of cybernetics and information theory (Bunz 2012; Mersch 2013) as well as the psychology of perception (Bawden and Robinson 2009; Manson 2014; Shenk 1997), the problem of epistemological noise, as posed here, is ultimately co-extensive neither with a phenomenology of noise, (because the question posed here starts from Shannon’s counter-intuitive relation between information and noise), nor with its cultural history, (insofar as it focuses on the conceptual implications of thinking about noise).

And although noise music and noise art are what opened my mental shutters to the prospect of thinking about noise, this book is about neither, leaving this avenue open for future projects (Attali 1985; Brassier 2007; Hegarty 2007; LaBelle 2006). This book looks also, albeit obliquely, at the emerging field of the philosophy of information, and more specifically at its effort of reconciling information concepts underlying science and technology with the humanities. Information philosophy has recently established itself as a specialist discipline in philosophy. It comprises new fields that have arisen from thinking about electronically mediated information, dealing with the transmission, circulation and conversion of one form of information into another (Dodig Crnkovic and Hofkirchner 2011).

Its teeming developments are so diverse that they can barely be overseen and do not yet appear to constitute a coherent whole in the eyes of its contributors – encompassing fields as varied as logic and computation, cognitive and neurocognitive sciences, dynamical systems and actor network theories, cybersemiotics and biosemiotics, information systems and epistemology, and information culture and information ethics (Capurro and Hjorland 2003; Diaz Nafria 2010; Floridi 2010). Leaving many references, precious ideas and references out, indeed cutting large chunks of the work that prepared this journey into noise was the necessary An Epistemology of Noise 12 sacrifice so as to allow the form of the argument to emerge.

This cut is but the performative aspect of the problem this book ultimately faces: how do we draw that line that makes the form of an argument emerge, even an argument about noise? What can we afford to exclude? How much variety, and hence how much uncertainty can we retain, without dissolving the very movement of thought, whose emergence we only begin to comprehend? In this sense, we will ultimately come to think of maximum noise as an unthinkable freedom of choice. While this book can claim none of the academic fields it visits as its own, it seeks to understand the problem of the conceptualization of noise as a problem that relates them, without reducing them to any single dominating view.

The oblique relation between these multiple domains requires that we understand the resonance of the idea of noise as something that, like the reverse of a carpet, reveals the messy connections that sustain the neatly separated forms of the academic organization of knowledge. To look under the carpet no doubt implies also a certain impertinence towards the well-established and well-deserved boundaries of specialist knowledge, at the risk of necessarily exposing one’s ignorance in comparison to those who have laboured hard to establish a more secure basis of expertise in any one of these fields. Perhaps the risk implied in being – not unlike noise – excessive of boundaries of discourse, resonates with George Canguilhem’s lightly humorous concession that ‘the philosopher is indiscrete everywhere’ (Canguilhem 1993, 19).

Part One. Draw a straight line and follow it. La Monte Young, ‘Composition 1960 #10 to Bob Morris’ The conceptualization of noise takes a new turn in relation to Claude Shannon’s definition of information as ‘information entropy’ – this much is certain. Just what it means to rethink noise in relation to ‘information entropy’ is the question posed in this first part of the book. The aim here is not to elucidate the concepts of information and noise to the engineer who is in no need of such speculations for practical purposes, but to understand how this new way of thinking about information feeds into the broader scientific and cultural understanding of noise.

The latter turns out to be no passive receptacle of technoscientific concepts, but feeds back into the new technologically inspired understanding of noise, creating a new culture for theoretical and experimental practices. Shannon’s audacity consists quite simply in correlating both information and noise with uncertainty. Both concepts are henceforth derived from the statistical unpredictability he associates formally (mathematically) with physical entropy. While information entropy clearly implies a degree of desirable uncertainty, e. the novelty of the message, Weaver will say that noise can be discarded as ‘spurious uncertainty’. Yet it is, in both cases, unpredictability that is expressed via the calculus of probability and statistical analysis, constituting what is called ‘entropy of the message’.

As Weaver explains in his introduction to the second edition of Shannon’s Mathematical Theory of Communication (MTC) of 1964: The quantity which uniquely meets the natural requirements that one sets up for ‘information’ turns out to be exactly that which is known in thermodynamics as I How to Draw the Line between Information and Noise An Epistemology of Noise 16 entropy. It is expressed in terms of the various probabilities involved – those of getting to certain stages in the process of forming messages, and the probabilities that, when in those stages, certain symbols be chosen next.

(Shannon and Weaver 1964, 19) Before getting a better grasp of the status of the concept of ‘entropy’, both as a concept in physics and as a metaphor in statistical analysis, we will compare two statements further down, one by Warren Weaver in his introduction to Shannon’s MTC and one by Norbert Wiener in his book Cybernetics (Wiener 1961). These two statements show that there was no disagreement between those generally acknowledged as the founders of information theory and cybernetics respectively, regarding the method of calculating information probability; however, they also reveal the fact that the same mathematical method nevertheless justifies two diametrically opposed definitions of information: one of information as ‘information entropy’ and the other, on the contrary, of information as the ‘negation of entropy’.

Noteworthy is that these radically opposed definitions of information did not appear to constitute a problem even worthy of mention by either Shannon or Wiener. The introduction to Shannon’s MTC in fact begins by acknowledging Shannon’s conceptual debt, not only to Wiener’s mathematical work, but to his philosophy (Shannon and Weaver 1964, 3, n. 1). And yet, it proceeds to define information positively as a measure of entropy, and entropy as a ‘measure of one’s freedom of choice’. Contrary to Wiener’s definition of information as the negation of entropy, for Shannon, greater information goes hand in hand with greater uncertainty.

A completely predictable message, by contrast, has only one possible outcome and is therefore redundant; it tells us nothing new. In Warren Weaver’s words, [I]nformation is a measure of one’s freedom of choice 9) […] in these statistical terms the two words information and uncertainty find themselves to be partners 27) […] entropy (or the information, or the freedom of choice […]) 13). (Shannon and Weaver 1964, 9–27) In other words, the redundant message presents no ‘freedom of choice’, because it contains no ‘information entropy’. Information is null, if there is no uncertainty about the state of the message.

Norbert Wiener, too, acknowledges the shared origin of the statistical conception of information in his own and Shannon’s work, amongst others: This idea [of developing a statistical theory of the amount of information, in which the unit amount of information was that transmitted as a single decision between equally probable alternatives] occurred at about the same time to Concepts: Information Entropy, Negentropy, Noise 17 several writers, among them the statistician R. A. Fisher, Dr. Shannon of the Bell Telephone Laboratories, and the author. (Wiener 1961, 10–11) However, Wiener and Shannon arrive at diametrically opposed ideas of what information is, because Wiener defines information precisely as the opposite of ‘information entropy’, namely as the negation of entropy (which the physicist Leon Brillouin later entrenches as the dominant technoscientific definition of the concept of information, by inventing the neologism negentropy): The notion of the amount of information attaches itself very naturally to a classical notion in statistical mechanics: that of entropy.

Just as the amount of information in a system is a measure of its degree of organization, so the entropy of a system is a measure of its degree of disorganization; and the one is simply the negative of the other. (Wiener 1961, 10–11, emphasis added) For Wiener information is precisely the reduction of freedom of choice, and thus the reduction of uncertainty. Information, in Wiener’s cybernetic theory, is a measure of increased constraint, associated with ideas of organization and order, bound to decrease entropy. Here entropy is not a measure of information, as with Shannon, but, on the contrary, a measure of its presumed opposite,

e. disorder or noise. There is a startling matter-of-factness in the way both mathematicians provide diametrically opposed definitions of information, without mentioning their fundamental divergence. This may be indicative of the low priority that discursive definitions have for the two mathematicians. The real emphasis is instead on the mathematical innovation, which both share without disagreement, even complimenting each other. The concepts of information and noise are treated as theoretical tools that must be not only fit for purpose, meaning the communication of mathematical theories to a broader public, but also tailored to different needs, be it the transmission of a trans-Atlantic telephone conversation, or a successful targeting of a self-directing missile, to name just two of the most frequently cited examples of information theory and cybernetics.

What this tacitly accepted dichotomy between ‘information entropy’ and the ‘negation of entropy’ reveals is, first of all, that there is freedom of choice in the discursive interpretation of mathematical formalization. This freedom of choice is nothing other than the ambiguity of non-mathematical concepts. It is thus not at the level of mathematical stringency itself that ambiguity arises about the conceptualization of information and noise, but at the level of freedom of choice in its discursive interpretation. This source of discursive ambiguity is An Epistemology of Noise 18 important to underline, since it is at the level of discourse rather than at the level of mathematics, that the concepts of information and noise are translated into other scientific domains – notably biology (Morange 2006), and from biology to sociology and economy etc.

– often via a tacit adoption also of the cybernetic paradigm of self-regulating machines with feedback (Mersch 2013). It is here, at this crossroad of conceptual circulation, that we must be most attentive, because concepts reveal themselves to be more than just theoretical tools: they are prisms through which we see and discover the world at the same time as being the tools with which we transform the world. Their consequences go well beyond mere functionality in a theoretical apparatus for this or that technological or scientific purpose: concepts contribute to shape cultures and precondition value judgements, while being in turn also imbued with cultural preconditions and slanted by pre-existing value judgements.

Claude Shannon’s definition of information as ‘information entropy’ has the singular merit of having prepared the ground for a philosophy of noise that evades the Manichean opposition between information and noise, echoing that between order and disorder, life and death. It also evades the mere relativism according to which what we define as information or noise is a question of individual perspective. To demonstrate the cultural relevance of this conceptual feat, we will tackle the difficulty that arises when the concept of noise is no longer applied only to the channel of communication, but also to other domains, where the distinction between information and noise is not a given.

