Grounding agency in a world without certain grounds: perspectives for a relational understanding of agency - An exploration of the relation between enactivism and Actor-Network Theory

(1)

GRADUATE SCHOOL OF HUMANITIES

Department of Philosophy

UNIVERSITY OF AMSTERDAM

Grounding agency in a world without certain grounds:

perspectives for a relational understanding of agency

An exploration of the relation between enactivism and Actor-Network Theory

Joris Melman

15-08-2016

MASTER THESIS

Supervisor: Dr. J. Bos

(2)

(3)

2

1. Introduction

The last years have seen the emergence of different theories that attempt to destabilize the notion of the sovereign human subject. Two of the most prominent theories in this respect are Actor-Network Theory (ANT) on the one hand, and the so called 4E approaches in cognitive science on the other hand. Interestingly, although these theories have both been inspired by, and led to a renewed appreciation of philosophical areas such as process philosophy and

phenomenology, both emerged initially outside of the traditional philosophical arena. However, despite this shared origin outside of philosophy, the background of both theories couldn’t have been much more different for the rest. Whereas ANT firstly began in the 1980’s as an approach to provide a social explanation of scientific practice, the 4E approaches (which is a broader banner for approaches studying the mind and cognition as extended, embedded, embodied and/or enacted) emerged in the 1990’s as a critique of the traditional representationalist view in cognitive science that sees cognition as taking place solely in the brain.

Despite the fact that such an entirely different background entails a different method, vocabulary, and theoretical framework, both theories seem to arrive at rather similar

philosophical conclusions and positions. Most importantly, both in their own way aim to describe how humans are fundamentally situated and embedded in their own particular environment or network, which in terms of their philosophical positions is combined with a rejection of the subject-object dichotomy, a monist and relationist outlook, and a

post-humanism providing a new perspective on the relation between humans and their environment. In order to account for this situatedness of humans and their entangledness with the world then, both are required to rethink the concept of agency. And although they do so in a different way, both theories in this way pose questions for the common sense understanding of agency as something limited to humans, to consciousness in general, and even to the physical boundaries of an organism. Against the traditional anthropocentric conception of agency, they put forward a more ‘ecological’ understanding in which the very possibility of action only arises in alliances between living beings and their surroundings. Indeed, it can be said for both that it is this reconceptualization of the role of humans that to a considerable extent explains why they have drawn so much attention. (Because indeed, both theories seem to establish firm roots in their respective fields and beyond.)

Accepting the idea for now that there are indeed some remarkable similarities between both approaches, the question is what this signifies. Is it just that two schools of thought happen to have a resembling worldview, or could it be that we are witnessing the first signs of the emergence of a new zeitgeist? It might be too early to speak about the latter, but the two approaches do fit into a broader trend of approaches attempting to emphasize the

entangledness of humans and their environment. In these approaches, the aim is usually to think from a less anthropocentric viewpoint, and to think more from the perspective of the (material) things surrounding humans. Indeed, some speak of a ‘material turn’, a term used to describe how scholars in a wide range of fields are reconsidering how humans and things constitute each other. Having a strong focus on materiality, this development is especially linked to fields which have always had a strong focus on the role of objects and other

(6)

5

‘nonhumans’, such as anthropology, archaeology and geography. But at the same time, scholars in fields such as political theory and sociology are affected just as well. In the work of Bennett (2009) and Coole and Frost (2010) for example, a renewed understanding of materiality is combined with political theory. And together, this has thus led to new ways of describing and understanding of for example the role of technologies in society, the source of environmental problems, and developments in the geopolitical sphere.

Although both approaches are in a way related to this emerging field and contribute to its development, ANT is more directly connected to it. Providing a perspective in which humans, material objects, and non-material entities can from a semiotic point of view all be described as equally important elements of social assemblages, it is in a better place to address the

situatedness of humans as a societal issue compared to the 4E approaches. For the 4E

approaches on the other hand, such ‘societal’ ambitions are not as central. And in addition, the 4E community would probably not identify with a strong emphasis on materiality. However, what makes the 4E approaches fit to this development is that it sketches a perspective on humans are fundamentally situated and embedded in their (material) environment from a more naturalistic perspective. Having the interactions between a cognitive subject and its

surroundings at the centre of its attention, it offers the tools to sketch in a somewhat clearer and more systematic way the relation between subject and object. And indeed, the development of the 4E approaches proves that it has a lot to say here. For instance, the work of Malafouris on material engagement makes a strong case for how the 4E approaches could contribute to a new understanding of the relation between human subjects and material culture at a fundamental level. And interestingly, Malafouris uses arguments derived from precisely the 4E approaches and ANT next to each other, indicating that they are not only similar in a general, sketchy way, but can also be combined in making a coherent argument.

Comparing the two approaches with each other, we thus see that a similar perspective on fundamental issues such as agency and the subject-object relation is combined with the use of entirely different tools for developing such a perspective. This therefore makes that both approaches are very interesting to relate to each other. More in particular, this thesis holds that doing so provides new insights for the field attempting to develop a non-anthropocentric approach to agency. On the one hand, it could reinforce certain ideas and positions when both approaches arrive on them on the basis of different arguments. At the same time, distinguishing areas where both approaches conflict might put into question certain elements of their

respective arguments. And in this way, problematic parts of a non-anthropocentric, relationist perspective can be identified that still need to be dealt with. Which in turn shows new

directions for research as well as theory development.

The aim of this thesis then is to relate the two approaches to each other in a more or less systematic way. This means to – rather than only showing that there are certain similarities (which nevertheless certainly is a first goal) – make a case for how the philosophical positions and theoretical arguments of both match or overlap, how they differ, and – importantly – how their respective positions can be used to criticize the position of the other. Because indeed, it will be argued that the two approaches are in a good place to reveal shortcomings of each other, precisely because their similarity is combined with an entirely different background.

(7)

6

More in particular, this thesis will focus on the concept of agency. As indicated above, agency plays a central role in any attempt to reconceptualise the place of humankind in its environment, and so also in these two approaches. For both the 4E approaches and ANT, the concept has the function of showing how humans are not the sovereign, rational actors mastering their environment modern science often claims them to be, but are just as much shaped by their environment as vice versa. Thus, being at the centre of both approaches – at the same time being a relatively ‘concrete’ topic – agency appears to be a productive entry point for understanding how both approaches relate to each other. And what makes it even more

interesting is that despite the fact that both have an understanding that overlaps in a way, these understandings are at the same time still quite different from each other. As will be discussed later, the 4E approaches on the one hand still tend to take natural categories (such as

organisms) and their boundaries as a starting point, and therefore also have a conception of agency that is more bound to individual actors. ANT on the other hand usually sees such actors rather as a consequence of the relations they are embedded in, instead of existing prior to such relations. Hence, it does not accept such preconceived actors as the starting point of an analysis, but more fundamentally undermines the idea that agency is coupled in a certain way to such individuals.

