This paper was written
originally for the 1997 Keele Conference on Actor Network Theory. I requested that it not be published in
the book of that conference.
Cite or use only with
author's permission; @ Adrian Cussins
5/30/97; minor changes January 2001
Norms, Networks, and Trails
— Relations between different topologies of
activity, kinds of normativity, and the new weird metaphysics of Actor Network
Theory. And some cautions about
the contents of the ethnographer’s toolkit —
Adrian Cussins
Adrian@haecceia.com
§1 Actors and Networks
There are two parts to actor network theory (ANT); a part labelled by the term ‘actor’ and a part labelled by the term ‘network’. Consider ‘actor’ first. For ANT, actors are heterogeneous; they include non-humans as well as humans, instruments as well as technicians, nature as well as society. Moreover, ANT’s actors have a hybrid ontology: they are actants, loci of behavior and activity, or quasi-objects / quasi-subjects. What this means, in a very few words, is that ANT allows that the ontology of actors is up for grabs: what gets to be an agent, what gets to be a human, what gets to be an object, what gets to have the power to change events, what gets to have intentionality, is part of what has to be explained. ANT must track these changing ontologies of agency and intentionality, of subjecthood and objecthood, and must describe—for any ethnographic site—how its ontology gets fixed and how it changes. Since a metaphysics is a way in which an ontology is fixed, or a way in which an ontology changes, there is a sense in which, for ANT, metaphysics is an explanandum. A theory of science and technology cannot take its metaphysics as given; not as given by theology, nor by a priori transcendental philosophy, nor by a philosophy of naturalism, nor by ‘commonsense’, nor by some ‘perfected’ or ‘ideal’ Science. Actors can change locally the metaphysics of the world. Therefore we must face the following, perhaps disconcerting, consequence of ANT: that in being a student of science and technology one is thereby a metaphysician. With all this, outrageous as it is, I agree.
But then there is the part labelled ‘network’. Technical and scientific activity has spatial structure in terms of which a theorist or ethnographer can understand the course of activity. ANT assigns that structure the single spatial topology of a network or networks. And, most importantly, for ANT the concept of a network is the basic explanans. It is, if you like, the most important part of the toolkit that ANT supplies to the ethnographer, as she enters the field: Follow the Networks! So now we have an explanandum and an explanans: metaphysics is to be explained by examining networks of activity. ANT is right about the explanandum[1], but gets into some difficulty over the explanans. Scientific activity does sometimes have the structure of a network, but this too is something that must be explained.
That networks are something that require explanation, and cannot therefore be part of the basic toolkit that ANT supplies to the ethnographer, is established in this paper. What, then, takes the place of networks in the basic toolkit? Here I make some suggestions.
The difficulties arise in those versions of ANT in which the only topology for epistemic activity, or at any rate the only topology which is good for epistemic activity, is the topology of networks. Let us call these versions “ANT(1)”. Like Mol and Law (1994),[2] I will suggest an alternative topology for the basic explanatory structure of scientific and technical activity. Unlike Mol and Law’s fluid metaphors, I will be recommending a topology of trails[3]. There are networks, but they are built out of trails, and if we want to understand how networks are maintained, and how they break down, we need to explore trails of epistemic activity. We’re not just trading metaphors here, because—as I will show—a notion of trails and a notion of networks give us two very different conceptions of what science and technology are about. The reason has to do with normativity. Because trails and networks are very different ways of getting around and structuring the environment — and because the relation between trailblazer and trail differs so importantly from the relations between ‘user’ and network — their theoretical analogs entail very different accounts of the goals and the structure of scientific and technological activity. The two parts of ANT(1)—actant metaphysics and network topologies—are in some tension: I will be suggesting that ANT(1) cannot retain its metaphysical insights if its model of the point of knowledge-work is based exclusively on networks.
The paper begins with some introductory remarks about normativity and its metaphysical commitments (§2) before introducing the distinction between trails and networks (§3) as having contrasting normative and topological structures. What kind of normativity would science and technology have if networks were the basic explanans? What kind of normativity would science and technology have if multiple topologies are possible, none of them having a privileged normative role, but trail-theory having a special explanatory role? In §4 I develop and criticize the metaphysical and ethnographic consequences of network-based theories. In §5 I consider the possibility that a theory based on trails of epistemic activity can support an account which is constructivist, realist and historicist.
§2 On the importance of normativity and the distinction between pre-emptive realism and constructive realism
When ANT is criticized by some as favoring might over right, it is being criticised for losing normativity in an explanation of science: as if ANT is claiming that what works in science is just a question of might. When a familiar kind of philosopher of science criticizes ANT he does so because the norms of truth and rationality play no role in the explanations provided by ANT. More traditional historians and psychologists are often perplexed because the cognitive norms—of genius, insight or brilliance—play no role in ANT’s explanations. And when the authors of “Leviathan and the Air Pump” criticize ANT, they do so because the norms of morality and the virtues—such as the virtue of trust—play no role in the explanations of science and technology. Normativity, then, must be central to any proper assessment of the merits of ANT, yet the literature around ANT provides surprisingly little explicit consideration of normativity. I want to be explicit about normativity, to use it in my assessment of ANT, but to argue not that ANT loses the norms of right, truth, rationality, intelligence, objectivity and morality, but rather that, as things stand with the theory, we should be cautious about accepting and acting on ANT’s conception of these virtues.