In vivo, rather than in the well specified and controlled situation of the channel of communication, the distinction between information and noise is never ready- made, but always presents itself as a vital decision or as an epistemological problem. At stake, hence, is the difference between information and noise in the making, e. the moment when information must be selected prior to the transmission of a message, necessarily implying a decision, an act of selection whereby information can stand out from noise. Is not the challenge of every form of research the problem as to how we can identify what counts as information and what, in turn, can be discounted as noise?

The dividing line between information and noise is so fundamental to all forms of enquiry and experimentation that the consequences of Shannon’s ‘entropic ideas’ vastly exceed any technological framework, making the conceptualization of information and noise philosophical problems in their own right. Shannon’s ‘entropic ideas’ require us to rethink our most basic attitudes concerning information and noise. Rather than opposing noise to information, as the presence of entropy to its absence, he divides the presence of ‘information Concepts: Information Entropy, Negentropy, Noise 19 entropy’ from the presence of ‘noise entropy’. The dividing line between information and noise now runs within entropy, rather than between entropy and its negation.

This is a subtle but fundamental shift that effectively challenges the principle of the excluded middle, according to which a proposition is either true, or its negation is true, and which implicitly underscored the analogy of the information/noise opposition with that of sense/non-sense, and even organization/chaos. A new division between desirable and spurious uncertainty now competes with the classical opposition between truth and error or, as in the excluded middle, between the truth of a proposition and its negation. The philosophical consequences are profound, for the process of information can now also be understood as a cut across the fabric of uncertainty.

Information becomes the process whereby this cut progressively gives rise to a form of measurable uncertainty. Shannon’s ‘entropic ideas’ thus have a profound philosophical and, more broadly, cultural importance, if only we are willing to consider their conceptual relevance beyond the technical realm. Common criticism instead holds that Shannon’s concept of information applies only to electronic signal transmission, and is utterly misleading in any other context. Complicit with this criticism is the equally common position that opposes culture and technology. Endowing only ‘cultural’ artefacts with signification, this view reduces all aspects of technology and even of science to their mere utility.

In its extreme form it represents a technophobia that pits culture, and even nature, against science and technology in a relation of hostility. The widespread cultural condescension towards the mere utility of the sciences and technology corresponds in kind to the technocratic dismissal of culture as mere recreation (or product of consumption). In fact, both attitudes are but two sides of a coin. Both fail to recognize the cultural potential of Shannon’s deconstruction of the traditional opposition between information and noise and the revaluation of uncertainty that it entails. Just as technocracy implies an aberration of cultural values, so technophobia fails to rescue culture, because it is itself the symptom of a redundant, conservative idea of culture.

It is redundant because it wilfully ignores the fundamental role that forays into mathematics and more broadly science and technology have always played in the visual arts, literature and music. The fascination with mathematics, science and technology has characterized the art of Greek Antiquity no less than the metrics of Arabic poetry, it has fuelled the European Renaissance five hundred years after driving Bagdad’s cultural prominence, finally it has been an indelible aspect of twentieth-century art, literature and music, and is even more so of the artistic practices of new millennium. An Epistemology of Noise 20 The redundant opposition between technology and culture atrophies not only the quality of engagement between the arts, the sciences and technology, but in turn also atrophies the status of creativity attributed to science and technology, by denying it its cultural relevance beyond its utility.

French philosopher Gilbert Simondon was right to speak of an enslavement of technology and to see in it a factor for mutual alienation in culture. To place Shannon’s ‘entropic ideas’ within this cultural frame of debate thus means overcoming the consensus that there is an opposition between technology and culture. The first task is to work against this alienation, so that we can recognize Shannon’s as a minimalist definition of information and noise of the highest cultural relevance. It is minimalist insofar as it deals with the conditions of possibility of information, precisely by bracketing out signification: it separates out signification from both the means and the process of transmission – thereby revealing the structural and procedural conditions of information processes, much like minimalist art did with artistic expression in an industrialized world.

Taken outside the narrowly technical context of signal transmission, we can begin to see that Shannon’s ‘entropic ideas’ also offer an iconoclastic definition of information and noise, one that breaks the spell fusing signification with the means and process of its transmission as if they were one. Some of the most beautiful words regarding the reconciliation of culture with science and technology have been written by Gilbert Simondon in On the Mode of Existence of Technical Beings (Simondon, trans. Malaspina and Rogove 2017, 15–16): Culture has constituted itself as a defense system against technics; yet this defense presents itself as a defense of man, and presumes that technical objects do not contain a human reality within them.

[…] The most powerful cause of alienation in the contemporary world resides in this misunderstanding [caused] by its absence from the world of significations, and its omission from the table of values and concepts that make up culture. Guiding us here is the ethos, rather than the method deployed by Simondon in METO (On the Mode of Existence of Technical Objects), where he gives a genetic account of the modalities of technicity across what he calls the evolution of technical individuals, technical elements and technical ensembles. It is not the objective here to construct a genetic analysis that would be in any way comparable to what Simondon did for the concept of technicity.

No comparable historicizing claim will be made about the splitting of the idea of noise across technics and religion, and across theory and ethics. Nor will there be an attempt to determine the role of aesthetics in mediating such a split. Concepts: Information Entropy, Negentropy, Noise 21 The objective here is more modest. It is to test two widely held presumptions about noise, and to do so in a number of different contexts, so as to reveal their intrinsic relatedness. The first is the implicit presumption that we can rely on an intuitive notion of noise, in order to bridge its definitions across different techno-scientific and cultural settings.

The second presumption is that, rather than intuition, it is a formal, e. mathematical, definition that presides over the multiple uses of the concept of noise across the spectrum of scientific discourse. What emerges instead are far from uniform conceptions of noise, some of which profoundly counter intuitive. Although ubiquitous, both the idea of noise and information reveal themselves to be conflicted, both displaying a fundamental ambivalence towards novelty and change, as signaled by Shannon and Wiener’s mathematically identical, yet discursively opposed definitions of information.. To equate information with unpredictability is intuitive enough, if it is to tell us something new, something that does not follow automatically from what came before.

It is equally easy to accept that a message we can fully predict is redundant, if it gives us no new information. What is much less intuitive are the consequences Shannon and Weaver draw from this unpredictability. By aligning the concept of information with uncertainty and by quantifying it as such and without concessions, we arrive at the apparently paradoxical conclusion that more information means more uncertainty. This appears paradoxical, in the sense that it contradicts the equally common assumption, even the doxa (opinion or dogma), that information is what reduces uncertainty, rather than increasing it.

Quantifying information according to the degree of uncertainty it presents, according to the ‘entropy of the message’, has therefore caused alarm (Janich 2006). Shannon appeared to have fundamentally misunderstood what we mean by information. The fundamental role he gives to contingency in information contradicts what we commonly associate with the purpose of information, namely that information is reliable only if it reduces uncertainty and makes experience less contingent. Shannon’s definition of information thereby appears dangerously close to that of noise. In his article ‘What Is ’ José Maria Diaz Nafria’s makes the implications of this difference between the ordinary sense of information and Shannon’s definition of information entropy abundantly clear.

If the only criteria for the quantity of information is its unpredictability, expressed in terms of entropy, then a critical signal consisting of few bits would be discounted as low in information, while the high entropy of irrelevant background noise would measure the greatest quantity of information: Just one binary digit may tell us if the universe is about to collapse, thus being very informative, and all millions of terabits on the web could just as well II Entropy as ‘Freedom of Choice’ An Epistemology of Noise 24 be generated by the whim of electrons in a rheostat, being thus completely uninformative.

(Diaz Nafria 2010) It must be clear from this example that Shannon’s entropic ideas about information are not a mere extension or deepening of the ordinary notions of information and noise, but a challenge to the ordinary conception of information of the highest order, since, as Diaz Nafria’s example makes plain to see, nothing distinguishes outwardly ‘information entropy’ from what we would ordinarily call noise. While information is, as a matter of course, meant to tell us something new, the logical consequence that this novelty decreases predictability and thus increases uncertainty appears to be going one step too far.

Shannon’s quantitative measure of information has since been interpreted almost as a form of sacrilege against the ‘true’ understanding of information, which ought to increase certainty. It is also discounted as incapable of telling us anything about what matters, which is not quantity, but quality of information, and which is a prerogative of its signification. We could say, on the other hand, that Shannon’s understanding of the relation between information and contingency is indeed paradoxical, but not because of the misplaced conceptual ambition. It is paradoxical in the sense that it is free of cultural pre-conceptions and therefore offends such pre-conceptions, transgressing their doxa: in this sense the conceptual innovation inherent in Shannon’s concept of ‘information entropy’ indeed acts as a form of conceptual noise, when it is exported from its technological application to other domains.

Let us be clear, Shannon’s definition of information as an ‘uncertainty relation’ does not contradict itself, but the doxa according to which one ought to obtain from information simultaneously both novelty and a reduction in uncertainty. Shannon’s definition of ‘information entropy’ instead frustrates this paradoxical need (novelty and certainty) and thereby enables us to think about contingency as belonging to the conditions of possibility of all processes of information, including but not only of those processes we associate with signification in the semantic communication between sapient beings. What, then, is the relation between uncertainty and information, and hence also between information and noise?

The answer to this question is not as obvious as it might at first seem and unfolding it may change the way we think about both noise and information. It is this question that is posed, in mathematical terms, by Shannon’s MTC. Shannon gives an engineer’s answer to this question, which Warren Weaver translates for a broader readership, expressing it in the following way in his 1964 introductory essay to the MTC: Concepts: Information Entropy, Negentropy, Noise 25 [I]nformation is a measure of one’s freedom of choice […]. (Shannon and Weaver 1964, 9) This definition is intentionally shortened here in order to indicate the importance we must attribute to it in the context of our enquiry into the conceptualizations of noise.