These differences in the conceptions of agency are not isolated differences. Indeed, what makes them interesting is precisely that they are indicative for more fundamental differences, or so this thesis holds. It is argued that rather than being merely coincidental, the differences in the conceptualization of agency are instead linked to differences in the ontologies underlying both approaches. Although neither approach is very explicit or consistent in the development of its ontology, ANT appears to take its relationism further than the 4E approaches do. Whereas ANT conceives of entities as being entirely determined by their relations, the 4E approaches (or at least the branch of enactivism, about which more in the next paragraph) still seem to leave some notion of essences that exist independent of such relations intact. And this difference being reflected in their both conceptions of agency, it is with the initial focus on agency that we are enabled to grasp the core of what differentiates enactivism and ANT. And as a consequence, this focus opens up the possibility to make an argument on how the approaches mutually expose problems of each other, as well as how they might complement each other.

In order to be able to do so in a clear and precise way, this thesis considers it important to use a consistent theoretical framework of both. It therefore chooses to use one focal point within each of both theories. Because both strands of literature consist of a rather wide ranging set of arguments instead of being neatly unified, speaking about ANT and the 4E approaches in a broad way might otherwise lead to fuzzy accounts. In the case of the 4E approaches, this thesis will mainly focus on the arguments of the enactivist account, more in particular those of what has been referred to as the autopoietic strand of enactivism. Although differences between other accounts are not that insurmountable (several scholars hold that the different approaches are compatible when understood in their proper philosophical background (e.g. Di Paolo, 2009)), the debate about their precise mutual coherence is still open (e.g. Kiverstein and Clark, 2009), and some argue that a diversity in approaches is actually to be preferred over having one monolithic field if the 4E approaches are to be productive (Vörös et al, 2016). (Autopoietic)

(8)

7

Enactivism then seems to be the most relevant of the 4E approaches, because it is in the framework of enactivism that a specific conception of agency is most important, which also entails that it is most clearly developed in this approach. In addition, it is this account that has the initial philosophical motivations of the ‘original’ enactivism of Varela et al. are most central. In presenting the arguments of ANT on the other hand, the focal point will be the thinking of Latour. This is because Latour’s arguments can be considered as the most influential in the field of ANT, but also because agency has been a constant issue in his writings, and because in ANT it is his thinking that seems to be most explicitly dealing with ontology and metaphysics in a way that is relevant for this thesis.

In pursuing its aim, this thesis will proceed as follows. First it will present the enactivist approach and its conception of agency, followed by an account of ANT and the way it deals with agency. Then it will make an argument about how the approaches and more in particular their conceptions of agency relate to each other, and highlight three main differences between these conceptions. After having subsequently briefly outlined the way both approaches are used next to each other in material engagement theory, it will then elaborate on how these differences can be traced back to the ontologies underlying both, and argue that these differences can therefore be understood better when going deeper into those ontologies. Now having an idea of the more fundamental relation between enactivism and ANT, it will then relate this back to their both conceptions of agency, and finally discuss how these conceptions ultimately relate to each other, and what this implies for attempts to study agency in a relational, non-anthropocentric way.

2. Enactivism

Since the early 1990’s, a revolutionary spirit has been going through the field of cognitive science. Being a reaction against the then dominant cognitivism, it has led the field to take a “dramatic anti-representational turn” (Ramsey, 2007: xv). Two decades later, it seems that this spirit is more and more consolidating, and that non-cognitivist theories of cognition are establishing firm roots in both cognitive science and the philosophy of mind. Mostly building on the enactivist account as originally developed by Varela et al. in The Embodied Mind (1991), many scholars are conceptualizing the mind and cognition as embodied, extended, embedded, and/or enacted. However, even though these projects are together mostly referred to as the 4E approaches (although this term is not used unanimously, and some distinguish other sub-approaches such as affectivity, leading them to speak of the 4AE or even 6AE sub-approaches), important differences exist between the arguments made under this broad label. These differences lie not only in the precise arguments that are made about the functioning of

cognition and the mind, but also in the philosophical framework underlying these arguments. It is therefore difficult to make sense of how exactly they relate, which indeed still is subject of discussion. As outlined in the introduction, this thesis therefore chooses to focus on the

arguments made in the approach called enactivism, more in particular the so called autopoietic enactivism (a term coined by Hutto (Hutto and Myin, 2013) in order to distinguish it from his own radical enactivism).

(9)

8

The aim of this chapter is to outline how this (autopoietic) enactivist approach

conceptualizes agency. But because this conceptualization cannot be made sense of without understanding it in the context of the more basic enactivist argument, the chapter starts with a section outlining its theoretical and philosophical basis, giving a brief account of the ‘classic’ enactivism of Varela et al. Important to note, this is not just the basis of the approach which is currently referred to as enactivism, but can be seen as the basis of the 4E approaches as a whole. After this outline then, the chapter will continue with an account of enactivism’s

conceptualization of agency. But because this account can be somewhat misleading when read on its own, it is important to again contextualize it, this time by placing it enactivism’s broader framework. A final section serves to do just this, at the same time allowing for a more detailed elaboration on the enactivist argument.

2.1 Enactivist foundations and epistemologies

Having emerged as an explicit reaction against cognitivism, it is important to understand enactivism by first looking at its relation to this cognitivist approach. Traditional cognitivism emerged in 1950’s, when first steps were made in the development of digital computers and information theory. Following these developments, scholars aimed to create a theory of the mind in which cognition and mental phenomena are explained in computational terms. In such a theory, the notion of representation became essential. In order for syntax-driven computation to be possible, there need to be certain symbols that can be computed. In the words of Fodor: “there is no computation without representation” (1981: 122). And in cognitivist theory, these symbols are derived from the brain translating sensory input to certain symbolic information that re-presents the world. Hence, the information that the brain uses for computation is a representation of a pre-given world. In the words of Varela et al., cognitivism thus understands cognition as “information processing as symbolic computation–rule-based manipulation of symbols” (1991: 42). The mind more broadly is understood as a process of perception (input), cognition (information processing) and agency (output). In this way, the cognitivist approach detaches cognition from its material embodiment, and makes it an abstract process that can take place in “any device that can support and manipulate discrete functional elements–the symbols” (ibid.). Taking the digital computer as a model for cognition, cognition is seen as a passive, isolated activity that takes place when the mind receives sensory input from some source.