Why talk about normativity at all? Normativity is the central concept in the intepretive sciences. Interpretation is about making sense, and to make sense of something we have to see its point. But we can only understand the point of an activity through understanding its normative structure: normative structure is the structure of the point of an activity. What distinguishes a merely physical process from a process with which the hermeneutics of interpretation can get to grips[4] is that the latter but not the former is normative. If a Martian scientist, who is trying to understand the human practice of playing chess, attempts to understand the activity entirely in terms of syntactic regularities of movement of thus-and-so shaped pieces on a board, then the Martian will understand nothing about chess. (Martians are merely philosophers’ analogs of the anthropological outsiders of Science and Technology Studies (STS). Anthropological outsiders, including STS ethnographers, are sometimes as prone as the Martians to offer non-normative, purely descriptive accounts of systems of activity). Chess is essentially an activity organised as a game in which the players compete to win. Winning is the primary norm which not only governs the activity of chess-playing, but which is constitutive of the game of chess. On any individual game, the players may not try to win, but that is possible only against the background of the mutually-understood activity of chess-playing-as-striving-to-win. What are chess players doing? They are trying to win. Interpretation must, in one sense, start with this central fact, for without it we lose the point of the activity. Without the norm of winning there will be no significance to any board position, no sense in which one position is better than another (better for what?). Even computer programs for playing chess are based around a metric for determining relative evaluations of board positions; the very idea of such a metric rests on a conception of chess as a normatively structured activity. The norms of most human activities are not as well defined as those of chess, nor are they bounded in the same way, but it is true of any activity which is a proper object for an interpretive science, that rendering that activity intelligible is rendering the structure of its guiding normativities. If one gets wrong the guiding normativities, then there is a breakdown in interpretive understanding. Consider, for example, “Dr. P.” as described by the neurologist Oliver Sacks (the “anthropologist from Mars”):
“What is this?” I asked, holding up a glove.
“May I examine it?” he asked, and, taking it from me, he proceeded to examine it as he had examined the geometrical shapes.
“A continuous surface,” he announced at last, “infolded on itself. It appears to have” — he hesitated — “five outpouchings, if this is the word.”
“Yes,” I said cautiously. “You have given me a description. Now tell me what it is.”
“A container of some sort?”
“Yes,” I said, “and what would it contain?”
“It would contain its contents!” said Dr. P., with a laugh. “There are many possibilities. It could be a change purse, for example, for coins of five sizes. It could ....”[5]
(Anthropologists, talking about the cultures they study, can often sound like Dr. P talking about a glove. How should we understand what goes wrong here?). Dr. P. is doing something with great facility and aplomb, but his failure to grasp the constitutive normativity of the glove is a gross failure of interpretation which is simultaneously sad and comic. It defines him as an outsider in the community of understanding. The point is that even the most basic levels of undersanding — the manipulation of everyday objects — consist in grasping the normativities that are appropriate to the situation. Members of a practice act as they do because of these same norms, and wherever this is so there is a proper object for an interpretive theorist.[6]
To put this point as baldly as possible: the central, constitutive fact in understanding the relevant practices is that, in one case, the activity is structured around the norm of winning, and in the other, the identity of the object is fixed by its functionality as a glove. And is it not compelling to continue this line of thought by saying that the central fact in understanding the practice of science is that it aims at significant truths about the world. That science is governed by truth as chess is governed by winning?
And now alarm bells are ringing in STS ears, for the norm of winning in chess is fixed by the rules of chess independently of any ethnographic knowledge of situated activities of chess playing. We explain the situated activity of particular players playing chess by reference to the norm of winning at chess, but we do not explain what it is to win at chess by reference to the situated activity of particular players playing chess. (The norm of winning in chess is explained by the context-independent rules of chess). That is, the governing norm of chess is explanatorily prior[7] to the situated activity of playing chess. Whenever it is correct to make a claim of this kind about an activity, the metaphysics of the activity is that of pre-emptive realism. If the governing norm of an activity is taken by a theorist to be explanatorily prior to the situated and historical dynamics of the activity, then the theorist is adopting the metaphysics of pre-emptive realism: the ethnographer of the activity must use his or her independently given understanding of the governing norm of the activity in order to make sense of the ethnographic data. For the pre-emptive realist, the norms that govern activity are explained prior to a local or historical account of activity and skilled practice.
Pre-emptive realists about science are theorists who suppose that science is governed by truth as chess is governed by winning. That is, for such theorists, truth-in-science is explained prior to the situated and changing activities of doing science. Consider, for example, the following recent claim: “On my conception, the cognitive goal of science is to attain significant truth ... This goal is independent of field and time, independent of how we think it might be achieved. ... On my account, the goals of science do not change over time.”[8] That is, the governing goal of science is independent of the practice of doing science (“independent of field and time, independent of how we think it might be achieved”), but in order to understand the practice of science we have to understand it in terms of the goal of attaining significant truth. This is a pre-emptive position because the structure of normativity pre-empts the range of possibilities for scientific activity and practice, in a way similar to how the rules of chess pre-empt what is possible on the chess board. Normativity is prior to activity.
Don’t be insensitive to the motivation for the priority of normativity to activity! If the ‘rules’ don’t pre-empt what is properly possible in the ‘game’, then the 'rules' become part of what is negotiated by the 'players'. If the 'rules' become part of what is negotiated by the 'players', then we end up with the comical but also absurd activity of "Calvinball" from the Calvin and Hobbes cartoon strip. [See figure 1]. The dialectical position is this: unless we agree with the pre-emptive realist, it looks like we’re putting STS in the unfortunate position of both Dr. P. and Calvin. But if, reacting against this kind of 'anything goes', we agree with the pre-emptive realist, won’t we lose the constructivist insights that STS has won over the last twenty years?
The way out here is to recognize that there are different kinds of normativity. (It is possible to be a normative system without being like chess). I want to distinguish between “elite norms” and “mundane norms” so that I can express the position of “constructive realism”. Elite norms include truth, reference, objective existence, rationality, winning, being efficient or well-designed, intelligence, disinterestedness. By contrast, mundane norms include such matters as skillful hand and eye coordination, cleanliness of pipettes and beakers, local coordination of communication across representational interfaces, pulling up ones’ socks (metaphorically and literally), expertise in navigating complex social, class-based, gendered and material environments. Getting around, getting about, getting up and getting through are the domains of mundane norms; the commonplace norms of everyday, mundane activity, and — as STS has taught us — of the mundane activity of elite environments of scientific laboratories and centers of technology.