As a technical definition it is, first of all, abstracted from the habitual association of information with signification. The common criticism of Shannon’s concept of information is therefore that it is a purely quantitative measure that is indifferent to the signification of a message, which consequently ignores the inherently qualitative aspect of information, which is its signification. The criticism levied against Shannon’s ‘entropic ideas’ is certainly important and valid at the level of interpretation and evaluation of a message, but it has also detracted attention away from two very subtle philosophical gestures that Shannon’s approach implies: one concerning information as a process rather than a given, the other concerning the role of contingency, and hence of uncertainty, in this process.

It is necessary to render these explicit, as the concept has de facto been translated and applied to a great variety of disciplines in the natural and human sciences, often as a language that facilitated profound methodological upheavals, like the shift from classical to non-classical mechanics, from classical physics to quantum physics, from classical biology to molecular biology and biophysics. Shannon’s definition of information as ‘entropy of the message’ offers us a contribution to our understanding of contingency, which challenges both the doxa according to which information is a correlate of certainty and the certainty with which we can distinguish information from noise when the concept is taken outside the technical paradigm of the channel of communication, for instance in biology or economic theory, where the scenario of the engineer who transmits a ready-made message as information no longer prevails.

Shannon’s contribution, which follows from the fundamental realignment of information and uncertainty, is fundamental insofar as it enables us to place information and noise on an equal footing, where both represent a measure of ‘entropy’ or unpredictability, prior to the assignment of signification, purpose or representation; prior, in other words, to the levels of decoding, interpretation and evaluation. If we follow through with Shannon’s ‘entropic ideas’, our fundamental assumptions about information must be rethought, taking contingency and hence noise into account, not only as that which impinges on the fidelity of the message, not only as that which obstructs the decoding and interpretation of information, but as an uncertainty fundamental to the process of information

An Epistemology of Noise 26 itself. This attention to contingency and uncertainty is what will enable us to rethink the definition of noise, to take it outside the channel of communication, in other words to think about noise in vivo, where the distinction between information and noise is always a process in the making. The emphasis is here on the conceptual consequences that must be drawn from Shannon’s alignment of information with contingency, and more specifically concerning the less Manichean, less oppositional relation between the concept of information and that of noise. We can now ask, in light of the indelible aspect of contingency in the information process, on what grounds can we draw the line between ‘information entropy’ and noise?

Meaning quite literally that the distinction between information and noise is a problem of ground or foundation of knowledge. If both noise and ‘information entropy’ are measures of ‘entropy’, understood as in purely probabilistic terms as ‘freedom of choice’, then how can we be sure which measure of choice informs and which exceeds and deforms? If uncertainty increases with both information and noise then, in this greater uncertainty, with what certainty do we draw the line between information and noise? The conceptual operator upon which the idea of information uncertainty hinges, is entropy. But how are we to understand the idea of entropy, when it is no longer a concept bound by the theoretical and empirical constraints of the field in which it arose as a key concept: thermodynamics?

Shannon’s theory of information is said to have emerged from his work on Boolean logic, applied to electrical switches. It owes more, in fact, to Norbert Wiener’s use of the calculus of probability in cybernetics, than to a direct engagement with Boltzmann’s statistical theory of physical entropy (Atlan 1979). The resulting formalism, however, is not just metaphorically, but formally analogous with the statistical expression of physical entropy. Let us see how Shannon’s designation of information as ‘information entropy’ builds entropy as an indispensable metaphor into this ontologically arbitrary concept of information. Shannon’s choice of expressing the unpredictability of the message as ‘information entropy’ (H) reflects the fact that its mathematical formulation in information theory, is indeed almost identical with Ludwig Boltzmann’s statistical formulation of molecular entropy in thermodynamics (S): H = – Σ pilog pi, S = – K Σ pilog pi Both information [H] and the physical system [S] measure the number of possible states, either as a message or as a physical entity.

This probability, attached to the number of possible states, is the sum of probabilities of the ‘presence’ [p1, p2, …, pn] of ‘signs’ or particles [i], multiplied by a logarithm [log]. [H] is thus a measure of the uncertainty over the occurrence of one amongst all possible events H(p1, p2, …, pn). If all probabilities [pn] are equal, then the greatest possible ‘freedom III Information Entropy and Physical Entropy An Epistemology of Noise 28 of choice’ corresponds to the greatest possible uncertainty regarding the actual state of either system or message, with respect to all its possible states.

More simply put, in the state of maximal entropy the movement of particles is determined by nothing but their random collision. The more random the movement of particles is, the greater is the number of their possible positions, speed and direction. Maximum entropy thus corresponds to the equal probability of all possible states of the system, which is often illustrated with the idea of the random collision of molecules in a canister of gas. All configurations of free particles can thus be said to occur with equal probability. Conversely, the probability of predicting the state or behaviour of the physical system, meaning the exact configuration of its micro-complexions, including the positions, speed and directions of its particles, is at its lowest.

Entropy and noise indeed become identifiable, and to this extent more predictable, thanks to a better knowledge of the statistical framework which encompasses them, which we owe notably to the important work of Boltzmann and Shannon, no less than Wiener and many others. Nevertheless, it would be wrong to overstate the statistical determination of noise and call it predictable, because ultimately entropy and noise remain a measure of ‘freedom of choice’, characteristic of an undetermined state of the system or message. The quantity of information in Shannon’s sense is analogous to the probability with which the observer of a physical system can predict what Max Planck called the micro- complexions of a given system, and the probabilities of finding the system in any of these complexions: Maximal disorder corresponds to the greatest number of possible complexions with equal probability for all […].

(Atlan 1979, 31, 1) Disorder here means that no external order imposes any form of constraint that would compel particles to behave in one way rather than another. Both entropy and ‘information entropy’ must thus be defined by sophisticated statistical measures expressing the receiver’s uncertainty as to the determination of the system, message or event. Increased quantity of information, in this sense of ‘information entropy’, is thus not the equivalent with increased certainty about the system, even if certain forms of noise have identifiable and reproducible characteristics in statistical terms. To predict the probability with which signals occur in a message, Shannon uses a mathematical expression that is almost identical to Boltzmann’s.

What is significant, however, is that he leaves out the term ‘k’. ‘k’ is the physical constant that expresses the calorific value of flows of energy, understood as displacements of thermal charges, wherever a disparity exists between energy levels, for instance in electrical currents or in flows of matter. This algorithm, ‘k’, is what anchors Boltzmann’s formula in physical reality. Physical potential arises from unequal energy levels that compel a process of equalization. Potential therefore expresses a form of constraint on the system, which is forced by the disparity of energy levels to evolve in such a way as to equalize this difference.

Potential thus increases the probability of a system to evolve in the direction of this equalization of energy levels, and its way of doing so will depend on the interaction between its constituent elements. Potential thus effectively reduces the number of possible states of a physical system: compared to the state of maximum entropy where all possible states occur with equal probability (for instance, the molecules randomly bouncing off each other in a canister of gas), potential obliges the system to actualize an equalization of energy levels. Now, if potential reduces ‘freedom of choice’ by compelling a process of equalization of energy levels, then the greater the disparity of energy levels – the greater the potential – the more powerfully the system is entrained to evolve in a particular way, as for instance in the flow of an electric charge.

Even if a margin of indeterminacy persists as noise, potential is what reduces the number of possible states of the system, forcing it to evolve according to its constraints. To transpose the idea of potential information to Shannon’s ‘entropic ideas’ thus runs us into difficulties, if we want to preserve the idea of ‘freedom of choice’. To understand potential as a form of constraint no doubt offends common sense. It forces us to pause before the usual idea of potential as synonymous IV The Idea of ‘Potential Information’ An Epistemology of Noise 30 with opportunity in the sense of ‘freedom of choice’.

When we say that an event has the potential to occur, we usually mean that it may or may not occur: potential thus represents an added option and hence greater ‘freedom of choice’, as when one says ‘he has the potential to become a great pianist’. When we say a child has potential, what is meant is that it has the opportunity to develop certain capabilities. On the contrary, a child seen as lacking in such potential is presumed to have fewer opportunities and thus a lesser ‘freedom of choice’. The child’s perceived aptitude, however, is better understood as an optimum fit with given requirements.

The potential to become a great pianist, for instance, compels a child onto a certain path of development. Because potential is valued as increased ‘freedom of choice’, we ignore the constraints that turn perceived potential into a constraint: the child with perceived potential is compelled to succeed. But what is success, if not the growing constraint to proceed from one level of achievement to the next? It becomes an almost mechanical sequence of expectation and compliance – which ultimately feeds into the social reproduction of relations of power. If to succeed, more often than not, increasingly narrows the noose of the expectation not to fail, then is the ‘freedom of choice’ we attribute to potential not ultimately greater for the one who is not pressed into the mould of expectations, corresponding to a perceived potential, or who deliberately confounds these expectations?

Art critic Martin Herbert, for instance, in his collection of essays Tell Them I Said No, looks at the work and lives of ten artists who, in different and sometimes extreme ways, refused to play the game of celebrity that enchains artists to an ‘overly educated’ and ultimately conservative audience (Herbert 2016; Judah 2017). In a market driven society it is undoubtedly heretical to question the link between potential and ‘freedom of choice’. It is only when we turn the idea of potential around and express a negative potential that the compelling nature of potential as constraint becomes more evident, and we can begin to think about potential as reduction of ‘freedom of choice’.

If we say, for instance, that a certain group of underprivileged children will potentially fail to thrive in society, according to available sociometric parameters, then the idea of ‘potential’ reveals its negative characteristic of constraint more readily. In both cases, however, expectations are present, be they valued positively or negatively, and potential can be said to reduce ‘freedom of choice’ insofar as it represents criteria, norms and structural conditions that compel a child, or any other observable phenomenon with potential, to evolve according to its greatest probability. What Concepts: Information Entropy, Negentropy, Noise 31 needs to be retained here is simply that the physical concept of potential, with its noble Aristotelian heritage, is by far not an anodyne synonym for the possible, when it comes to understanding the relation between probability and ‘freedom of choice’.