Against this abstract ‘brain-in-a-vat’ like conception of cognition, Varela et al. attempted to come up with a new understanding of cognition that does more justice to our lived, everyday experience, in this way ‘re-enchanting the concrete’ (Varela, 1992). In this understanding, cognition is not an isolated, passive process, but is something that emerges from embodied processes between organisms and their environment. Rather than a clean process consisting of input, information processing and output, cognition is something closer to an “Escher

Spaghetti” (Clark, 1999: 7), in which brain, body and environment interact with each other in a complex way. Mind and cognition are not always there, passively waiting for some input from outside, but only come into being because they are enacted in concrete, situated interactions. In a sense, they are actively done.

(10)

9

More specifically, Varela et al. laid down a conceptual framework in which cognition is conceived of as embodied action, which is opposed to understanding it as a process of

representational information processing. As embodied, firstly, cognition is understood as essentially bound to the experiences that are made possible by having a body that is endowed with certain sensorimotor capacities. And by conceiving of it as action, secondly, Varela et al. make the point that “sensory and motor processes, perception and action, are fundamentally inseparable in lived cognition” (1991: 173), and that by having evolved together they both constitutively depend on each other. The sensorimotor processes of an organism (or system) that constitute the embodied action that its cognition entails then, are understood as

structurally coupled to their environment. This concept refers to the observation that if two

systems share a history of interaction, there is congruence between their structures. To say that the sensorimotor loops of an organism are structurally coupled to its environment thus means to say that the cognitive capacities of this organism are constitutively bound to its history of dynamic interactions with its environment. Therefore, understanding cognition as embodied action means to understand it as emerging from the interactions between an organism and its environment. Cognition is thus enacted in the sense that it emerges in a dynamic interaction between systems: cognition essentially is “a history of structural coupling that brings forth a world” (p. 206).

In this sense, a cognitive agent is shaped by its environment as well as it shapes it. Or in other words, “knower and known are mutually specified” (Maturana and Varela, 1987: 253). Cognition is not the detection of information from a pre-given outer world by a subject, and neither the construction of a world by a subject. Instead, a cognitive actor and the world are two sides of the same coin. Building on Merleau-Ponty’s observations that the world is inseparable from the subject, and the subject is inseparable from the world, Varela et al. argue that the subject and the world co-emerge from recurrent dynamic interactions. Thus, a monist and relationist view of cognition comes to the fore.

It should be clear that the motivations of Varela et al. were just as philosophical as scientific. They did not just aim to provide a theory that more satisfactorily describes certain cognitive processes, but attempted to develop a whole new understanding of the mind and its functioning, bringing about “nothing short of a paradigm change” and offering “a whole new set of metaphors for doing cognitive science” (Kiverstein, 2012: 741). As a consequence, they based their approach not only on the findings of scientific disciplines such as that cognitive disciplines and system theory, but also to a large extent used the philosophical insights derived from the phenomenology of Heidegger and Merleau-Ponty as well as Buddhist philosophy. Indeed, in making sense of enactivism, it is essential to understand it in the light of the profound philosophical motives that formed the basis of it.

As a most important philosophical goal, The Embodied Mind attempted to overcome the epistemological divide between realism and idealism. As the authors point out, they try to find a “middle path between the Scylla of cognition as the recovery of a pregiven outer world (realism) and the Charybdis of cognition as the projection of a pregiven inner world (idealism)” (p. 172). Essentially, the authors argue, both these extremes have a notion of representation at their core. In both cases, this is a consequence of the ‘Cartesian anxiety’, meaning a craving for finding an

(11)

10

absolute ground in either the material or the mental. And this craving manifests itself for both in the notion of representation. In the case of realism, “representation is used to recover what is outer”; in the case of idealism “it is used to project what is inner” (ibid.), meaning that either the mental is conceived of as a mere epiphenomenon of the material, or the material is conceived of as a construction of the mind. In other words, the represented part is always the part that stands further from the absolute ground that either approach has chosen to accept as certain. A debate between these two perspectives is therefore ultimately bound to result in a chicken-and-egg discussion about what was first: a pre-given external world or an inner cognitive system. But rather than responding to this discussion and picking a side, the intention of the authors is to not let it arise in the first place. Instead, their epistemology is one in which chicken and egg, world and perceiver, specify each other. The aim is to rather than crave for an absolute ground, “embrace the fundamental groundlessness of all phenomena” (Vörös et al., 2016: 191), an aim which Varela derived from Buddhist philosophy, more in particular the notion of ‘emptiness’ (or sunyata).

In order to justice to this epistemological standing point, first- and third-person

approaches are placed on the same footing. Indeed, the third-person tools of cognitive science are therefore combined with the first-person methodologies as developed by phenomenology. But this does not mean that enactivism’s attitude towards traditional cognitive science is necessarily a hostile one; it rather views itself as a necessary extension and complement aimed at transforming the style and the theoretical underpinnings of the research community. In this way, rather than a precise, finished description of what the mind and cognition are like,

concepts such as enaction are meant to be metaphors that serve a heuristic purpose.

Following this ‘classic’ version of enactivism, a whole range of scholars has moved away from the traditional cognitivist way of studying the mind and cognition. This has led to a rather diverse range of projects, in which cognition is not only conceptualized as enacted and

embodied, but also as extended, situated, embedded and (bodily) affected: the 4E approaches. However, the many differences that exist between the specific arguments made in these approaches have led to the question whether it is still justified to speak of approaches with a central theory at its core (e.g. Kiverstein and Clark, 2009. But see also Vörös et al., 2016). Indeed, the question is still open as to what extent the different approaches are actually consistent with each other.

And as Vörös et al. (2016) point out, it could well be argued that the differences between these approaches can be traced back to their relation with the deeper philosophical concerns that were at the core of classic enactivism. Because as they argue, the philosophical core of enactivism in the end is essential in such a way that “one cannot incorporate enactive views on the mind into cognitive science without changing the epistemological and metaphysical framework in which it is embedded” (Vörös et al., 2016: 194). More specifically, the differences between the exact arguments made with the different E-acronyms might to a considerable extent be traced back to their different conceptions of symmetry in subject-world relation. As Vörös et al. argue, “the main issue […] revolves around how much constitutive force we should ascribe to the agent in her relation to the world” (ibid.: 195). (As a side note, it could be argued that the epistemological idea of knower and known mutually specifying each other here turns

(12)

11

into a more ontological argument concerning agent and environment specifying each other. This will be dealt with later in chapter 5, but in this chapter I will follow the scholars in the enactivist community in referring to their ‘symmetrism’ between subject and object as an epistemological concern.) In other words: when we want to understand the interaction between and organism and its environment, do we give relative priority to the ‘interiority’ of organisms or agents – thereby moving closer towards a more idealist position – or do we prioritize the (material) environment of the organism, meaning that we come closer to a more realist position by accepting notions of pre-given entities?