An ethnographer can characterise an epistemic activity system in terms of mundane normativity (but not in terms of elite normativity) without making any metaphysical commitment as to how the distinctions between nature and society, between world and mind, between non-human and human, or—in general—between objectivity and subjectivity, are drawn in the system. The application of elite norms, by contrast, is only possible given some way of drawing the distinction between the objective and the subjective. For example, one can only say of something that it is true or false relative to a distinction between how things are and how things appear which is or is assumed to be already in place. One can only talk of so-and-so’s brilliance relative to a distinction between mind and world. One can only talk of efficient or well-designed technologies relative to some distinction between society and nature. For this reason I sometimes call the elite norms “the norms of objectivity”: the elite-mundane distinction is really the distinction between norms which do and norms which don’t presuppose some conception or culture of objectivity. Because the application of mundane norms does not depend on a conception of objectivity (a way of distinguishing the objective from the subjective), an ethnographer who has the tools to explore mundane normativity can examine an activity system without either making the mistake of the anthropologist from Mars, or (a priori or whiggishly) presupposing the ontological assignments for the system. It is possible to avoid Dr. P’s howler whilst retaining the insight of the first part of ANT: that metaphysics varies from activity system to activity system, and within activity systems, and that the ethnographer’s job is to empirically discover the metaphysics of the system that is being studied.
So now we can state the position of the constructive realist. The constructive realist agrees that an ethnographer must presuppose normativity in doing her ethnography; that an ethnographer must enter the field with normative concepts in her toolkit. But the constructive realist differs from the pre-emptive realist in requiring that these normative concepts be concepts of mundane normativity, and not be concepts of elite normativity. For the constructive realist, the ethnographer tries to explain how the distinctions of elite normativity are established, maintained and coordinated with each other in the ethnographic site: metaphysics is an explanandum AND the ethnography is normative. Our capsule characterisation of pre-emptive realism is modified to read: elite normativity is prior to activity, whereas the capsule for constructive realism is: activity is prior to elite normativity, or activity is prior to objectivity.
Can Science and Technology Studies account for the relations between the norms of activity and the norms of objectivity in a way that sustains constructive realism? I shall argue that ANT cannot.
We have two tasks: (a) describe the structure of mundane normativity, and (b) use this structure to explain elite normativity. As we will see, ANT(1) tries to do this, in a single step, by means of the stabilization of networks. A theory which uses trails to describe the structure of mundane phenomena gives a multi-step account of the construction of elite properties (see below).
While we’re here, we might also note how other metaphysical positions can be characterised in terms of their explanatory relations between elite norms and mundane norms (though it is only the contrast between pre-emptive and constructive realism that is of importance for this paper). Some have reacted against asymmetric explanations of scientific success and scientific failure by taking the view that the elite norms should be eliminated from STS explanations. For example, it is sometimes said that truth-talk, reality-talk, rationality-talk, genuis-talk and so on are only power-plays by the subjects in the discourse, and STS theorists need only understand the socio-political explanation for whoever it is who gets to establish their truth (etc.) talk. In this view of “Elite Eliminativism”, elite norm-talk has no place in responsible STS explanations. Eliminative materialism is a form of elite eliminativism, as are some forms of social constructivism[9]. Given the distinction between elite and mundane normativity, some eliminativist positions may be characterised as holding that explanations of science should only appeal to mundane normativity. Other forms of eliminativism reject the explanatory role of any kind of normative concept. And then there are accounts which are reductionist rather than eliminativist: they hold that elite norms reduce to mundane norms. That is, there really are elite norms, but they are nothing more than the effects of social forces, or of the enrollment of human and non-human allies[10].
By attending to the explanatory relations between the different kinds of normativity, we can determine the metaphysics of a theory of epistemic activity: pre-emptive realism, constructive realism, eliminativism or reductionism. So let us now introduce trails and networks as normative structures, and go on to examine whether there are, and what are, the metaphysical assumptions built into these structures.
§3 Tales of Two Travellers
Tales of travelling, or better, of getting-about, ought to be central in a discussion of normativity. The reason is this: in describing getting-about one is describing the most mundane form of guidance. And I will be suggesting that the central concept in an account of normativity is guidance. A ‘norm’ is anything which guides behavior or which guides practice[11]. In common with ANT, I will use spatial concepts in a very general way, to include social / technical / theoretical spaces in their extension. Then we can say that the basic form of a norm is “Go this way!” or “Go there!” or, perhaps, “Don’t go over there, go here!” Often we think of norms as propositional goals, such as moral codes or social rules or rules of method, which we strive for, or aim at fulfilling. But norms can be tacit rather than explicit, and they need not have propositional form. Some norms are instituted by outside agency: by God, by the monarch, by the institutions of the law and the police, by the philosophy of science, or by the sociology of science[12]. Some norms are like this, and they too guide activity, but norms can also derive from within an activity system, be maintained without external imposition, and not require the support of specialized institutions. Some norms govern activity as water which is poured into a Pyrex beaker conforms to the shape of the beaker. The shaping, in such cases, is all one-way. But norms need not be pre-emptive in this way; some norms adapt to the resistances and affordances of their activity systems. In other words, adaptation can be two-way: the activity system to the norm and the norm to the activity system.
Most guidance is local, situated in the environment in which practice takes place. Guidance may not require any intentions to act in a certain way, to conform to certain rules, or to maximise some expected utility. Think, for example, of entering an unfamiliar room, and finding one’s way from one side of the room to another. One is guided here by the positioned solidity of the furniture, as one’s body — or one’s informational systems — gently bump up against the materialities of the environment. These ‘bumpings’ are forms of guidance, and thus they are norms, not with the status of the norm of truth, or the norm of efficiency, but fully normative nonetheless. Such mundane norms are situated in the environment, they are changed by the environment and by the flow of activity through it, and they change the environment and its flow of activity. Therefore the meaning of these norms is not independent of their material embodiment.