It is thus important to stay alert when considering whether ‘information entropy’ must be understood as ‘potential information’. Physical potential implies that an event is more likely to occur, thus in fact reducing the number of possible events, as when one switches on a light circuit and the electricity is compelled by the physical potential to rush through the wire. Both potential and freedom of choice are manners of speaking about a possible event, yet the difference of inflection between potential, perceived as an option, as greater ‘freedom of choice’ and potential as greater probability of occurrence, hence reduction of choice, is not without consequence.

In the state of maximal entropy, on the contrary, initial differences in energetic potential have equalized through interaction, until the system as a whole finally reaches a state of energetic equilibrium, where flows of matter or energy from one part of the system to another are highly improbable, at best random effects, because the micro-constituents of the system are no longer exposed to the tension of discrepancies between energy levels, no longer compelled by the physical potential that arises from these differences. Consequently, each state of the entropic system occurs with equal probability or, differently put, with the greatest ‘freedom of choice’.

Coming back to Shannon’s formal mathematical definition of information, this means that to define ‘information entropy’ as potential information is to inverse it completely: potential, strictly speaking, would be a negentropic factor, negating entropy. In Boltzmann’s definition the algorithm [k] serves to indicate flows and displacements of thermal charges. Yet by leaving out the reference to this physical aspect, Shannon transforms Boltzmann’s mathematical expression of entropy into an ontologically arbitrary measure of probability. Shannon thereby unmoors probability from Boltzmann’s empirical measure of calorific conversion of energy and work related to thermal displacements in a physical system.

Although Shannon himself applies this formula to the problem of electronic signal transmission, his concept of ‘information entropy’ and hence also of noise is now devoid of any ontological reference: it could inform us about the probability of occurrence of any phenomenon involving large numbers, be it the flow of signals, flows of people, of goods or unicorns – in short it is ontologically arbitrary. An Epistemology of Noise 32 The abstraction of Shannon’s quantitative measure of information is undoubtedly what facilitated its translation with great ease into every imaginable field of research involving mass phenomena.

It lends itself to statistical analysis, not only in communication technology, but also in economical or biological systems or any other domain. This gives Shannon’s concept of information an eminently analogical, if not paradigmatic function. Shannon’s definition is lower in theoretical constraints than Boltzmann’s formula, giving it greater polyvalence, but by the same token also increasing ambiguity of its interpretation: with respect to Boltzmann’s definition of entropy, Shannon’s concept of ‘information entropy’ is thus itself a prime example of what ‘information entropy’ does, when it increases the number of possible interpretations, namely increasing also uncertainty.

The analogy between information and entropy, nevertheless, remains paradigmatic in the denomination of information as ‘information entropy’. The application of the term information to communication technology also reinforces this analogy with physical processes, because the engineer must deal with the effect of physical entropy in order to ensure the message is sent without loss due to perturbations, such as thermal noise, during the transmission of an acoustic or electric signal. The physical analogy thus persists, but as we have seen on a metaphorical rather than formal, mathematical level (Morange 2006). As a result the notion of physical potential remains, despite the obliteration in Shannon’s concept of the physical referent [k], a key feature of the concept of information.

What persists is also a margin of ambiguity, when one speaks of ‘information entropy’ as the information one lacks, or as ‘potential information’, as the German physicist and philosopher Carl Friedrich von Weizsäcker does, using the term entropy here with specific reference to Shannon’s ‘information entropy’: Positive entropy is potential (or virtual) information. (Weizsäcker 1994, 167) However, the idea of possibility which is implied in both physical ‘potential’ and the ‘virutal’, risks blurring a distinction that is perfectly clear to the physicist and much less clear in ordinary language. What von Weizsäcker means is that entropy, which denotes the number of possible states of a system, corresponds to virtual information and he specifies this in the parenthesis.

Von Weizsäcker thus speaks of ‘potential’ information in order to make the idea of ‘information entropy’ more accessible. However, the virtual refers to the number of possible states, which increases with ‘information entropy’, while physical potential, as Concepts: Information Entropy, Negentropy, Noise 33 we have seen places a constraint on the physical system and thus decreases the number of possible states by making one event more likely than another. Where the notion of information ‘potential’ is introduced, it is thus in fact re- introduced as an extrinsic criterion for the evaluation of ‘information entropy’, more specifically of its hoped-for use-value as information in the traditional sense of certainty and constraint.

Better put, the idea of potential information introduces the idea of the capacity of ‘information’ to perform work, to make sense, which in turn is specific to the recipient of this information and the use s/he can make of it. What remains ambiguous and unspoken is the necessary conversion between the uncertainty that ‘information entropy’ introduces as ‘freedom of choice’, as under-determination, and the implied sense of potential information leading to negentropy, e. of increased certainty and constraint. Implied is that the actualization of potential information is equivalent with this conversion of uncertainty into certainty.

And nothing could be further from certain than the spontaneous consolidation of uncertainty into certainty. For what this requires, is also that the nature of the boundary between information and noise changes, from being a limit that curtails the uncertainty of the ‘entropy of the message’ vis-à-vis the unlimited uncertainty of noise, to a border that opposes information and noise as certainty and uncertainty. There is thus continuous ambiguity at the level of conceptualization when the notion of physical entropy is transformed into the pure probability of ‘information entropy’, at once untethered from the physical paradigm, yet indelibly tied to it through metaphor and philosophical tradition.

It is this ambiguity that constitutes ‘epistemological noise’ when Shannon’s concepts of information and noise are exported, alongside negentropy and often without distinction, to other domains, like biology, sociology and economics, where the physical paradigm risks becoming prematurely reductive.. The controversial aspect of Shannon’s definition of information is that it is one of randomness and unpredictability, for which Shannon uses not only the mathematical expression but also the term entropy. Shannon in fact continued the line of enquiry into the mathematical treatment of signal transmission begun by H. Nyquist and R. V. L. Hartley at the Bell laboratories.

He openly declared his debt to Norbert Wiener’s work, who in turn pointed out the innovation that Shannon’s ‘entropic ideas’ represented for information theory (Shannon and Weaver 1964, 3). As Weaver points out in his introduction to MTC, the notion of entropy was already associated with the notion of information in physics, notably in the work of L. Szilard, and had proven useful in quantum mechanics and particle physics, notably in the work of von Neumann (Neumann 1932, chap. V). Yet albeit being fundamental to the physical sciences and engineering, it is easy to see how the notion of entropy becomes a source of confusion when introduced into common language.

There is of course no obvious reason why the concept of information should be introduced in this way into our understanding of physical processes, or why concepts from physics should pertain to our understanding of information1. The biophysicist Henri Atlan already takes this controversy into account, referring to Leon Brillouin’s 1959 Science and Information Theory, but also to the more general problem of the use of intuitive concepts in physics, such as ‘energy’, ‘force’ and ‘speed’, which has been discussed notably by Cornelius Castoriadis, G. Hirsch and -M. Levy-Leblond (Balibar, Lehoucq, and Lévy-Leblond 2005; Brillouin 2013; Hirsch 1976; Lévy-Leblond 1976).

Outside of the purely scientific engagement with the notions of entropy and information, Shannon’s entropic definition of information also provoked and still provokes controversies as being excessively technical and alienating, if not V Physical Concepts of Information and Informational Concepts of Physics An Epistemology of Noise 36 contradictory (Capurro and Hjorland 2003). The term ‘information entropy’ evokes the paradoxical notion that information is reduced to disorder, if not chaos or, on the contrary, that entropy corresponds to the idea of homogeneity and un-differentiation, which is the opposite of what one would normally associate with the idea of a signal or message that stands out against the indifference of background noise.

The mathematical formalization Shannon uses is, as we have seen, almost identical to the way in which Ludwig Boltzmann first formalized the statistical measure of entropy in a physical system, as expressing the average of all its possible microphysical configurations, occurring with equal probability under specified constraints. It is understandable that this notion of ‘information entropy’ is incompatible with what one ordinarily calls information, if ‘information entropy’ evokes simultaneously the ideas of disorder and of homogeneity, and which to boot becomes a measure of the information we lack: Dr. Shannon’s work roots back, as von Neumann has pointed out, to Boltzmann’s observation, in some of his work on statistical physics (1894), that entropy is related to ‘missing information’, inasmuch as it is related to the number of alternatives which remain possible to a physical system after all the macroscopically observable information concerning it has been recorded.

(Shannon and Weaver 1964, 3, n. 1) Yet how can the quantity of ‘information we lack’ correspond to the ‘quantity of information’ we receive? The natural answer to is to say that it is precisely the opposite that is the case, that information is the opposite of the ‘information entropy’, namely its negation, and to explain this with the minus sign that precedes the symbol Σ in the equation used to measure ‘information entropy’: H = – Σ pilog pi This definition of information as negation of entropy has become core to the now dominant definition of information as negentropy.

But to accept this neologism, without enquiring into the theoretical conversion it implies, risks discarding too hastily the philosophical potential of Shannon’s ‘entropic ideas’, thereby risking to transform the concept of information into one of redundancy. To negate entropy, is to negate all possible alternatives, and hence to affirm an identity that cannot change. Another way of solving this dilemma is to say that the more improbable the occurrence of a particular sign is a priori, the more informative it is a posteriori (conversely, if its occurrence was certain a priori, then it would Concepts: Information Entropy, Negentropy, Noise 37 bring no new information a posteriori) (Atlan 1979, 33).