Although ‘symmetrism’ seems to be a core tenet in Varela’s classic enactivism for holding a middle ground between the Scylla and Charybdis of idealism and realism, it turns out to be very difficult to uphold this ideal of symmetry when developing the enactivist framework more in detail. Whereas some defend a view in which the environment that organisms encounter is pre-given in a way (e.g. Noë, 2012), others give relative priority to the ‘interiority’ of organisms. This latter position is found especially in the 4E-approach which is commonly referred to as

autopoietic enactivism. This approach having a strong background in biology, it argues that although the relation between subject and world is reciprocal, it is ultimately a-symmetrical. In the words of Thompson, “although the physical and energetic coupling between a living being and the physicochemical environment is symmetrical, the living being modulates the

parameters of this coupling in a way the environment typically does not” (2011: 121). Opposing the living with the non-living, it holds that a living being’s capacity to regulate its interactions with its environment leads to an a-symmetrical relation between both. And this is important, because (as will be discussed more in detail later), there is a direct link between such a

conception of the symmetry between subjects and the world and its understanding of agency. And since it is especially autopoietic enactivism that has gone into to this link, and it is also this strand in the literature that has the clearest conception of agency, it is this approach that will be taken as a basis in the remainder of this thesis (which will from now on mostly be referred to as enactivism, without autopoietic).

2.2 Agency in enactivism

In order to make sense of the conceptualization of agency in current enactivism, it should in the first place be understood in its relation to the relevant theories in cognitive science and

neurophilosophy. Because despite the broader philosophical background of enactivism, its arguments are in the first place aimed at cognitive scientists rather than philosophers in

general. As a result, agency is not – as philosophers would tend to do – treated in the context of the structure-agency debate, but rather in the context of its (more practical) use by cognitive scientists, as for example in the project of artificial intelligence. Although it could be argued that the underlying epistemological debate about symmetrism can be related to the structure-agency debate in a way (Varela et al.’s symmetrism for example being read as an attempt to overcome the structure-agency dichotomy), the scholars themselves do not refer to it when conceptualizing agency. Instead, the focus is more on demarcating what defines an agent, an aim which coheres with attempts to study agency empirically, as in cognitive science the concept of agency has the benefit that it more neutrally captures the notion of a behaving

(13)

12

system, without stepping into heavily loaded debates surrounding notions such as ‘conscious being’, ‘mind’, or ‘intentions’. As a consequence, empirical evaluations of the practical and theoretical implications of using a certain definition of agency may have an effect on its definition.

In the autopoietic enactivist community, several accounts defining agency can be found (e.g. De Jaegher and Froese, 2009; Di Paolo, 2005; Froese and Di Paolo, 2011; Torrance and Froese, 2011) but the most comprehensive account is probably that of Barandiaran et al. (2009). As a starting point, Barandiaran et al. provide the following ‘non-controversial’ basic definition of agency: “agency involves, at least, a system doing something by itself according to certain goals or norms within a specific environment” (2009: 369). Three observations can be made on the basis of this definition. First, this definition mainly attempts to describe what an agent is; it is therefore assumed that agency is something that belongs to an individual, distinguishable entity (note that individual here is not linked to the idea of a human individual). Second, this entity (or system) does something by itself so that it is distinguishes it from its environment. And third, this system does so according to a certain goal or norm. These three observations are developed into three requirements that form the basis of Barandiaran et al.’s definition of agency: individuality, interactional asymmetry and normativity.

The requirement of individuality firstly is a consequence of the aim of developing a concept of an agent rather than the concept of agency. For Barandiaran et al., agency is

something that can only arise with the existence of a distinguishable entity and the constitution of its individuality. “Neither a specific environment nor agentive relations with this

environment can exist without the constitution of an agent as an individuated system” (p. 369). The problem here is obviously what exactly establishes an individual as an individual. Why do we choose to regard certain elements together as an individual (or system), and others not? Which elements are part of the individual, and which are not? As Barandiaran et al.

acknowledge, “what belongs to the clock ‘as a clock’ or to planetary systems remains unclear without an observer introducing an arbitrary criteria [sic] on the definition of the system in terms of the function it sub-serves for her” (p. 369-370). However, the observation by some other agent does not seem to be a proper ground for defining individuality. Instead,

Barandiaran et al. therefore propose to define the individuality of agents (in contrast with systems in general) on the basis of their own capability to define their identity as an individual, thereby distinguishing it from their environment (a point that is intricately bound with the concept of autopoiesis, which I will return to later). This defining is done as an ongoing endeavour through the actions that the individual generates.

In order to define itself as an individual agent, this entity should be the source of activity and not an entity that only passively undergoes the effects of external forces. This idea is captured by the second requirement, that of interactional asymmetry. Although agentive individuals are structurally coupled to their environment, they are capable of modulating this coupling with their environment from within, which means that an asymmetry arises between the environment and the individual agent. The sole responding of a system to its surrounding is not agency. A sailing boat whose course is (passively) determined by a combination of its physical structure and the external forces of water and wind for example is therefore not an

(14)

13

agent. In order to be an agent, an individual needs to be “capable of engaging in some modulations of the coupling and doing so at certain times” (p. 372).

However, not all interactions that take place with a well-identifiable individual as the genuine source of this interaction qualify as agentive actions. The authors for example hold that the spasms of someone suffering from Parkinson’s disease are not actions. For an interaction to count as an agentive action, the authors argue, this interaction should not be random, but must be following certain goals or norms according to which the agent is acting. This is the basis of the third requirement of normativity. Norms here are not universal laws, such as an element following the laws of gravity. In order for something to be a norm, it must be a failure if it is not followed. That is to say, there must be a certain purpose that is acted upon, because if there is no purpose, it would not make sense to speak of a failure either. When the sun ceases to follow the laws of nature for example, this is no failure. For a human being however, certain

interactions with their environment can be either a success or failure, because they can

contribute to their continuation as an individual or to the fulfilment of a certain other norm. As with the self-constitution of an individual, these norms are defined by the agent itself, not by some outside observer. Still, this needs to be qualified: the origin of norms can sometimes lie outside an individual (for example in the case of social norms), but it is always the individual itself who internalizes them, establishes them as its own norms, and acts according to them.