The paradigm of mundane norms are trails, because trails simultaneously guide those who follow them, and in the very act of guidance are themselves shaped. Thus trails are not like God’s law or the rules of a monarch which are external to practice and imposed on practice, because the changing pattern of trails reflects the changes in practice, as well as, and simultaneously, altering practice, causing it to conform to its guidance. Trails are contingent, historical, embodied and fully local entities, but they establish normative boundaries: this is right, this is wrong; this lies on the path, this lies off the path; this is where you are and this is where you are going. Thus trail-norms are not like the rules of rational method, the rules of a Bayesian maximiser, or rules which are constitutive of an activity, such as the rules of chess. Rules such as these are given as necessary, whereas trails are symbols of contingency. Trails can be global, extending to the limits of some space, yet they are built and maintained locally. Trails vary, are responsive to haphazard local configurations, provide choice points and multiple routes, yet make the difference between arriving and being lost. Trails are fully material entities, constituted out of patterns of disturbance within some material medium. Yet trails are also artefacts. (Perhaps trails are the first artefacts). Thus, trails are simultaneously human and non-human, or animal and non-animal. People and animals gather on trails, follow trails because of the droppings of other animals, because of the increased likelihood of finding other animals (mates or prey or colleagues), or in order to get to go where they are going only because others have been there. Thus trails are simultaneously natural and social and historical. Trails are purely physical, yet they are also symbolic, having a significance that goes beyond the natural laws that govern their embodiment. They represent directions of movement, places to go, places to be, and places where one has come from, yet they are not part of any symbol system. Trails are technologies of reproducibility, for bringing one home again, or for getting back to the water hole that has been out of sight for so many months. This re-producibility or re-presentation is gained in the face of changing and nonconforming environments: because trails flexibly skirt obstacles, and provide work-arounds for the unpredicted, they can provide a robust reliability in returning to the same place again and again.
There are surely some similarities here between trails and networks, especially in the idea that they are simultaneously local and global. But it is no accident that the models for networks tend to be taken from the massive social, technical and human phenomena of modernity, like Thomas Hughes’ “Networks of Power”, whereas the models for trails are simple animals in their environmental niches, and include humans in a way which emphasizes human continuity with the animal. Trails and networks provide for very different forms of guidance, and therefore serve as very different models of normativity. It is compelling to say that trails and networks have different and compatible virtues as topologies for activity. But as we will see, when networks alone are given the explanatorily basic role, trails become inadequate networks: the actants that form the trails of activity are not yet properly aligned, and as an assembly, they are poorly unified. Caught up in the magnificence of networks, we can no longer see the virtues of trails.
Trails, according to this network-centered view, are unstabilized networks, and therefore lack the essential and internal virtue of networks; trails become wannabe networks. Contrasting the animals shaping and being shaped by simple patterns in the earth, with the global wonders of contemporary technology and science, we suppose that the only sense in which trails provide the foundations of normativity is the sense in which trails are a primitive form of normativity. There is a temptation to think of trails like Shirley Strum’s baboons[13]—as primitives—handicapped by their lack of stabilization, and thus forced to laboriously mark out everything in the ever-changing patterns in the earth, or in the ever-changing distance relations amongst the members of the troop. But we ought not to submit to such temptations; trails provide the foundations of normativity, not because of their primitiveness, but because of their adaptability, their flexibility, their intertwined multiplicity, and their negotiability[14]:
• Trails allow for many different kinds of interest (whereas a network serves the interest of its ‘macroactor’)
• Trails do not depend on the alignment of others’ interests although they can be built so as to locally support alignmnent (whereas the functionality of networks requires alignment)
• Trails provide for the cooperative negotiation of interests (whereas unity is imposed in networks, and assemblies of actants are “forced to act as one”). Networks can entrap, whereas trails can allow actants to innovate without having to reject what has already been built.
• Trails have no center, and thus are fully distributed (whereas networks depend on ‘centres of calculation’)
• Trails are robust enough to support many different ‘forms of life’, both in its literal and in its Wittgensteinian sense (whereas networks are brittle in the face of diversity).
Impressed by these virtues of trails, and noting that networks—however good they are as networks—may lack these virtues, we recognize that the different topologies of activity play distinct normative roles, and have contrasting advantages and disadvantages. An ethnographer, thus impressed, should treat networks and trails as independent variables, logged along distinct dimensions, in her description of epistemic activity. If the qualities of trails (“trailishness”) are distinct from—and not merely the inverse of—the qualities of networks (“networkishness”), then an account of the elite normativity of an activity system should enquire into the changing contexts in which sometimes truth requires increased trailishness but decreased networkishness, sometimes increased networkishness but decreased trailishness, sometimes decreased both and sometimes increased both. Activity is described as a trajectory through a space whose dimensions are the dimensions of mundane normativity, including orthogonal dimensions for trailishness and for networkishness. And we can enquire, empirically, into the system’s elite normativity by enquiring into the path of this trajectory (more on this below).
Consider, by contrast to these remarks, how the metaphors of trails and of networks play out in that center of calculation on Boulevard St Michel[15]. In a story of the construction of spaces and times, trails are associated with hacking one’s way through a third world jungle and with the chaos and disruption of anarchic political protest. Networks, on the other hand, are associated with the clean and efficient TGV. We find ourselves first, in the story of the trail, gendered female, and in the “deep jungle”:
The first voyager sets about in a deep jungle and cuts her way with a hatchet along a trail which is barely visible. Each minute, she opens a few centimetres of a pathway, but she ages more than one minute. She sweats. Her body bears the traces of her efforts, each meter can be read in the bloody scars made by thorns and ferns. The path is cut as she gets along, but she is lacerated as well. A suffering body strives among other suffering bodies, lianas, grass and woods. She will no doubt remember all her life each minute of this extenuating trip across the jungle. The reason she will remember it is because each centimeter has been won over through a complicated “negotiation” with other entitites, branches, snakes, rods, that went in other directions and had other ends and goals.
Such a bloody and chaotic tale can be balanced only by a stirring tale of French national pride, of obedience (and a gender transformation). Here is the story of the network, in which we are invited to find ourselves in the situation of the author:
See by comparison how comfortable is the other traveller, her twin brother, who came to this conference, for instance, like me, by TGV! He sat quietly in his first-class air-conditioned carriage and read his newspaper, paying no attention to the number of places crossed by the speedy train which all looked to him like landscapes projected on a movie screen. He did not age more than the three hours of the trip. His body does not bear any trace of the voyage except a few wrinkles on his trousers... The trip for him was like nothing. All the atoms of steel, all the electrons, all the gates, all the switches, all of the efforts of the train companies, SNCF and CFS, ... obeying to the millisecond the world famous Swiss exactitude and the almost as famous French TGV quality of service. No negotiation along the way, no event, hence no memory of anything to mention.