While this appears to be a good compromise, it leaves us with the abyssal question: how do we turn the unexpected, and hence that which we could not anticipate or know a priori, into something we know a posteriori? The difference between a priori and a posteriori is a little more complicated than a mere before and after the fact, if we accept that these terms have been irreversibly conditioned by Kant’s critical philosophy. The idea that we can turn the a priori unknown into what is known a posteriori implies an epistemological conversion that isn’t entirely straightforward.

Let us recapitulate the idea by which the paradox of Shannon’s information entropy could be brought back into the fold: information entropy is what is unknown a priori, but known a posteriori and, crucially, the more unknown it is a priori e. the more unexpected it is), the more knowledge it procures e. the more it informs us in the traditional sense of the word information) a posteriori. Now, if the a priori is a critical term that designates the conditions of possibility of cognition, e. the concepts without which there is no coherent unified experience, and the a posteriori designates that which is experienced on the basis of these concepts, then I am not sure what such a conversion of a priori uncertainty into a posteriori certainty could mean.

It could mean making the absence or indistinctness of concepts (or our uncertainty about the a priori conditions of thought) the prerequisite for our certainty about experience. In other words, it would mean a rejection of Kant’s critical legacy, and a return to dogmatic intuitionism, where experience, if not irrationalism, supplants reason. This, it appears to me, is not a solution to the paradoxical relation of information and uncertainty, but an even greater paradox. If, on the other hand, the terms a priori and a posteriori here do not refer to critical philosophical concepts, but are simply used as erudite terms for ‘before’ and ‘after’ the fact, then we are still faced with a difficult epistemological conversion, namely of the virtual (the purely possible) into the actual.

In this case, the more uncertain we are about the virtual possibilities inherent in a situation, in other words, the more unexpected the evolution of this situation is, the more knowledge this transformation will have imparted on us once it has occurred. This inverse relation of the virtual and the actual may indeed provide a fruitful conceptual framework for thinking about information. However, what is certain, is that such a way of thinking about information is revolutionary by default, if by revolution we mean the radical and unexpected transformation of a situation, (and not a sudden reversal understood as a return to something pre-existing).

Such an approach, whereby the maximal value of information is An Epistemology of Noise the most revolutionary, implies an epistemological attitude to information that could not be further from the idea of negentropy, if the latter is understood as the negation of alternatives. The conversion of uncertainty into certainty is implicit and therefore taken for granted, when ‘information entropy’ is defined as potential information, or as information we ‘lack’. This conversion, however, cannot be the same as a simple actualization of a potential, neither in the metaphysical, nor in the physical sense. The problem of uncertainty and ‘freedom of choice’ and its transformation into the opposite, into information as certainty and constraint takes as its starting assumption what still needs to be explained, namely how knowledge constitutes itself in the face of contingency and what role uncertainty plays in the constitution of knowledge.

What is taken for granted is thus the fundamentally dynamical problem of information at the heart of epistemology: information can only be understood as a process rather than a given, a factum or a datum. Intuitively the idea of quantifying information as ‘bits’ suggests a simple encounter of form and content, as if one could quantify a certain amount of information as when one measures how full a cup is. This risks obscuring one of the most important aspects of Shannon’s entropic concept of information, which quantifies not the individual signal or message, but its relation of probability with the set of all possible messages given particular constraints – such as for instance a string of letters in relation to a finite number of possible letters in an alphabet: a message can be composed of a selection of discrete symbols, which could be letters, words, musical tones or any imaginable other signal, each however belonging to a set of symbols or a spectrum within which there is a certain ‘freedom of choice’ in terms of probability.

Each choice furthermore stands not on its own, but always in relation to previous choices having already occurred in a discrete or continuous transmission of information. The previous state is thus factored into the probability with which the next symbol is chosen as the most likely, in what is called the Markoff process. It is this progressive relation of probability, which turns out, as Weaver says, ‘to VI Information as Process Rather Than Content An Epistemology of Noise 40 be exactly that which is known in thermodynamics as entropy’ (Shannon and Weaver 1964, 12). ‘Information Entropy’ is thus a measure of the probabilities involved in progressing through stages of selection, indicating the probabilities with which, at each stage, certain symbols will be chosen next.

It is thus never, the individual message that is carrier of information, but its relation with the set of all possible messages under equivalent constraints, a relation that changes as the transmission progresses: The concept of information applies not to the individual messages (as the concept of meaning would), but rather to the situation as a whole, the unit information indicating that in this situation one has an amount of freedom of choice, in selecting a message, which it is convenient to regard as a standard or unit amount. (Shannon and Weaver 1964, 9) When Weaver says the ‘unit of information is called a “bit”’, what is thereby quantified is not a signal or message, but a changing relation between the actual and the possible, within a given frame of constraints.

To predict the actual symbol or even message on the basis of the set of all possible symbols or even messages, is to anticipate the relation between a set of n independent symbols and the probability of choice p1, p2, …, pn. It is this relation of probability that finds mathematical expression in Shannon’s formula: H = – Σ pilog pi The calculus of probability therefore measures how rich in entropy information is, in terms of the progressive relation between our ‘freedom of choice’ and our capacity to predict. This means that the quantity of information is never measured as content or amount of the transmitted message alone, but as a function of the relation between this message and all possible messages with equivalent constraints.

Information is thus understood as a dynamical relation of probability that measures a process rather than a content. It is this progressive sequence of probability between the actual and the possible that becomes the raw material of communication, quantified in terms of ‘freedom of choice’ prior to any possible interpretation and evaluation of the message as being significant or not within a semantic context. The habitual sense, in which information is considered like a vessel, a carrier of a certain amount of signification, is thus transformed by Shannon into a measure of the relation between the set of all possibilities, allowing a certain ‘freedom of choice’ in terms of probability, and the probability of prediction based on already actualized choices:

Concepts: Information Entropy, Negentropy, Noise 41 The significant aspect is that the actual message is one selected from a set of possible messages. If the number of messages in the set is finite then this number […] can be regarded as a measure of the information produced when one message is chosen from the set, all choices being equally likely. (Shannon and Weaver 1964, 31) For one, this implies that information is never a given, because it characterizes a progressive modulation of certainty and uncertainty. Information presupposes as essential the structural and operational synergy between context and individual message, as between the uncertainty of ‘freedom of choice’, and the progressive modulation of certainty during the evolution of the individual message.

Information, then, is the progressive unfolding of this relation between uncertainty and certainty.. The philosophical significance of Shannon’s shift of emphasis, from individual signal to process, can be appreciated if we look at it through the lens of a medieval problem that was once known as the problem of individuation. The idea of individuation must ring unfamiliar to the contemporary ear, evoking at best a vague impression of scholastic disputes, perhaps echoing faintly around the names of Aquinas, Scotus or Ockham. Readers of Leibnitz or Wolff will recognize its gradually effacing traces in early modern philosophy.

However, with Descartes and the modern empiricists, the problem of individuation appears to slide into oblivion, and what we are left with, qualifying the object of experience, is the idea of ‘All Things, that exist, being Particulars … ’ (Locke 1975, 409 in Barber and Gracia 1994, 2). The individual has since become the starting point of critical reflection and even the axiom of any possible rationality, be it as the cogito or as the touchstone of empirical investigation. By the same token it has become the stronghold of what we now call information according to common sense.

The individual, as object of experience, is what informs us on ourselves and on the world (constituting either a bundle of faculties or an aggregate of attributes). Even though science has proceeded to dissolve individuality all the way down to quantum fields, and has dismantled any residual faith in its intuitive givenness through the neurocognitive sciences, the individual has nevertheless ossified into a tenacious idea of personhood. The concept of the individual has congealed into a political and moral sine qua non. We like to flatter it, when we qualify the individual as a subject, paying no mind to the pejorative connotation of subjection, which implies that we are subject to other powers, and that we thereby glorify what the Ancients considered a passive substrate to an active principle.

A narcissistic investment in the idea of the individual thus makes it difficult to render this notion unfamiliar once more, or even to recognize the VII To Think about Information as a Process of Individuation An Epistemology of Noise 44 impact it has on our way of thinking about the world, and about what can inform us and how. To question the legitimacy of the individual’s primordial role in the contemporary Zeitgeist is perhaps even threatening to some, as it touches the centre piece of contemporary humanism: the individual and its identitarian reclamations. Yet, what is left of the idea of humanity appears to be a fragmented, hedonistic individualism.

It has the merit of keeping the economy alive with its voracious need to accessorize individuality and to soothe its fear of dissolution with consumption. However, by the same token, the idolatry of the individual also heralds the potential demise of humanity. Biologist Eugene F. Stoermer and atmospheric chemist Jozef Crutzen even proposed to call our current geological era the Anthropocene, indicating that the presence of humans on earth now has the power to catalyse a process of such magnitude that the planetary survival of all forms of life is put in doubt. (Crutzen 2002).

If the question of individuation appeared to belong to the Middle Ages, it may yet acquire a new urgency in light of the consequences of today’s unbridled individualism. It is the singular merit of Gilbert Simondon (1924–1989), to have put the ossified concept of individuality back into motion, by reviving the question of individuation. An atom, a biological cell or, indeed, a person is no longer considered a given, either in the form of a monadic entity or as an always already constituted whole. Instead, whichever entity or term we call ‘individual’ is seen as the end product of a process of individuation, whose most final stage of individualization is but the exhaustion of its potential for further individuation.

Simondon thus proceeds to put the metaphysician back on the school bench. He takes us on a theoretical tour de force through modern physics, biology and psychosocial theory – without neglecting a critical appraisal of information theory and cybernetics, within a wider philosophical analysis of technical reality. He postulates that any theoretical problem or even existential crisis can be said to reflect a field of tension such that, in analogy with an electromagnetic field, whatever discovery, concept or idea is inserted as a new element is at once seized by this field and polarized. The process of individuation is thus compelled by the potential that characterizes this field of tension.