Together, these three requirements (which all need to be fulfilled in order to be able to speak of an agent) form a minimal definition of agency. But in order to make sense of where these three requirements come from and how these relate to each other, it is important to understand how their principle of organization is based on the notion of autopoiesis. After all, it is this concept that this branch of enactivism is named after. Autopoiesis originated as a

concept in cell biology developed by Maturana and Varela (1980) in order to define the nature of living systems. As a definition, autopoiesis refers to the ability of a system to reproduce and maintain itself. An autopoietic system “is organized […] as a network of processes of production […] of components such that these components:

1. continuously regenerate and realize the network that produces them, and

2. constitute the system as a distinguishable unity in the domain in which they exist” (Varela, 1997: 75).

Autopoiesis is thus closely related to agency’s first requirement of individuality. It is precisely by continuously regenerating itself and its boundary that an autopoietic system demarcates itself from its environment as a unified and integrated system, or as a living individual (Barandiaran et al., 2009: 375). An individual is therefore not a pre-given entity, but only exists through the actions by which it maintains itself. An individual enacts itself.

By being the essence of what constitutes an individual, autopoiesis also underlies the requirement of normativity. Because it is the purpose of a living individual to regenerate itself that forms the basis for its normativity. It is because some things are favourable for its

continuation and others are not that it is possible for a system to say that things are good or bad for it, or more general, that things matter for it. An autopoietic system (or an individual) can

(15)

14

make sense of a certain situation, or give meaning to its environment only because of its

autopoiesis. If a system would not have such an internal normativity (which we might even call a teleology), there is no ground for appreciating the world in a certain way, because it would not matter what effect this world would have on it.

And subsequently, this argument leads naturally to the requirement of interactional asymmetry. As we saw, it is only when a system can modulate its coupling with its environment that we can speak of an agent. This is to say, a system that responds to its environment only by internal reorganization is not an agent. In order to be an agent, a system’s adaption needs to not only take the shape of internal adjustment, but also of operating across physical boundaries. An agent only starts to act when it actively regulates its interactions with its surrounding for adaptive purposes, which means that these interactions become of an asymmetrical nature (e.g. Di Paolo, 2005). And it is only on the basis of an agent’s normativity that such active regulations can be initiated. Systems without such a normativity have no incentive to regulate their

coupling to an environment, and have no ‘guidelines’ for doing so. And because this normativity is based on a system’s autopoiesis, the interactional asymmetry is ultimately based on an agent’s autopoiesis.

In this way, enactivism’s understanding of agency is one in which agent and environment are structurally coupled, but at the same time do not affect each other in the same way. That is to say: in a case in which an autopoietic system is only structurally coupled with its

environment, this means that environment and system are mutually affecting each other and are thus mutually constraining. This is a first ‘loop’, and things are still symmetrical here

(illustrated by image A in figure 1). However, when this structural coupling is in turn modulated on the basis of the internal normativity of a system, this implies a second, higher-order loop, which in turn implies asymmetry (illustrated by image B). And it is only when this second loop emerges that we can speak of acts, or agency, and only then that the self-constituting (or autopoietic) system in the coupling becomes an agent. Hence, whereas the spasms of a person suffering from Parkinson are not actions, a bacterium performing certain metabolic chemotaxis can be an agent.

2.3 Contextualizing agency: emerging autonomous domains and location

In this way, enactivism’s understanding of agency reads like a consistent application of Maturana and Varela’s concept of autopoiesis. And although it allows for a refreshing

perspective on agency in which it at least is not exclusively bound to humans or consciousness Figure 1: Agency as a second-order loop (source: Di Paolo, 2009).

(16)

15

in general, it is not without problems. Firstly, there is the problem that it emphasizes the interiority of living systems or organisms. Especially with the principle of interactional asymmetry, a kind of ‘bio-idealism’ appears to emerge in which the interiority of the agent is prioritized over its external environment. If there is an agent asymmetrically regulating the coupling with its environment, this seems to entail a perspective in which a certain autopoietic core of an individual pre-exists the supposed symmetrical enaction of subject and object. And although this is in itself no problem, it seems to be in tension with the epistemology (or ontology) that forms the very basis of the enactivist argument. Aren’t subject and world meant to co-emerge? And shouldn’t therefore more weight be given to the constitutive role of the environment? Secondly, it might be argued that this account of agency is a very narrow one that understands it as a purely biological and system theoretical concept. However, would it not be possible to conceive of agents acting at other levels of organization than the level of biological and metabolic organization? Is the autopoietic framework in its current form sufficient to explain operations at for example the social level, or – maybe even more difficult – interactions with technology?

These are important questions to answer. Not only in order to more thoroughly understand enactivism’s conception of agency, but also because it is important to know how enactivism deals with these issues in order to relate it to ANT. This section will therefore address these questions. Treating them as related issues, it will first outline the notion of emerging domains. With this notion in place, we can then say more about how enactivism tries to make sense of non-biological domains in its framework. Subsequently, it will relate this to enactivism’s understanding of the relation between agents and their environment, and use this in order to describe the way enactivism localizes agency. In doing so, this section will rather than take a position on the plausibility of the enactivist argument, aim to present the way in which enactivism attempts to deal with the questions asked here.

Starting with the concern of the role of non-natural categories in the enactivist

framework, a main way of dealing with this has been enactivism’s introduction of the system-theoretical notion of non-reducible relational domains. By using this notion, Froese and Di Paolo (2011) mean to emphasize how interaction between autonomous entities can open up a new relational domain that itself cannot be reduced to the entities that enabled its very

emergence, but acquires an autonomous organization of itself. Although the original interactors have enabled the emergence of the new interactive domain, their own individual normativity can now be affected by the newly emerged structure. In order to grasp what the authors mean by this, we could imagine two agents (for example two human beings) sharing an environment. In this situation, the acts of one agent A can have an effect on the other agent B in such a way that the sensorimotor system of B is affected, just like other elements of agent B’s environment can do. And when agent B responds to this stimulation, an interaction process could start in which both agents respond to the acts of the other, meaning that we can speak of coordinated behaviour. In other words, the agents become coupled to each other as they are with the rest of their environment. However, in this case the agents are not alone in regulating their coupling, because the other agent also regulates its coupling. In other words, their coupling becomes a third-order coupling in which the agents regulate a regulated coupling (see figure 2). And this

(17)

16

means that the interaction process now itself becomes an autonomous process in a way, meaning that it can in turn affect the agent’s on which it depends. The normativity of the individual agents alone is not enough to make sense of the interaction process, but instead, the interaction process becomes self-sustaining. We could therefore say that just as the agents themselves are enacted, so is the newly emerged social domain. And hence, enactivism argues that “interindividual relations and social context do not simply arise from the behavior of individual agents, but themselves enable and shape the individual agents on which they depend” (De Jaegher and Froese, 2009: 444).