Necessarily there are silent others who make possible all this “TGV quality of service”:
The train passed at 3O0 kilometers an hour without stopping in the very place, Culoz, where all the trains for the Alps and Switzerland used to stop a few years ago... the natives of this little town who, before, had the dignity of being able to stop the train, to board it or alight from it, were now cut into two halves and could not cross or stop the train anymore... This little station had counted, it no longer counts. It interrupted the trip, it no longer interrupts it. It was a station, it is no longer a station. The rails, well aligned, run into only one direction, from Paris to Geneva. So the difference between our two voyagers comes from the number of others one has to take into account, and from the nature of those others. Are they well aligned intermediaries, making no fuss and no history and lending themselves to a smooth passage...?
We can see, in these contrasting stories of trails and networks, a crucial characteristic of ANT. This is the principle of one-dimensionality: what is good qua trail is bad qua network, and being good as a network entails no longer being a trail. That is, in these Latourian stories of getting about, the properties of trails and of networks are inversely related to each other: the more trail-like a form of guidance, the less network-like it is; and the more network-like a form of guidance, the less trail-like it is. ANT(1)’s topologies of activity are arranged along a single dimension, with the qualities of networks best displayed towards the right end, and the qualities of trails being characteristic towards the left end. For ANT, the only ‘good’ for trails is to transform themselves into networks, and thus lose their trailishness. Trails do not have their own positive virtues, as they are manifestations of phenomena which are not yet networks. Or else they are ‘lapsed networks’ which have lapsed back into trails:
The inhabitants of the city who are cut into two by the line may decide to protest and to demonstrate and sit on the tracks or even put logs on the rails and set them on fire (not in Switzerland of course, that would be unthinkable, but let’s say on the French section!). ... Buses will have to take them away from the station and they will lose hours because of the angry demonstrators ... imagine a revolt along all the points of the trip, along each station of the railways and then also along each of the roads taking the buses to turn around the striker blockades. What would happen? Well, we would be back in the jungle we started with!
A second characteristic principle of ANT is that the elite-norms of science (most prominently, truth) are equivalent to the virtues of networks. I will refer to this equivalence relation as “The Truth Relation”. Since networks characterize the structure of epistemic activity, the Truth Relation has the effect of establishing a direct correlation—a one-step connection—between activity and elite normativity. That elite-norms º the virtues of networks (or, elite-normativity º networkishness) is arrived at by two stages. First is the claim that, given the kind of entity that networks are, good networks are stabilized networks. ‘Good’ here is a relative term: what is good for ants may not be good for wildebeest. The TGV rail lines are better as networks if they are stabilized, allowing no interference from the (sometimes) pesky electrons or inhabitants of Culoz. The implication goes also in the other direction: if the rail lines are good networks then they are well stabilized. Therefore, virtue for networks and increased stabilization are equivalent. The second stage in arriving at The Truth Relation is the claim that stabilization is equivalent to truth, or to whichever other elite norm is appropriate, such as efficiency or brilliance or rationality. This identity has been a constant, from Latour and Woolgar (1979) [eg, "... The activity of creating black-boxes[16], of rendering items of knowledge distinct from the circumstances of their creation[17], ... is precisely what occupies scientists the majority of the time", page 259, endnote] through to Aramis: "Truth [is] the result and not the cause of the stabilization of scientific controversies"[18]. (To be fair, identifying stabilization with truth, or more generally with elite normativity, is not confined to ANT. The idea that good science is post-closure science and and the idea that closure is achieved by stabilization / black-boxing are widespread in science studies and SSK.)
Let us add together these two characteristic principles, that topologies for epistemic activity are arranged along a single dimension, and that the elite normativity of science is equivalent to stabilization and thus to network-virtue. The sum of these principles provides a picture of a single dimension along which the epistemic activity of true Science lies towards the right (networkish) end, and science which is not-yet-made, which has not yet reached closure, or which has been revealed as fraudulent lies towards the left (trailish) end. Position along the dimension is marked as degree of stabilization. Networkishness is Good (for science, for technology, for business, ...) and Trailishness is Bad (for science, for technology and for business). The problem for cold fusion is that its socio-natural allies are disobedient, and, unlike “the atoms of steel, all the electrons, ... all the efforts of the train companies”, refuse to align themselves. Success for cold fusion is only brief, after which its aligned epistemic structure is transformed back into that of hazardous and chaotic trails, rather like the ‘not-any-more-railway’ after the interventions of the inhabitants of Culoz and the other towns along the line.
In summary, ANT (or, at any rate, ‘let’s say the French section’ of ANT), is distinguished by the following:
(1) Network topologies are the basic explanatory structure for science and technology
(2) Good networks are stabilized networks, whose parts are ‘black-boxed’
(3) Bad networks are trails
(4) Goodness for Science is truth, and other elite norms (this much is necessary for any form of realism),
(5) Therefore, (from (1), (2) & (4)), Truth in science is made by making stabilized networks, and black-boxing their components. That is:
(i) Goodness (virtue) in Science = Goodness (virtue) for Networks; (ii) Truth = Stabilization; and:
(iii) Trailishness is Bad for Science (given (3))
(6) The norms for networks and trails are arranged along a single dimension.
(7) Therefore, normativity in science is one-dimensional
These claims which are characteristic of, if sometimes implicit in, ANT(1), especially that truth = stabilization, that trailishness is bad for science, and that normativity in science is one-dimensional, are, I believe, both false and at odds with ANT’s constructivist, actantial metaphysics. As a way of pursuing this, I want to respond to an objection that I have misrepresented ANT(CSI)’s position, which, after all, in various places has recommended two-dimensional accounts of scientific and technological activity. By considering this objection we will gain a better understanding of ANT(1)’s single-step explanatory connection between activity and elite normativity, how the single-step deprives the ethnographer of her proper toolkit, and how it results in an unwanted commitment to pre-emptive realism.