This impetus is comparable, yet subtly different from Aristotelian entelechy, because form no longer predetermines the outcome, but enters in a recurrent causality of form and matter. In an ambitious synthesis of Plato’s concept of form and Aristotle’s hylomorphism, which he updates with the scientific concept of the Concepts: Information Entropy, Negentropy, Noise 45 ‘field’, Simondon revives the formal power of the idea, but embeds it in a revised hylomorphic schema: the structuring or organizing principle of form enters a reciprocal, mutually determining relation with a field of tension. More simply put: what Plato considers a superior reality, the idea or form, is no longer aloof of matter, but seized in a hylomorphic relation.

In turn also Aristotle’s hylomorphism is reformed. Form and matter are no longer abstractly linked, as active and passive principles. The field of tension that receives structuration, comparable to Aristotle’s matter, is itself active: it polarizes and affects the idea or form, as much as it is structured by it. There is thus an embeddedness of the formal power of ideas and concepts in an empirical field. The idea becomes constitutive of this field and its process of transformation, but is also polarized by it. Simondon’s account of the process of individuation thus not only comprises the emergence of form,

e. the gradual or sudden structuration of a domain, but implies also the concurrent transformation of the field itself, whose pre- individual state gives rise to a milieu associated with the process of individuation. Individuation co-evolves with its own milieu. Both the final individual and its associated milieu are thus seen as by-products of a same process of differentiation. Rather than being the first object of consideration, a given, the individual is thus what comes last. Conversely, the milieu is not what precedes individuation – in other words, it is not simply that to which the emerging individual adapts – but is itself a correlate of individuation (Simondon 2005a).

Crucially, what qualifies the genesis of form, for Simondon, is information. Information is not an aspect of the individual alone, such that one could compare the content of a piece of information to the complexity of an individual entity (or signal). Rather, information is whatever catalyses a process of differentiation, comparable to the effect that a crystalline germ has on an oversaturated solution, but information also qualifies whatever modulates this process, amplifying or regulating it. In other words, information is both what triggers a process of differentiation and what acts as an organizing principle. Whatever catalyses and modulates the process of differentiation, as a resolution of tensions or the solution of a problem, can thus be qualified as information.

Information is thus, for Simondon, what links the epistemological and the empirical aspects of individuation. It is the idea or concept (or form) that catalyzes a process of differentiation, that modulates this process and that, hence, informs empirical reality, which in turn polarizes the idea. However, the coupling of reason and experience is also subject to a formal analogy between thought processes and empirical processes, whereby An Epistemology of Noise 46 no norm, no system detached from its content can be defined: the individuation of thought alone can, by accomplishing itself, accompany the individuation of beings other than thought; it is thus not an immediate nor a mediate knowledge that we can have of individuation, but only a knowledge that is an operation parallel to the known operation; we cannot, in the habitual sense of the term, know individuation; we can only individuate, individuate ourselves, and individuate within ourselves; […] an analogy between two operations, which is a certain mode of communication.

(ILFI, 36. Emphasis in the original; my translation) Information thus becomes a key concept for individuation. The new inflections that the concept of information receives by way of information theory and cybernetics are present throughout Simondon’s two main works: his major doctoral theses entitled L’individuation à la lumière des notions de forme et d’information (published as a whole only in 2005 by Éditions Jérôme Millon (Paris) and as yet untranslated),2 and his secondary thesis, written in accordance with academic requirements at the time, entitled Du mode d’existence des objets techniques, which was published as early as 1958 in by Aubier, Flammarion, and whose official English translation has been published by Univocal Press only in 2017 (Simondon, Malaspina and Rogove 2017).

It is in this key role given to a processual understanding of information that we find an affinity with Shannon’s mathematical approach to information as an evolving relation of probability. The philosophical magnitude of Shannon’s processual understanding of information can be gleaned by comparing it to Simondon’s metaphysics and epistemology of individuation: there is no ‘piece’ of information whose quantity could be determined in and of itself, any more than there are individuals that can be abstracted from a process of individuation without rendering them sterile and lifeless. The point that can be made here, without imposing too violent a reading on Simondon’s work, is that information is fundamentally misunderstood, if it is taken to characterize an entity (a piece or information, a message in and of itself) rather than a process, from which the transformation of context cannot be dissociated.

Shannon and Simondon both operate a Copernican revolution, replacing the individual (being or message), traditionally at the centre of the attention, with a careful attention to the co-evolution of both individuation and its context or milieu. However, Simondon’s own critique of the insufficiency of any quantitative concept of information must be born in mind. It would be wrong to Concepts: Information Entropy, Negentropy, Noise 47 suggest that Simondon’s theory of individuation lends itself to an appraisal of Shannon’s concepts of ‘information entropy’, and especially to the philosophical revaluation of the concept of noise, such as it is argued for here.

Yet without seeking to harmonize the difference between Simondon’s concept of quality of information and Shannon’s ‘information entropy’, we can still find resources in Simondon’s theory of individuation which enable us to shed new light on Shannon’s conception of information as a progressive relation of probability. For one, Simondon enables us to grasp the philosophical enormity of a concept of information that puts the individual (message) last, and brings to the foreground a progressive relation of probability between the individual state of the message and the set of all possible messages under a given set of constraints.

To think with Simondon, without thinking exclusively in Simondonian terms, thus helps to critically redress our understanding of information, and notably to reject the common conflation of information with ‘data’, understood as something given. However, Shannon’s definition of both information and noise as entropy, distinguishing only desirable from spurious uncertainty, no doubt strains the analogy with Simondon’s theory of individuation. For Simondon the role of information remains essentially one of differentiation and structuration. It is ultimately based on an opposition to entropy, understood, as Wiener and Brillouin do, as the final state of equilibrium or the death reached by a closed system.

Although, technically speaking, Wiener’s concept of negentropy is based on the same mathematical model as Shannon’s, imposing an understanding of information as process rather than entity, in cybernetics it is always in the service of an already constituted and correctly functioning entity, a machine or an organism, that information is required to counteract entropy, notably through feedback processes. This is not to say that Simondon settles for Wiener’s definition of information as negation of entropy, any more than Shannon’s. Both fall short of providing a qualitative definition of information, such that it could encompass individuation.

What Simondon finds lacking in the quantitative definition of information is the notion of potential, the tension that polarizes and thus gives a sense, if not a signification to information. Simondon rejects the idea that information, measured in bits, could in any way encompass what we must understand by the quality of information, which is what characterizes not only the capacity to inform or regulate reality, but, crucially, its capacity ‘to illuminate new domains’ (Simondon 2005, 549). An Epistemology of Noise 48 Simondon does, however, incorporate a great number of aspects into his theory of individuation, which would appear to lend themselves to a revaluation of various aspects of noise, such as we are aiming at here, by relativizing its opposition to information.

He has a subtle understanding of demodulations of structure, even of crisis, which he sees as a necessary reload of potential for novel structuration and reorganization. He mentions, for instance, processes of dis-adaptation in developmental psychology, as necessarily preceding reorganization at a higher level (Simondon 2005, 545). Several key terms in his philosophy express a deep appreciation for indeterminacy or even dissolution of structure or form. In his On the Mode of Existence of Technical Objects he values margins of indeterminacy in the functioning of so-called open machines, as necessary for their capacity to respond to input from the environment.

In his lecture Communication et Information Simondon even delves into the motivating or bonding aspect of social noise (Simondon, Simondon and Chateau 2010; Simondon 2010). Furthermore, Simondon’s notion of a pre-individual state lends itself to an analogy with noise if we think of the pre-individual state of being as characterized by the equal probability of all possible states, and thus rich in the greatest possible ‘freedom of choice’. As such a primordial noise could now be thought, in light of Simondon’s concept of the pre-individual, as a positive ground for differentiation, in other words, as ground for the emergence of form and its transformation.

This would mean that a primordial maximal uncertainty not only grounds but co-evolves with the process of individuation. Noise in the ordinary sense, understood as interference, would thus be the remaining uncertainty, tethered to the process of individuation by its associated milieu. What Simondon calls the ‘pre-individual’ state of being is thus not entirely alien to our idea of a subtler difference between ‘information entropy’ and noise entropy, both understood as a margin of ‘freedom of choice’ or uncertainty. However, even this limited analogy requires an important proviso. The analogy works only if both concepts, the ‘pre-individual’ and ‘maximum entropy’, are treated as regulative ideas, acknowledging that no situation is absolutely closed, in the sense that the absolute absolves from any relation to an outside.

There are approximations of absolute closure of a system (for instance, in nanotechnology or quantum technology) that come at the cost of a technological apparatus and of mathematical constructions of great sophistication. However, no process of individuation and no process of information ever takes place on absolutely blank slate: nowhere in the empirical world is a closed system Concepts: Information Entropy, Negentropy, Noise 49 realized in absolute terms, e. without having to take into account the role of the observer or the impermanence of this closure. Every system known to mankind is always already situated in a reality that is densely packed with pre-existing processes of individuation.

To acknowledge this is to admit that empirical reality is always already a noisy mess of competing processes of individuation, involving dissolution of form and a wrangle of certainty and uncertainty. However, even if we acknowledge that no process of individuation is absolute and that multiple processes of individuation may compete noisily, it must also be acknowledged that Simondon explicitly stops short of an affirmation of noise, (or of phenomena involving de-differenciation, indeterminacy or even chance), as constitutive of information processes. Simondon compares for instance the tension of form, which he sees as a precondition for the quality of information, to social phenomena such as pre-revolutionary tensions.

It is conceivable that in such situations, he says, a ‘thought coming from elsewhere’ (le fait qu’une idée tombe d’ailleurs) triggers a sudden structuration (ILFI, p. 550). Just as a chance correlation of molecules may set off the process of crystallization, so a ‘chance encounter’ may set of a revolutionary process. However, and this is crucial, ‘it is very difficult to admit that chance has a value of creation of good form’). This is because the quality of information is more than a fortuitous aggregation: its structuring effect must be more than just fleeting, it must sustain a structuring power, and sustain what in French is called sens and which we can only partially translate as both signification or direction.