This concept of emerging autonomous domains is not only applied to social interactions, but also to mare basic processes. At a biological level, it could also be said that “even though the interactions between an animal’s neurons can enable the emergence of a domain of behaviour and cognition, the latter domain cannot be reduced to the functioning of the nervous system” (Froese and Di Paolo, 2011: 3). And to move to an even more basic level: the emergence of an autonomous individual (like an organism) can be enabled by interactions between chemical elements, but this organism cannot be reduced to these elements. The argument remains that although these domains only emerged on the basis of the relations of pre-existing entities, they now acquire an autonomous character themselves (autonomy being used as a more general version of autopoiesis – the use of which usually is restricted to organic matters – referring to systems that generate themselves as a unity as an organizationally closed system). And in this way, the concept of emerging autonomous domains is used to link all main concepts of enactivism, and thus all different scopes of analysis with each other.

To explain this very briefly, we take basic individual, autonomous entities as a theoretical starting point. Then, we move to (metabolic) systems that are able to modulate their coupling to their environment (agents). When agents in the form of metabolic systems subsequently form a new system together with non-metabolic systems (such as nervous systems), a new, more specific domain of mentality emerges. In this domain, it are system identities rather than molecular identities that are maintained, because the acts of such systems are underdetermined by chemical metabolic reactions. These acts are therefore called cognitive acts (meaning that cognition is conceived of as only a more specific form of general acting, namely one that is irreducible to its enabling metabolic substrate). A next step then is the interaction between different cognitive agents, which is where we enter the domain of the social. And one step further, we will also (especially among humans) find social interactions that cannot be

Figure 2: Schematic representation of multi-agent interaction. Regulating the interaction process itself, a third-order coupling emerges (source: Froese and Di Paolo, 2011).

(18)

17

explained by only looking at the individuals that are physically present in an interaction. Instead, such behaviour might be guided by norms that pre-exist the interaction itself. Here we might enter again a new domain, namely that of culture. (Figure 3 provides a schematic

illustration of the relations mentioned here.)

Without going further into the details of this argument for reasons of space, it is

important to understand how the phenomena in different domains relate to each other. Because on the one hand, there is a certain operational asymmetry between the different domains, since the more basic, biological domains are necessarily (historically as well as developmentally) the basis of the more specialized domains (such as the social domain), which could not have emerged without the first. However, this does not mean that the interaction between domains is one-sided. Instead, once a higher-order domain has been established, the enabling individuals in the underlying domain might in turn depend on this newly emerged domain. As Froese and Di Paolo argue, “even cultural norms can be re-inscribed back into the normativity operative on the metabolic level” (2011: 29). And importantly, we therefore cannot speak of a hierarchical dependence between the different domains. Instead, the aim of this perspective of emergent, non-reducible domains is to bring forward a view in which there is a non-linear

interdependence between phenomena from different domains. In turn, and importantly, this means that “the investigation of any particular phenomenon can be used as an entry point from which to explore the entire range of the enactive approach” (ibid.: 5). In other words: no a priori

Figure 3: Schematic illustration of the relation between the domains of the enactive approach and the core concepts belonging to them. To use the original caption: “any inner layer necessarily depends on all of the outer layers, although complex relations can obtain between layers

in both directions such that the emergence of new domains transforms the background conditions of operation. The necessary and sufficient conditions for each qualitative phenomenon specified at the bottom of a layer (e.g., ‘sense-making’) are specified by the operational requirements at the top

(19)

18

priority is given to material, natural or biological phenomena over non-material phenomena or ‘higher-order’ phenomena such as human-made institutions.

So it is in this way that enactivism attempts to provide one systematic framework that not only accommodates natural phenomena, but also social or cultural phenomena, at the same time maintaining a conceptual continuity “from life to mind and from cell to society” (Froese and Di Paolo, 2011: 30). Indeed, what it aims to show is how enactivism does not give another epistemological or ontological status to non-biological phenomena such as social processes than to biological phenomena such as cells. At the same time however, it is obvious that there

remains much work to be done here, because in practice the higher order domains such as social and cultural processes (let alone issues such as technology) are still underdeveloped. Enactivism is well aware of this problem, which it refers to as ‘the cognitive gap’, and proposes that interdisciplinary collaborations might be needed in order to do progress (notably, Froese and Di Paolo explicitly mention anthropology in this regard (2011: 29)). However, a more fundamental problem might be that although enactivism argues that there is no hierarchical dependence between lower and higher-order phenomena, there still is the historical or

chronological dependence. In time, enactivism can only explain social phenomena for example when material categories are already in place. And in this way, natural categories do seem to be prioritized over higher-order categories.

Leaving this issue to later sections in this thesis, a second thing to note about the notion of emerging domains is that it also is meant to show how a perspective in which autonomous individual systems asymmetrically interact with their environment, can at the same time retain an understanding for the fundamental intertwinedness of these systems with their

environment. What the idea of emerging domains in a way aims to do is to make less absolute the boundaries that demarcate an individual from its environment. Because although one individual autopoietic system can be operationally closed, it can give rise to dynamics that acquire an autonomous status of themselves. And in this way, enactivism’s definition of agency does not entail to say that an agent is a sovereign autopoietic individual mastering its

environment, for the agent is not the master of its actions. When the interactions of an agent lead to the emergence of a new domain, this domain can in turn affect the normativity of the agent (let’s say that cultural norms are formed which are then internalized by the agent). Therefore, despite the fact that agents are operationally closed, they are not independent in determining their behaviour.

Indeed, the reliance on the notion of interactional asymmetry should not conceal the fact that enactivism maintains a view in which agent and environment are fundamentally

intertwined. Because despite this ‘asymmetric’ conception of agency, enactivism does not see the environment of an agent as the passive surroundings of this agent which are open to be mastered by its will. In fact, the word ‘environment’ does for enactivism not refer to the pre-given surroundings of an organism at all. Instead, as the organization of an organism has emerged in a specific interaction with the surroundings, this organization determines which aspect of the environment are ‘relevant’ to it, or in other words: which elements can affect it or are needed for its continuation. In this way, “the environment is not just what lies outside the system as demarcated from the observer’s point of view but is specified by the system through

(20)

19

the set of boundary conditions that affect it” (Barandiaran et al., 2009: 375). And more fundamentally, the emergence of an individual is understood as the result of a process of interaction. Not only must an individual continuously interact with its environment to self-generate, but its whole constitution is based on the interactions through which it enacts itself. An individual is therefore still seen as deeply entangled with, or structurally coupled to its environment.

