Disassembling the actant: A valediction to actor-network theory

Introduction

On 9 October 2022, Bruno Latour died, aged 75, after a long illness with pancreatic cancer. His death was widely and deeply mourned, with prominent obituaries not just in the French press but also overseas, most notably in The New York Times and The Guardian in the UK (Schmidgen and Custance, 2015: 4–10). His Actor-Network Theory is, for many, comparable with the theory of communicative action, structuration theory, genealogy or the habitus. Richard Sennett described him as ‘the most creative intellectual of our generation’ (Jeffries, 2022) and he was ranked as the 17th most cited sociologist in the 2010s (Korom, 2020). He made a profound – and controversial – contribution to sociology.

A great sociologist is dead. Yet, this article is not an obituary. Eulogies have already been published by his colleagues and friends (Czarniawska, 2022). While acknowledging the scale of Latour’s contribution, this article rejects Actor Network Theory. Sociology is currently at a cross-road, which reprises a similar conjuncture over a century ago when Gabriel Tarde and Emile Durkheim were engaged in a debate about the future of sociology. Tarde was defeated, sociology took the direction which it did, heavily influenced by Durkheimian concepts of the collective. Bruno Latour saw himself as the anti-Durkheimian successor of Tarde. He wanted to build a new programme of sociology which returned sociology to its authentic Tardian origins. The question is whether, after his death, his hugely influential research programme, Actor-Network Theory, should become a dominant paradigm in sociology. This article claims it should not. It argues that just as Durkheim superseded Tarde a hundred years ago, sociology of the 21st century should, therefore, seek to marginalise ANT. It should affirm its Durkheimian roots. In this way the paper contributes to a neo-Durkheimian consensus which is apparent in some Anglophone sociology today.

Of course, the contemporary intellectual situation today is plainly more diverse and diffuse than at the time of Tarde’s death. At that time, there was a more or less binary choice in France, between Tarde or Durkheim. Today, there are other important currents of sociological thought. So, any re-emergent Durkheimian consensus today will be limited; it will be far from absolute. Weberian, Marxian, feminist, and post-colonial theory are equally important currents in sociological theory. Yet, in the light of the influence of ANT, sociology might be at an important conjuncture. In order to analyse an increasingly digitised, automated society, will sociologists adopt Latour’s ANT, with its flat ontology and empowered actants? Will, in short, the repressed Tardian tradition return? Alternatively, might the emergent neo-Durkheimian consensus be capable of understanding the accelerating effect of new technologies in an intensely globalised world in social and collective terms? Clearly, it is impossible to adjudicate definitively on this debate in a single paper. However, this article seeks at least to direct sociologists away from ANT, in order to advance the Durkheimian consensus which seems to be emerging. It does this by focussing exclusively on one issue: the actant.

Sociologists should reject ANT because it is profoundly flawed philosophically. There are a several errors which his critics have noted in Latour’s work but, in order to refute ANT, this paper focuses on one: Latour’s concept of the actant. The actant has featured heavily in contemporary Latourian scholarship (Appadurai, 2015; Janicka, 2022; Khong, 2003; Sayes, 2016; Zielke, 2022). These scholars claim that actants are central to his whole system. I agree and this paper draws upon the insights of these scholars. Yet, it makes an alternative argument. In order to sustain ANT, Latour invested material, non-human objects, actants as he called them, with agency. Together they influence, channel, enable and steer humans to create and sustain networks. However, in order to explain networks, Latour is finally forced to retreat to a surreptitious appeal to groups; to human communities involved in collective projects. In effect, Latour is a covert Durkheimian, relying on the very collective phenomena which he dismisses.

The paper begins by examining the current conjuncture in sociology arguing that while Latour represents a new Tarde, it is also possible to see the emergence of a neo-Durkheimian school. In order to support this school, the paper then describes ANT and the role of the actant within it. It then goes on to expose the fallacy of the actant, principally by an analysis of Latour’s The Making of Law. It offers a neo-Durkheimian analysis of the law against Latour’s approach. Finally, the paper considers whether recent technological developments in Artificial Intelligence might posthumously vindicate Latour’s concept of actant. It claims that even AI cannot redeem the actant. In each case, the paper promotes a Durkheimian approach to sociology in which the human social group remains the prime explanatory variable.

The new Tarde

Throughout his career, Bruno Latour professed ANT as a radical alternative to conventional sociology which he saw as an extension of the work of Emile Durkheim. Like Durkheim, sociologists explained action as the product of social facts, collective social processes, and group membership. Latour rejected this faith in collectives as a form of metaphysics. In place of Emile Durkheim, he advocated Gabriel Tarde as the true originator of sociology (Tarde, 1903, 1969, 2012; Latour and Lepinay (2009), Latour et al. (2012); King, 2022).