The quality of information mediates information’s power of structuration and the tension that characterizes a domain capable of receiving information. This mediation is quite literally the sense that information makes, the signification or organizing power it catalyses. A purely fortuitous process, in turn, would be subject to an equally fortuitous dissolution. In other words, what we could call a noise phenomenon, and which Simondon here characterizes as coincidence or chance (hazard), may act as a trigger for spontaneous structuration but – and this is the crux – the process of information itself remains a process of structuration for Simondon and ultimately a negation of entropy.

While this appears to cut short any analogy between Shannon’s conceptual audacity and Simondon’s return to the problem of individuation, it must be born in mind that Simondon is already several steps ahead of the problem we are addressing here. His emphasis on the quality of information is already a problem of signification. For us, on the other hand, what is at stake is a rather limited An Epistemology of Noise 50 problem that does not yet encompass the question of signification, but only the presence of uncertainty among its conditions of possibility. What Shannon enables us to think is not an absolute value of noise as novelty – which one could provocatively call ‘pure information’, if one were to attribute a maximal information value to maximal entropy.

It is, rather, the fact that we can now think of information as a subtler difference than that between organization and chaos or sense and non-sense, a difference that takes place within the conceptual space of entropy, within the space of uncertainty: if information can be thought as qualified uncertainty, then noise too can be released from the theoretical exile of negation into which it was thrown. Noise can become possible information. In other words, unqualified uncertainty can be understood as one of the preconditions of qualified uncertainty and, hence, of information. Information entropy, understood as ‘freedom of choice’ or as equi-probability of events, is thus never absolute, especially when we consider that this term qualifies the quantity of information not merely as ‘entropy’, but as ‘entropy of the message’.

‘Entropy of the message’ designates a variability that is always already limited by the condition that the source of information continues to employ the same set of symbols and that this set is finite. It is a ‘relative entropy’ that implies a certain amount of redundancy, in other words of repetition (frequency) giving it a head-start on the margin of predictability. As Weaver puts it: If the relative entropy of a certain source is, say 0.8, this roughly means that this source is, in its choice of symbols to form a message, about 80 per cent as free as it could possibly be with these same symbols.

(Shannon and Weaver 1964, 13) If the relative entropy of a source (of continuous signal transmission, like for instance a radio transmission) is given a value of 0.8, then the remaining 0.2 corresponds to constraints that are placed upon the message, in other words, to what makes this message minimally predictable and hence, to what will be redundant within it: One minus the relative entropy is called the redundancy. (Shannon and Weaver 1964, 13) Constraints on the entropy of the message can be, for instance, statistical rules governing the use of symbols, or the set of letters in an alphabet or syntactical rules.

The predictable part of the message is what can be reconstructed and is therefore considered to be inessential to the novelty of the message, and in this sense ‘redundant’. It is what separates the ‘entropy of the message’ from complete randomness or noise. Interestingly Weaver goes so far as to call the redundant part of the message unnecessary, which appears to suggest that the message can still be a message without it: VIII Redundancy and Necessity An Epistemology of Noise 52 [T]his fraction of the message [that] is in fact redundant in something close to the ordinary sense; that is to say, this fraction of the message is unnecessary (and hence repetitive or redundant) in the sense that if it were missing the message would still be essentially complete, or at least could be completed.

(Shannon and Weaver 1964, 13) Weaver is of course right in the sense, for instance, that most vowels can be left out of a typed message, without making it impossible to reconstruct the message. The journalistic convention of replacing letters in offensive words with the symbol * e. f***) is indicative of the fact that what is redundant need not be reiterated. Redundancy is nothing other than the predictable part of a message. Weaver indeed goes on to note that language has a very high level of basic redundancy: It is most interesting to note that the redundancy of English is just about 50 per cent, so that about half of the letters or words we choose in writing or speaking are under our free choice, and about half (although we are not ordinarily aware of it) are really controlled by the statistical structure of the language.

(Shannon and Weaver 1964, 13) The conceptual presence of necessity here slips into Weaver’s expression nonchalantly, in the form of a negation: redundancy is the part of the message that is not necessary. However, to define redundancy as unnecessary, meaning inessential or accidental, could lead to a misunderstanding of far reaching theoretical consequences. While Weaver appears to say something obvious, the concept of necessity is one that cannot leave philosophical analysis of information and noise indifferent. Weaver’s way of putting it, namely that redundancy is the part of the message that is not necessary is potentially misleading, not least because his introduction seeks to lay the conceptual foundations for a new understanding of the broader theoretical relevance of MTC.

It is a far from negligible slippage of logic to describe redundancy as unnecessary, because it shows and even performs that the necessary, that which cannot not be and which constrains ‘freedom of choice’, is what, as self-evident, can be left unsaid and hence un-thought. What does it mean for the redundant part of the message to be unnecessary? The necessary is, in simple terms, whatever is absolutely indispensable and hence of utmost importance (as for instance in the expression of the ‘bare necessities’ for survival). Necessity can also be understood as a constraint, such that its stringency or unavoidability is recognized in law even where a necessity contravenes the law, as in the expression ‘state of necessity’: the necessity to

Concepts: Information Entropy, Negentropy, Noise 53 safeguard the interests of a person may, before the law, result in the impunity of an incriminating act (dir. Jaen-Marie Pierrel et al.). Necessity thus designates what is required by a situation (material, practical, technical or vital necessities), but more fundamentally, in the philosophical tradition, that which cannot not be, or which cannot be otherwise. In other words, necessity is the mother of all philosophical concepts: a categorial, logical or metaphysical necessity is what reason posits as valid in any circumstance and whose contradiction is an impossibility. For reason, necessity is nothing less than the axiomatic starting point of rational thought.

Everything else has been, since Greek Antiquity, attributed to the order of opinion, of mere phenomena or appearance. What is not necessary is contingent: either absolutely contingent or contingent upon a necessity that we may or may not know. The introduction of probability into reasoning is, therefore, a significant event in the history of thought. That something can be said to be 0.2 per cent certain and 0.8 per cent uncertain introduces the possibility of nuance and process: genesis and corruption are no longer excluded from the realm of reason. It is, in epistemological terms, the metaphorical equivalent of introducing colour into a black and white vision of truth.

All the more reason to take note that necessity is what Pascal, one of the founding fathers of the calculus of probability, called a ‘state of constraint or restraint that annuls freedom of choice’ (dir. Jaen-Marie Pierrel et al.; Pascal and Guern 1987). Now, if redundancy is the part of the message that imposes a constraint, that reduces ‘freedom of choice’ in terms of the message’s probability, then it is hard to see how it could be unnecessary to the message. Redundancy is what in the message remains constant, what is stable and not subject to degradation through noise.

Redundancy, in other words, is the very state qualified by the Latin root of necessity: non cedens, that which does not give in and which, in its regular form necée, is close to the idea of chastity: untouched by genesis and degradation. Is redundancy in the message not precisely that which remains untouched by ‘freedom of choice’, by entropic degradation of the message, by contingency, in short, by noise? Redundancy, without which ‘information entropy’ would be indistinguishable from noise, is thus not only necessary to the message, it is what, as self-evident, becomes the invisible or unthinkable a priori of information.

The consequences of underestimating redundancy as unnecessary are far from trivial, if we acknowledge that every form of organization is based on constraints that introduce redundancy. Every system is informed by constraints that discriminate ‘freedom of choice’ according to given (hence redundant) criteria of pertinence. An Epistemology of Noise 54 To euphemize redundancy as unnecessary thus risks losing from sight what founds every system: necessity, in other words that without which a system cannot be what it is. By becoming redundant in the sense of self-evident, necessity is what can remain unspoken and hence un-thought.

Necessity, that which makes a system what it is, is thus also what most easily subtracts itself from critical analysis, as an a priori of this system, of this organization, of this way of functioning, that is always already taken for granted. In linguistic terms, the redundant part of language is what can be taken for granted, what can be ignored. But we must not overlook the fact that it can afford to appear unnecessary, it can afford to be ignored, only as long as it operates as the condition sine qua non of communication.

Redundancy, contrary to Weaver’s claim, is necessary – and only therefore is it not necessary to reiterate – just as good manners do not need spelling out where they can be taken for granted, and just as the work of illegal immigrants can remain invisible to the public, while operating as a condition of possibility for an economy hungry for cheap labour, unregulated by workers’ rights. The question could be transposed onto the idea of work redundancy. It is especially relevant in view of the digital revolution, which exposes entire segments of professional activity to potential redundancy.

The question, as with redundancy in language, is this: which a priori are being taken for granted? What goes without saying? A critical approach would be to seek the blind-spot in redundancy, in order to debunk mere preconceptions masked as necessity. When the necessary conditions of the information process, of communication more generally and of organization (including social or political) are what is redundant, then they are nothing less than the a priori of our way of thinking and acting. Yet if we fail to address the informational value of redundancy by minimizing it as ‘not necessary’, then it becomes increasingly difficult to ask: when is an a priori a necessity, a sine qua non of being thus, of thinking and of communicating thus?

And when is it mere prejudice? The a priori restriction on the ‘freedom of choice’ in the message, is nothing less than the condition of possibility of communication, also because, without it, nothing would offset the uncertainty that is a correlate of the novelty of information. In other words, without redundancy the pure novelty [entropy] of information would be absolutely incomprehensible and equivalent with noise. It is only on the basis of redundancy that novelty demarcates itself from what is already certain. Redundancy is, furthermore, in this sense, also an essential concept for our understanding of physical entropy.