§4 Ethnographic Yardsticks: Response to an objection
§4.1 The 1-d ® 2-d Transformation, as told by Callon and Latour
In “One more turn after the social turn” Latour (1992)[19] considers the three guarantees of the modern constitution: the non-human origin of knowledge, the human origin of knowledge, and the “complete separation between the two”. Abandoning the first guarantee leads to social constructivism, abandoning the second leads to [pre-emptive] realism[20], and that therefore “the only one that might be discarded” is the third. The result of discarding the third is that instead of the dual transcendences—Nature and Society—of the modernist (Kantian) picture, we have only one transcendence: nature-society. Moreover, neither Nature nor Society can now be used to explain science and technology since both are the results of the practices of science and technology (that is Nature and Society, Mind and World are explananda)[21]. As Callon and Latour (1992) put it[22]: “Our general symmetry principle is to ... obtain nature and society as twin results of another activity, more interesting for us, and that we call .. network building... The phenomenon we wish to describe cannot be framed from the two extremes -- nature out there and society up there -- since on the contrary “natures” and “societies” are secreted as by-products of this circulation of quasi-objects”[23]. Here, then, are two clear statements of what I have been calling the first part of ANT, the part with which I heartily agree. We have only one metaphysical transcendence, and it is constructed in-site.
If metaphysics is constructed in-site, then any good ethnographer needs to be equipped with the resources necessary to follow this construction. ANT must therefore reject the one-dimensional metaphysics which is characteristic, in Latour’s account, of Kantian modernity (see figure 2). Why must ANT reject this metaphysics? Because Society / Nature, which used to be explanantia (as in figure 2) are now, in the new science studies, explananda. And these diagrams (figures 2 and 3) are “yardsticks”, that is instruments available for the ethnographer to gather data on scientific and technological practice, which can be used in explaining the changing course of that practice. So Callon and Latour add a second dimension, that of stabilization (figure 3).
The one-dimensional yardstick [Figure 2] allowed us to position any entitiy along the object-subject line. I showed that although this was useful it did not do justice to most of the discoveries of science studies: objects and subjects are belated consequences of an experimental and historical activity that does not clearly differentiate if an entity is 'out there' in nature or 'up there' in society. This means that any entity should be logged [not only according to its position along the object - subject line but also] according to its degree of stabilization.
Figure [3] is an attempt to define any entity by two sets of coordinates instead of one. One line is the distance to P’, the locus of phenomenon in Kant’s scenario, and goes either to the subject/collective pole or to the object pole. The other is the degree of stabilization going from O to P’, from unstability to stability. It is clear from the diagram that the one dimensional yardstick I critiqued above corresponds to only one value of the stabilization gradient. [Latour, 1992]
[Figure 2: (adapted from Latour (1992))the ‘Kantian’ “one-dimensional yardstick”, in which “empirical scientific knowledge appeared in the middle, but this middle, the phenomenon, was understood only as the meeting-point of the two purified sets of resources coming from the subject-pole and from the object-pole”]
Pre-emptive realism and social constructivism are one dimensional because they are committed to the idea that, as explanatory resources, Nature and Society must be kept as far apart as possible. Latour, by contrast, wants to play with “variable ontologies”, allowing the ‘distance’ between nature and society to change, according to the value of the “second dimension” of stabilization.
[Figure 3: (adapted
from Latour (1992) “adding a second dimension”)]
[for similar diagrams, see Callon and Latour (1992), p. 349 and Latour (1993), p.86]
It is important to see that ANT(CSI)’s demand for a second dimension is presented as arising out of the history of STS (as they tell it). The Bloor[24] of this history effects the first “social turn” in STS by rejecting explanations of truth by Nature and error by Society, and instead requiring that both truth and error should be explained by Society. It is then pointed out that although this requirement is symmetric with respect to truth and error, it is asymmetric with respect to Nature and Society. Hence we need to “turn through 90 degrees” by adding a second dimension which is perpendicular to the Nature-Society dimension. From the perspective provided by this second dimension, Nature and Society are seen symmetrically. Hence figure 3. And hence the claim “I showed that although [the first ‘Kantian’ dimension] was useful it did not do justice to most of the discoveries of science studies”.[25]
Unfortunately, it is just at this point of ANT(CSI)’s most important and radical contribution to metaphysics—where finally we can move beyond oscillations between pre-emptive realism and social constructivism—that the account becomes confused. It is confused for three reasons.
§4.2 The First Confusion: You’ve Forgotten that the Nature « Society Dimension is no longer an Explanans
Because of the historical construction of figure 3 type diagrams, the Nature-Society dimension has been retained, even though it no longer has any business in these diagrammatic ethnographic “yardsticks”. We saw above that in the ‘Kantian’ picture, Nature and Society are explanantia, because they are taken to be the basic explanatory resources. It is because they are resources that the “third guarantee of the modern constitution” is required: if the two poles were not to be held apart then they would lose their explanatory power. It is this picture of one-dimensional explanation that is characteristic of SSK, of pre-emptive realism, of social constructivism and of ‘happy medium’ stances. In order to move beyond this picture, the actantial metaphysics holds instead that Nature and Society are explananda. They are no longer resources, and therefore do not belong in these “yardstick” diagrams. The diagrams are diagrams of resources which are available to the ethnographer for the explanation of knowledge-based activities. But the whole point of the “second turn” is that nature and society are “by-products” or “results” of the circulation of quasi-objects / network-building. Therefore, they are not available to do basic explanatory work. So, GET THAT NATURE-SOCIETY DIMENSION OUT OF FIGURE 3! The only problem is that once it is taken out, as it should be, we are left with: one dimension.
It is not hard to see why this confusion has occurred. The very language of “one more turn (at 90 degrees) after the social turn” suggests something looking like figure 3. Moreover, it clearly shows how ANT treats nature and society symmetrically: that’s the beauty of gaining a perspective which is perpendicular to Nature « Society. Furthermore, because the two-dimensional diagrams are always introduced via a history of science studies, what is distinctive of the CSI account is situated with respect to SSK etc., and therefore both players have to be in the picture. (Otherwise the task of situating ANT(CSI) would fail). So as a didactic device it makes sense that the diagram should retain the dimension which is characteristic of explanation in SSK, in pre-emptive realism, and in social constructivism. How easy it is, then, to confuse this didactic picture with the picture of how explanation in the new science studies works. In the presentation in Callon and Latour (1992), the horizontal dimension is labelled “SSK’s yardstick” and the vertical dimension is labelled the “‘Paris’ yardstick”. It’s as if, why not have both? Indeed, the rubric for the diagram suggests this—note the “as well” and the “not only ... but also”—: “The one dimensional yarstick of figure [2] is allowed to position any entity along the object-subject line (their longitude). The two-dimensional yardstick allows us to position objects and subjects according to their degree of stabilization as well (their latitude), and thus to offer for each entity two coordinates. Of each entity we would not only ask if it is natural or social (projected [towards one end or the other of] the SSK yardstick) but also if it is unstable or stable [towards one end or the other of] the "Paris" yardstick).”