Still, this does not take away that the reliance on the notion of asymmetry does seem to entail that enactivism prioritizes the agent over its environment, or the subject over the object. Indeed, Wheeler (2010) for example has criticized enactivism for that by being too internalist, it is not capable of making sense of how components outside of a system can be part of the cognitive (and thus agentive) process as well, or in other words, how cognition and agency can be extended processes. Thus, that as a consequence of enactivism’s priority to the internal side of the organism, it has a too strict conception of the boundaries of an organism.

However, what criticisms such as that of Wheeler miss is that the enactivist argument, agency is not conceived of as a property of an agent. In order to exemplify this, Di Paolo’s (2009) response to Wheeler is illustrative. Discussing cognition instead of agency, he argues that cognition is a concept that emerges in relational processes, and is therefore not something that takes place inside one system. Although cognition has an asymmetrical source in the

normativity of an individual system, it only emerges in an interaction, and is therefore

essentially relational in nature. Therefore, it is neither internal to an organism or a system; and nor is it external: cognition in itself has no location. In the words of Di Paolo, “it simply makes no sense to point to chunks of matter and space and speak of containment within a cognitive system. Inspect a baby all you want and you’ll never find out whether she’s a twin” (2009: 19). In this way, cognition is not bound to brains or the boundaries of an organism; cognitive

engagements always also involve processes outside of the body.

This argument can then be easily coupled to enactivism’s understanding of agency. Because as outlined before, there seems to be no fundamental difference between agency and cognition apart from a difference in specificity. Cognition is only a more specific form of agency, namely one that is bound to the mental domain. And although this point has so far not been explicitly made in the enactivist literature up till now (after all, enactivism in the end is more focused on explaining cognition than on explaining agency), it seems to be a logical

consequence of the structure of the argument. Therefore, just as cognition is not located in a certain chunk of matter, neither is agency. Some confusion arises because talking about an agent does imply talking about a specific, demarcated individual. This makes it seem as if agency is something that is contained within this agent. However, following the reading of the enactivist argument as outlined above, this is not the case.

As a conclusion, we could therefore say that agency in enactivism just like cognition has no location. Agency emerges in asymmetrical interactions guided by the internal normativity of an agent. This asymmetry however does not mean that the agency is a property that can be ascribed to the agent in this interaction. For this very interaction does not exist without the environment of this agent. As figure 4 illustrates, agency emerges in the coupling between the system and its environment, and is therefore a relational concept. What is more, the agent itself

(21)

20

cannot be understood in isolation, because it constitutively depends on the environment for its very existence. Indeed, as Aguilera argues: it is interactions themselves that allow entities to emerge as agents enacting the world (2015: 260-261). This means agents are a result of a process as much as its cause. Agency should therefore not be conceived of as a property of a pre-existing actor, but as something which is enacted just as much as cognition is. The question that

remains then is to what extent this leads to a thoroughly relationist understanding of agency. Because despite the fact that the agent itself only emerges in the interaction, the autopoiesis that ultimately is the source of the agency seems to have already been there, without a need to emerge. Although an agent’s normativity can be affected by its environment, the fact that it is a normative (i.e. autopoietic) system is not a consequence of this environment. Does this mean that agency ultimately has its source in the essence of a living individual? Chapter 4 and 5 will elaborate on this question further.

3. ANT

Probably even more so than for the 4E approaches, it is difficult to speak of Actor-Network Theory as a consistent, systematic theory. In fact, it is not even agreed on whether ANT should be conceived of in the first place as a methodological approach, a social theory or a metaphysics. It is therefore difficult to present the arguments of such an approach in a way that does justice to it as a whole. Instead of attempting to describe ANT in the most comprehensive way, the aim here will therefore be to highlight its theoretical core and its metaphysics rather than its

application in empirical research (which is not to say that this part of ANT is less interesting or important). This is because it is this core that underlies its conception of agency, and that thus gives a deeper understanding of how agency is made sense of in ANT.

In doing so, it will mainly focus on the work of Bruno Latour. There are drawbacks to this choice, as Latour’s version is not the only version of ANT, and the other way around, it could be argued that not all of Latour’s work is ANT. In addition, attempting to describe Latour’s thought in a systematic way is not without risks. As Latour has not just been called not only an

(22)

21

anthropologist of science, but also a sociologist of associations, political ecologist and

philosopher of modernity (e.g. Blok and Elgaard Jensen, 2011: 19), connecting observations from all such different corners of the intellectual realm without paying much attention to the borders that usually divide them, he is difficult to pin down. And this is even more so because of his style of writing; Latour uses a playful, rather metaphorical style. One in which putting forward interesting theories, ideas and stories is just as important as providing a consistent foundational theory (if it is not more important). His goal is explicitly not to approach the style of the

sciences. Because, as Latour states, it is with him just as with most French thinkers that his “style is part and parcel of his very philosophical argument” (1987: 96), instead of being a mere means of communication. Rather than putting in plain language what he has to say, the goal is to find new metaphors, a new language to describe reality with. An attempt to still describe Latour’s thought in a relatively straightforward and systematic way is therefore at risk of misrepresenting it.

However, it seems to me that the advantages that are to this focus on Latour outweigh such disadvantages. The benefit of focusing on one author in general is that it allows for more consistency and clarity in the description of ANT (although consistency is never a primary goal with Latour), and the advantage of Latour’s work more specifically is that its motivations have in a sense always been metaphysical or ontological. Rather than providing a specific theory that describes the functioning of the world in certain causes and effects, his thinking seems to in the end always be related to his broader objective of providing an alternative to the dualistic

thinking of modernity. This, as will be explained later, makes it a suitable account to get a more thorough understanding of ANT’s conception of agency. In addition, I hold that it is especially this characteristic of Latour’s work that makes it interesting to compare it with enactivism. Indeed, although enactivism and ANT have so far been unconnected, as mentioned earlier, it has been said that Varela came to feel growing intellectual affinity with the work of Latour in the end of his life (Cohen Varela, 2002). And it is likely that this affinity is based more on fundamental or metaphysical grounds than on ANT’s method and more detailed arguments.