For the measure of entropy in a physical system is a direct correlate of the knowledge we already Concepts: Information Entropy, Negentropy, Noise 55 have of it and the knowledge we lack. The knowledge we have of a system, for instance of chemical rules of interaction between elements, acts as a constraint in epistemological terms: it reduces the entropy of the system. Without this knowledge the behaviour of the system is absolutely random to us. Conversely, what we call complexity is a correlate of low redundancy, in other words, of a low level of pre-knowledge about the system.

By complexity we must understand the degree of indeterminacy of a system, rather than its level of structural complication. Biophysicist and philosopher Henri Atlan defines complexity as the measure of the observer’s ignorance as to the precise determination of a system. Greater complexity of information denotes greater uncertainty. What he calls the ‘maximum maximorum’ of ignorance is the state of greatest complexity. It corresponds to the most basic measure of information in Shannon’s sense (H), which informs us only about pure multiplicity, nothing but the number of elements in a system (H= log N). Atlan calls this the first, ‘trivial and maximal’ measure of complexity.

It corresponds to the observer’s maximal ignorance of other factors, such as variety, frequency and other constraints. The second measure of information takes into account statistical distribution and frequency (H= Σ p log p). Its quantitative value is therefore smaller than the first, as its complexity is reduced. The third measure of information, finally, introduces redundancy through the addition of constraints [H = Hmax (1 – R)]. This corresponds to the least complex level of information, as determining factors carve away at the complexity of the pure multiplicity that characterized the first and maximal level of complexity.

(Atlan, 1979, p. 80).. In logic, deductive redundancy is achieved when each proposition is tightly correlated with the propositions deduced from it. Step by step, each proposition becomes resonant with the others and no element can be modified without compromising the whole (Blanché 2009, 10). The philosopher of mathematics Robert Blanché thus describes the process of deductive thinking as propagating a structure of constraint where, like the emerging lattice structure of a crystal: [S]tep by step, a tight network is constituted where, directly or indirectly, all propositions communicate. (Blanché 2009, 9–10) This is why one speaks of the deductive resonance of a mathematical theorem if, like a crystalline structure, it achieves a rock-solid correlation between each and all terms, where nothing is left to chance and no ambiguity can arise.

The logical coherence of the whole can then be called isomorphic, like the lattice structure of a crystal. This resonance is a form of redundancy in the logical chain. It eliminates ambiguity, just as redundancy in the transmission of a message serves to reduce noise. It was long hoped that logic would provide a language without ambiguity, in other words without noise. If the elimination of ambiguity could ensure the truth of all statements in a logical chain, by eliminating ambiguity and hence error, then deductive redundancy was deemed necessary in order to stabilize philosophical discourse.

If, however, all statements are not bound by necessity then they retain a trace of ambiguity. This is the case for unproven postulates, in other words, for all statements that are not yet proven axioms. A single postulate in a system of axioms becomes the gateway to a potentially radically diverging axiomatic system. A paradigm shift thus lies dormant in each postulate, in each trace of ambiguity and openness. This ambiguity and openness is what we will call noise, because it persists as a margin of uncertainty and freedom of choice, until it too IX Logic and Freedom of Choice

An Epistemology of Noise 58 is axiomatized. The fifth postulate in Euclid’s Elements long remained open to such uncertainty, because of the impossibility of proving it. According to this postulate, also called the ‘parallel postulate’: If a line segment intersects two straight lines forming two interior angles on the same side that sum to less than two right angles, then the two lines, if extended indefinitely, meet on that side on which the angles sum to less than two right angles. The impossibility of proving this postulate despaired mathematicians for centuries. The uncertainty it represented revealed itself as a freedom of choice only when, in 1829 Nikolai Ivanovich Lobachevsky could prove that the fifth postulate does not hold true in a geometry of infinite dimensions.

The search for the impossible proof of Euclid’s fifth postulate thus paved the way for the emergence of so-called hyperbolic or ‘non-Euclidian’ geometry. The classical three dimensions of Euclid’s geometry opened up to the infinite dimensions of Lobachevsky’s geometry, and as a consequence, also the space-time of classical physics could eventually open up to the possibility of special space-time relativity (Blanché 2009, 47). Eric Temple Bell called Lobachevsky the ‘Copernicus of all thought’, and saw in Lobachevsky’s work an incentive for mathematicians and scientists to ‘challenge other “axioms” or accepted “truths”’ (Bell 1986). For us, the ambiguity of Euclid’s fifth postulate can be thought as a borderline case, where uncertainty becomes freedom of choice, in other words, where noise becomes information.

If Euclid’s fifth postulate represented a source of uncertainty for centuries, Lobachevsky transformed this uncertainty into freedom of choice, opening up three dimensions to infinite dimensions, and opening up the idea of mathematical truth to the emergence of new paradigms. In the early twentieth century, Georg Cantor set out to harmonize the domain of mathematics and to rescue the unity of mathematics in light of the plurality of geometries. In the attempt to axiomatize set theory, Cantor encountered foundational paradoxes that shook mathematics to its core no less than Lobachevsky’s achievements in geometry, and thereby irreversibly unmoored its theoretical capacities from its last anchorage in intuition.

Mathematics and geometry were henceforth wide open to the prospect of infinite or transfinite sets of infinite numbers, chaperoned only by the paradox that the set of all stets cannot ground itself. In response to the crisis in the foundation of mathematics that was provoked by Cantor, logic, it was hoped, would reign mathematics in and prevent its speculative excesses. Concepts: Information Entropy, Negentropy, Noise 59 Robert Blanché gives an account of the axiomatization of mathematics that is valuable in this regard, because it helps understand both the necessity and the absurdity of deductive redundancy, when taken to its extreme and when erected as the sole pillar of truth.

Logic, it was hoped, would rescue mathematics from the paradoxes of axiomatic set theory, which appeared to have turned mathematics into a science where one never knows what it is that one is talking about, nor whether it’s true. […] By proposing to ground […] the entire edifice of mathematics on logic, Frege and Russell’s ‘logicism’ aimed further than returning to its principles: it intended bringing it to its term, reaching the rock, the ultimate foundation. (Blanché 2009, 70) It is history, by now, that rather than eliminating the antinomies that had sprung up within axiomatic set theory, disagreement about the validity of logical principles were, in turn, to shake also the foundations of logic to their core, ultimately demolishing ‘the idea of an absolute, unique and universal logical legislation’.

Blanché reconstructs how the axiomatization of logic finally led to the ‘disintegration of logic from within’, issuing forth into a pluralization of logics. Even if the intra-logical and axiomatic problems raised by a plurality of logics and of axiomatic systems could be set aside, what remains problematic is thus the ‘fit’ of logical redundancy and reason’s intrinsic complexity. Its formalized terms, albeit submitting to the utmost criteria of necessity, no longer represented anything but the mediation between ‘simple tautologies’. Although perfectly harmonized in a deductive redundancy, without what we may call the noise of ambiguity, they ‘say strictly nothing about the real, but […] for this reason, remain valid whatever content one applies’ (Blanché 2009, 71).

In a different context the philosopher of biology Marjorie Greene struggled with the limitations of Logical Positivism in conceiving of the problems specific to biology. Its incapacity to deal with the imprecision of the empirical world eventually turned Greene away from Logical Positivism, in which she saw a sterility when it came to the life sciences: In the Anglophone tradition (which I derive partly from the Germano-Austrian tradition) that which one called the received view dominated until recently. I participated personally in Carnap’s seminar in Chicago during the year 1937– 1938. Having previously studied zoology, I was rapidly disappointed.

It seemed impossible to treat the praxis of zoology with a purely extensional logic. I tried to explain this difficulty to Hempel, who was Carnap’s assistant in this seminar and An Epistemology of Noise 60 he replied: ‘We only say what we can explain with ’ […] twenty-five years later […] logical positivism was taught under its new less aggressive name of logical empiricism. It dealt with laws, theories, the deductive relation between theories and laws, the problem of confirmation etc. […] Today however, this old orthodoxy is, if not entirely buried, then in a – how shall I say – catatonic, vegetative state.

(Grene 2007, 24–25) Greene’s disappointment in articulating the deductive power of Logical Positivism with the open field of biological complexity is directly relevant to our interest in the question of ‘epistemological noise’, to the unavoidable ambiguity and distortions of concepts and theories, when we seek to cross- fertilize fields of knowledge and praxis. It shows that ‘epistemological noise’ is not merely a deplorable side effect of interdisciplinary communication, or of the transdisciplinary formulation of scientific problems, but a positive requirement, without which specialized discourse becomes not only sterile, but worse, redundant. The question that has emerged from the previous sections is: how do we draw the line between constraint and ‘freedom of choice’?

We cannot avoid complexity and ambiguity entirely without risking sterility of information, but we still need to impose a boundary between a level of complexity relevant for the formation of knowledge and infinite complexity, which consigns us to the power of oracles, not reason. Here lies the difficulty in distinguishing between the ‘entropic’ understanding of information and the entropy of noise. Yet this is precisely the question that is suspended when the engineer transmits a readymade message, regardless of whether it is the rambling telephone conversation of someone’s mother-in-law, an encrypted message, or Schönberg variations.

The a posteriori evaluation of what is received, as either spurious or significant, literally doesn’t come into the equation. Any value judgement that discerns a message with the mark of distinction, namely that it is informative, is either pre-given, a priori, in the decision to transmit a message as information, or it occurs a posteriori, as the result of a process of evaluation and interpretation. The relatively restrictive scenario, where an already selected signal is transmitted as information has this particularity: the decision upon which we base the distinction between information and noise has already been made when the message is transmitted.

The engineer treats the message as ready-made, but in order for information to be transmitted, in order for the entire technological apparatus of signal transmission to exist, a decision must have preceded transmission: namely the decision that information must be transmitted, and the decision that what is transmitted is information. Even if this is not a decision that concerns the communication engineer, it is a decision that needs to be made, repeatedly, every time a message is selected for transmission as information.