The last sentence of this quotation points us towards the second confusion.
§4.3 The Second Confusion: You’ve Forgotten that Nature«Society is now supposed to be a SINGLE transcendence
Recall §4.1: once the third guarantee of the modern constitution is rejected, there is no longer any need or basis for treating nature and society, object and subject, world and mind as having distinct ontologies. Latour (1992): “Instead of the two opposite transcendences of Nature and Society ... we have only one transcendence left. We live in a Society we did not make, individually or collectively, and in a Nature which is not of our fabrication. But Nature “out there” and Society “up there” are no longer ontologically different. We do not make Society, more than we do Nature, and their opposition is no longer necessary. ... Instead of providing the explanatory resources in order to account for empirical phenomena, this common transcendence becomes what is to be explained...”[26]. So, if we are to retain—contrary to §4.2—a Nature«Society dimension in our ‘yardstick’ diagram, then it must not be represented in terms of two poles, each with its own ontology (for that is to present Nature«Society as a dual transcendence). Figure 3 is intended to display the form of explanation in the new science studies, yet the bipolar representation of the horizontal dimension in this figure encourages us to ask of an entity, is it natural or social, or more natural than social, etc. Indeed in the last sentence of the quotation in §4.2 we find Latour submitting to precisely this diagrammatic encouragement, “Of each entity we would not only ask if it is natural or social ...”[27]. Yet once nature-society is seen as a single transcendence, this question makes no sense for the practitioner of the new science studies: “the very notion of culture is an artifact created by bracketing Nature off. Cultures .. do not exist, any more than Nature does. There are only natures-cultures...” [Latour (1993), pp. 103-104].
ANT(1)’s guiding norm for epistemic activity is: Increase Stabilization! (see §3). But at maximum stabilization, figure 3 returns us to what, in the Callon and Latour (1992) version of figure 3, is labelled “SSK’s yardstick”; ie. the figure 2 metaphysics of twin transcendences. How disappointing if success in ANT(1)’s own terms has the effect of reinstating the dualistic metaphysics, which was so effectively criticized in the first (‘actant’) part of ANT. Again we see the two parts of ANT(1) in tension with each other: given figure 3, network normativity is incompatible with actantial metaphysics. Somehow or other we have to understand the network dimension in such a way that normative success does not entail the ontological myths of modernity.
If, contrary to §4.2, metaphysical explananda are to appear in our diagram, then they had better manifest the metaphysics of a single transcendence. There are two ways to do this. First, position along the horizontal dimension could represent the degree to which the single transcendence of nature-society is attained in the ontology of the site, at a time and in a particular context. This would be a dimension of subject - object - ivity (what I call s/o-jectivity—see Cussins, A. 1992), such that position along the dimension measures the degree of separation between subjectivity and objectivity, between society and nature. Western science, under conditions of closure, would be logged as a high value of s/o-jectivity, whereas the epistemic practices of the Amazonian Achuar would be logged as a low value of s/o-jectivity (see Descola (1996))[28]. Or, secondly, if we retain the depiction of the horizontal dimension as having a ‘nature’ end and a ‘society’ end, then the projection of any more or less unstabilized entity onto this dimension will be marked not as a point, but as a scalar, an extent., whose center will always be in the middle of the axis (at P’ in figure 3). Entities with low stabilization will project onto a line that passes through P’, but that extends only a small way on either side of P’. Entities with high stabilization will project onto a line that has large extent on either side of P’. In a virtuous epistemic practice, as virtue is understood in ANT(1), the entities referred to early on in the practice (pre-closure) will project onto a small extent on the x-axis, whereas the entities referred to later on (post-closure) will project onto a large extent on the x-axis. The transition from the line E-F to the line C-D in figure 3 may be thought of in this way (although the corresponding diagram in Latour (1993) loses this insight, returning us to points projecting onto points). Thinking of the horizontal dimension in this way brings us to our third confusion.
§4.4 The Third Confusion: You’ve Used Two Dimensions to Draw a Picture that is Only Doing One-Dimensional Work. To Do Two-Dimensional Work, the Dimensions Need to be Orthogonal. But the Dimensions in Figure 3 are Not Orthogonal, because Position Along One Dimension can be Predicted from Position Along the Other Dimension!
Here is the bottom line: Figure 3 is really only a one-dimensional graph. (Like the pre-emptive realists and the social constructivists, we are again reduced to “drawing cartographic maps on the basis of longitude alone”! (Latour (1993), p. 85)). That an illustration is drawn two-dimensionally is not sufficient to show that what it illustrates is two-dimensional. In order to function as a genuine two-dimensional graph, the two axes must be orthogonal to each other, but we can see that the axes in figure 3 are not orthogonal because position along one axis determines position along the other. Thus, low stabilization entails a scalar whose x-axis projection has a small extent around P’, and high stabilization always entails a scalar whose x-axis projection has large extent around P’. In figure 3, if an entity is logged at position P on the y-axis then it would follow without the need for any further empirical measurement, that its x-axis extent must be given by the coordinates {C’, D’}. Or, in other words, the representation of an entity in figure 3 as a line whose left-end is at C and whose right-end is at D follows logically from the measurement of its stabilization as being equal to P. Similarly, if an entity, with lower stabilization, is measured at position Q on the y-axis, it follows a priori that its x-axis coordinates are {E’, F’} and thus that it is represented in the graph by the line that runs between E and F. Therefore, position along the x-axis is predictable from the position along the y-axis. Hence the axes are not orthogonal, and the figure does not depict a two-dimensional graph.
This is all very bad for the ethnographer. ANT(1) equips the ethnographer with a stabilization meter which she is to operate as she Follows the Networks. But if figure 3 turns out to be really only one-dimensional, then each reading of the stabilization meter is correlated one-to-one with a given metaphysics. Therefore ANT(1)’s ethnographer must violate ANT(1)’s own constraint that the ethnographer must not make a priori or whiggish assumptions about a site’s metaphysics. The ethnographer ought not to enter the activity system that is her site, already knowing what constitutes elite normativity in the site, because the formula for elite normativity (elite norms = high stabilization) is built into the instruments in her toolkit. What ought to be in her explanatory toolkit is not a formula telling her what objectivity and reference are, (etc.) but some resources for going and looking at the practice and so finding out what objectivity, reference and elite success are, for, and in the context of, that practice.