Before going into Latour’s work, it is important to situate it by pointing to the underlying motives for it. Because despite his ontological motivations, his work is not about contemplating from a distance. Instead, his philosophy is rooted in actual social and political practices; it is not set out to be a systematic ontology, but rather one that triggers reactions. Because in the end, his motivation for his project of finding alternatives to the modernist thinking is his all too real observation that modernity is bumping into its limits, the ecological limits of the planet being the most prominent example in this respect. Indeed, it is this topic that is gaining a more and more central role in his thinking over the years, making it more and more politically engaged (e.g. Latour, 2004; 2013; 2016). The reason that Latour’s work still can be called ontological is that Latour holds that if the goal is to find alternatives for modernity’s thinking, it would be foolish to use the very tools and theories that have produced it. What Latour therefore attempts to do is to address modernity’s thinking at the most fundamental level. So in this way, his motivations are metaphysical as well as practical. And although Latour’s philosophy can hardly be separated from his social engagement, this thesis will not highlight this side as much as Latour’s more theoretical, metaphysical side.

(23)

22

This chapter then will proceed as follows: first, it will introduce Latour’s work by

providing a more general outline of Latour’s thought and its relation to modernity. Next, it will set out the theoretical core of ANT as well as some of its core concepts, attempting to show how ANT can be conceived of as social theory as well as metaphysics. It is only when this

understanding is in place that it is possible to make sense of how agency is understood in ANT, which will be the aim of the last section of this chapter.

3.1 Latour as a non-modernist

Maybe even more than for Varela et al., an attempt to resolve (what they as) the dualistic fallacy it is at the core of Latour’s thinking. Because it is not only the body-mind or subject-object duality that he is addressing, but also that of society and nature, between fact and value, of human and non-human and – indeed – that of structure and agency. It is what Latour (with Whitehead) calls the ‘bifurcation’ of nature that he is addressing, which is “what happens whenever we think the world is divided into two sets of things: one which is composed of the fundamental constituents of the universe – invisible to the eyes, known to science, yet real and valueless – and the other which

is constituted of what the mind has to add to the basic building blocks of the world in

order to make sense of them” (2005: 225–26). At the core of this bifurcation, Latour argues, is the modernist striving to find a minimal certain ground as the basis for our thinking. Whether it is in Descartes’ cogito, in Kant’s ding an sich, in the (post-)structuralists language and world, and even in Heidegger’s Being qua Being, this Latour holds to be the essence of modernity. Indeed, Latour describes the work of modernity and the related critique as “that of a reduction of the world into two packs, a little one that is sure and certain, the immense rest which is simply believed and in dire need of being criticized, founded, re-educated, straightened up… Out on rough water, the critique always looks for a lifeboat” (Latour, 1987: 85). Latour’s thinking instead is one in which the rough see is accepted as it is. In fact, one in which “the oceans are the only firma terra” (ibid.). Instead of taking one part of reality as certain and then stretching it to all of reality, we need to accept all of reality as reality.

If there is one realm in which the dualistic thinking of modernity is produced, it is of course in the sciences. It is therefore no coincidence that Latour started off as an anthropologist of science (and still, ANT as a whole should in the first place be localized as a program in the science and technologies). Describing scientific practices ranging from Nobel-prize winning scientists in a bio-medical laboratory in California (1979) to a group of biological scientists in the rainforest of Brazil (1999) to Louis Pasteur revealing microbes (1988a), he puts forward a new understanding of science. Traditional scientific accounts describe science as a process of a sovereign scientist studying certain objects by the use of a scientific method, and then finding certain results. Against this ‘clean’ account, Latour introduces a much messier account in which not only scientists and their passive study-objects play a role. Instead, there are all sorts of other entities at work, such as technological artefacts, social hierarchies, conviction techniques, industries with certain interests, and even the study objects themselves. None of these entities are passive. Instead, they can all play an equally active role in the process, and it is only because of the way in which they are linked together in a network that they become active as they are

(24)

23

(leading Latour to speak of all of them as actants, of which more later). The scientific process then involves the association of all these different types of actants – human, non-human, institutional – in a network so that certain results are produced.

Scientists themselves then can be seen as a kind of politicians that are involved in the mobilization and representation of all these types of actants in a way that suits their interest. The main process in which they are involved is one of translation. Latour’s description of a group of scientists studying the movement of the border between forest and savanna in Brazil (1999: 24-79) might serve as an example. Looking only at the end-point of this scientific process, we find a report in some Parisian office that refers to a rain-forest which lies thousands of kilometres away. However, argues Latour, this is a distance that needs to be bridged by all kinds of small steps, each of which involves a translation. First, there is a group of scientists that travels to the rainforest in order to study it, using certain theories. Then, there are many tools that are needed in order to do so. One example is a specific box (‘pedo-comparator’) that is needed to collect different types of soil. It is because of this box that the Brazilian soil can become mobile. Subsequently, this box is transferred to a certain laboratory, which allows it to become studied in a particular way. And after being studied, the results are written down using certain writing devices. All of these steps – of which there obviously are many more than only the ones described – are essential for traveling the distance between the Brazilian soil and the final research-report. Neither of them is neutral in the sense that it leaves the forest as it

‘originally’ was, as all of them perform a small translation of the original forest that enables it to be linked to a package of paper lying in a Parisian office.

This thus leads to a picture in which scientific knowledge and facts are not in a special, higher position lifted above the rest of social reality. Instead, they are fabricated in certain social practices; they emerge from the messy negotiations of scientists, their associations with

technologies and other actants, from practical tasks. The activities that take place in scientific arenas such as laboratories are not special in a way that makes them reach a higher sort of ‘truth’ than other social activities. As Latour quotes Knorr-Cetina (1981), “nothing extraordinary and nothing ‘scientific’ [is] happening inside the sacred walls of these temples” (Latour, 1983: 141). The only thing that happens there is the fabrication of facts. Facts that are not eternal truths that have always been there and are now suddenly ‘found’ by scientists, but that have been actively produced by the practices of these scientists. Rather than describing reality, what scientific texts do is to add something to the world.

It is not difficult to imagine why this description of science has led many scholars to the criticism of relativism. It obviously is in full opposition to the way in which scientists

themselves usually picture their activity, which normally is as a ‘clean’, objective process that follows certain strict philosophical norms such as falsification. But as Latour argues, this picture is a result of scientists carefully trying to hide all practical, technological, political and social actants that are involved in their research. In this way they separate that what is natural, eternal and thus ‘Scientific’ from the social, meaning that only those things that are essential to the purpose of the scientist remain visible. Latour refers to this as purification. The facts and knowledge that science produces should be seen as black boxes, in which the actants are made