The pursuit of constructive realism depended on the ethnographic use of instruments whose normative commitments were exclusively mundane. To say that figure 3 is really one-dimensional is to say that ANT(1)’s instrumentation has elite normative commitments: the instrument itself determines what it is to be true or rational or brilliant in the site. What constitutes the elite norms is, as it were, already built into the meter with which ANT(1) equips its ethnographers. Which is to say that the determination of elite normativity is prior to the ethnographic investigation of the changing course of activity. Which is to say (recall §2) that ANT(1), despite its own protestations and best intentions, is a form of pre-emptive realism. Back to the drawing-board!
Once normativity is theorized as one-dimensional, the problems do not end with the commitment to pre-emptive realism. My aim in this paper is to diagnose why it is that treating networks as the exclusive topology of epistemic activity ends up entailing pre-emptive realism. But I might just mention a few other problems, to be treated elsewhere. Given the principle of one-dimensionality and the Truth Relation, ANT(1) also loses the possibility of accounting for the historicity of objects and objectivity. An object may be represented as a trajectory through a figure 3-type diagram, but, given that these diagrams are really one-dimensional, there is the possibility of virtuous movement and change only so long as maximum stabilization—and therefore, given ANT(1),objectivity —is not attained. Thus historicity is confined to the period prior to attaining objecthood or objectivity, or to the period after losing objecthood or objectivity. Hence objects and objectivity are not themselves historical entities; all they have is a pre-history!
Just like trails, networks have both virtues and vices. By establishing a one-to-one correlation between network-virtue (ie. stabilization) and elite normativity, we cause ourselves to be blind to the vices of networks. As Jordan and Lynch (1993) point out, one result is that “Latour’s general definition may serve to articulate the operative ‘theory’ of the corporate promoters and proprietors”.[29] A related result is the conception of users of science and technology as enrolled, enlisted, entrapped, absorbed, deskilled and dependent. There is much to say here, and much that has been said. If this paper can make a contribution, it is to show that these deleterious consequences are the result of adopting (despite misleading practices of illustration!) a one-dimensional account of normativity, and the assumption of a single-step connection between activity-structures and elite normativity. Elite normativity can then only be understood as an end-state, a golden state, a state which once achieved leaves nowhere else to go (a stasis). Once in such an end-state, the only virtuous act is to maintain the status-quo, and to avoid any risk of destabilization. The result is an ideological, even a fundamentalist conception of virtue. Back to the drawing-board!
§5: What’s Gone Wrong? And What Can We Do About It?
Now, finally, we can see the value in drawing the distinctions between elite and mundane normativity, of distinguishing between different topologies for characterising epistemic activity, and of discerning the normative commitments of the different topologies. For we have thereby put ourselves in a position to understand what has gone so badly wrong. Because networks are micro-structural, and locally built out of semiotic relations amongst inscriptions, we all thought that networks must be mundane normative structures. But the arguments of §3 and §4 have shown us that, despite appearances, ANT(1)’s networks are really elite normative structures, and are thus ill-suited as the basic explanantia in an enterprise which is constructively realist. This is not the fault of network topologies—which have many constructivist virtues—but it is the fault of the two principles that we saw (in §3) governed ANT(1)’s use of network topologies: the principle of one-dimensionality and the truth relation. So, all we have to do to put things right is to replace the ideas that trailishness is always bad for epistemic activity, that networkishness is always good for epistemic activity and that elite normativity is given directly (in one step) as high stabilization. We must give up on the picture that has networks and trails arranged along a single normative dimension, with the trails as mundane normative structures and the networks as elite normative structures. How?
In the early part of §3 we noted that trails had a set of epistemic virtues (and vices) that are quite different from the epistemic virtues—well described in the STS literature—(and vices) of networks. So we should let trail topologies and network topologies form distinct dimensions, each of them having only mundane normativity. Because trail virtues and network virtues are not correlated with each other, we can draw a genuinely two-dimensional explanantia diagram, with one axis marking degree of trailishness and the other axis marking degree of networkishness:
Figure 4: A genuinely multi-dimensional explanantia diagram
Our (ANT(n?)) ethnographer is now equipped with two quite different forms of instrumentation, one form measuring properties of networkishness, and the other form measuring properties of trailishness. (A Figure 4-type diagram can be drawn in different ways depending on which properties of networkishness and trailishness the instruments are attuned to). A third dimension is used to track a variable of change: this might be time, or changing location, change in actants, or changing context. Rather than attempting to draw a third dimension, I will indicate it by means of the arrow which accompanies the trajectory through the two-dimensional networkish / trailish space (see figure 5). Moreover position along none of the dimensions is correlated with elite normativity! (“Network” thus has a rather different meaning in the context of ANT(n) than it does in the context of ANT(1)). Hence, a trajectory through the space does NOT have its elite normativity assessed according to its distance from some end-state virtue (pre-emptive truth, etc.).
Since the instrumentation which tracks the changing variables in the ethnographer’s site operates without any commitment to elite normativity, and therefore without any metaphysical commitment, the ethnographer is free to empirically explore elite normativity as an historical and situated phenomenon. The ethnographer must attempt to empirically establish correlations between different plots of the dynamics of her activity system and assessments of elite normativity. She is therefore free to ask what constitutes elite normativity in this site, at these times and in these contexts. Metaphysics for her, is—as it ought to be—an explanandum. Elite normativity is presented not as an ideal end-state, but as a dynamicway-of-being, which is situated and historical, and therefore whose form (a description of which is given in the activity plot on the multidimensional graph) may change over time in a way which is context-sensitive. For an ethnographer who Follows the Trails, truth, reference, rationality, creativity, brilliance and objectivity are neither pre-emptive of activity, nor eliminated, nor reduced, but constructed as real, situated, and historical ways of being.
Figure 5: Tracking Elite Normativity through Multi-dimensional Plots of Dynamic Activity Systems[30]