Tarde was a jurist, criminologist, and sociologist in Fin de Siecle France; he held the Chair of Philosophy at the College de France from 1900 to his death 4 years later. He was Durkheim’s main opponent and rival. Tarde attempted to establish the new discipline of sociology in the late 19th century, but he proposed a quite different approach to Durkheim. Tarde rejected Durkheim’s appeal to social facts, the ‘substratum’ and the collective conscience. For Tarde, all science addressed the question of repetition; science explained how coherent, ordered repetition was possible. This was true of sociology. The sociologist’s role was to explain how regular, repetitive social action – that is, social order – was possible (Tarde, 1899, 1902: 20). This was not obvious because, according to Tarde, society could empirically consist only of individuals and their interactions; there was nothing else to it. Anything else was mere metaphysics. For Tarde, individuals were manifestly independent and discrete entities with their own motivations, self-identity, and will. Influenced by Liebniz, Tarde, therefore, argued for the science of ‘monadology’. Sociologists should explore the interactions, the flows of communication and influence, between individuals. Social order was possible, not because of some invisible collective forces, but because individuals were bound together by the rays of imitation, emanating from other individuals. Humans were inherently susceptible to each other. They could not help but be influenced by each other. Imitation was an intrinsic and ineluctable human capacity (Tarde, 1903: 37, 68); Tarde was fascinated by the phenomenon of hypnotism which was extremely popular at the time (Tarde, 1884: 509). For Tarde, imitation explained social order because humans did not imitate randomly. They were hypnotised by charismatic individual leaders, whose imitative rays were strongest, or by individuals who professed ideas, which were plainly superior. The result of these syllogistic duels between the best ideas and desires was that, as one individual imitated another, humans coalesced into vast social chains, unified around the same desires and beliefs (Tarde, 1897: 365; Clark, 1969: 69–70: King, 2022).

Bruno Latour himself accepted that large parts of Tarde’s system were odd: ‘we don’t accept all of Tarde’s idiosyncracies’ (Latour, 2005: 13). Nevertheless, against Durkheimian social facts, Latour found Tarde’s concepts of monads and imitation liberating: ‘Tarde refuses to consider society as a higher, more complex social order than the individual monad’ (Latour, 2002: 119). Individual actors had to be continually and persistently enrolled and recruited for collective phenomena to exist at all.

Tarde died in 1904 and, as Latour lamented, was almost immediately forgotten. Throughout his own life, Latour wanted to reverse this history. ANT was his attempt to resurrect Tarde and to re-direct sociology along the monadological path it should always have taken. Latour wanted to dispense with the social groups, human relations, and collectives as the critical explanatory variables for understanding practice. Latour, in fact, went further than Tarde. Tarde’s sociology was completely individualist; Latour claimed that sociologists have to include natural objects in their accounts. Where conventional sociology asserted the power of the human group, ANT was inclusive, even promiscuous; nothing was excluded. Non-human beings, objects, and entities were equally important. Social life consisted of the association of human and non-human things. Latour proposed a flat, Tardian ontology where things and people were invested with equal significance; they influenced each other mutually. Latour is now dead. The question is whether the history of sociology will repeat or reverse itself? Will Latour and his ANT project be forgotten as an interesting but failed project, as Tarde’s was, or will it attain a dominant position in the discipline, as Latour hoped?

There is some evidence that ANT might succeed, where its ancestor had failed. In an increasingly transnational, de-terroritorialised environment, in which data, artificial intelligence and automation are becoming ever more important, traditional sociological categories, like class, nation, gender, race, ethnicity seem to be inadequate. They seem to communicate a cohesive homogeneity which is alien to hybrid contemporary social realities. A Durkheimian commitment to social groups, collective consciousness, and social facts all seem hopelessly archaic. They seem all too human. They do not begin to address the appearance of disruptive technologies, algorithms based on machine-learning, or drones, which seem to be able to think and to act for themselves.

Latour’s ANT has already been hugely influential in sociology. Actor Network Theory very quickly transcended debates in Science and Technology Studies where it began, to be applied and utilised widely by many other scholars, who had no interest in science and technology at all (Law 1986; Law and Hassard, 1999). Latour became ‘the prince of networks’ (Harman, 2009). Indeed, the concept of a network became almost ubiquitous in the contemporary sociological lexicon, even when Latour himself was not acknowledged (e.g. Castells, 2009). Actor Network Theory’s influence has extended well beyond sociology too. In law (McGee, 2016), architecture (Yaneva, 2022), international relations (Barry, 2013), and security studies (Grissom, 2006; Wasinski, 2019), many scholars have actively embraced the theory. In cultural sociology, the influence of ANT is pronounced. For instance, many sociologists in this tradition are interested in the construction of personal identity. Sociologists have been deeply impressed by the rise of new personal digital sensors, which monitor individual performance. Deborah Lupton (2014) has analysed the way these devices are reconstituting human agency. Digital actants are reconfiguring human agency. The question is will ANTs influence grow posthumously.

The Durkheimian revival

Notwithstanding the great efforts of Latour and his followers to invalidate the sociological canon, the revival of Emile Durkheim’s work in the last two decades is very noticeable. There is little doubt that some of Durkheim’s work and especially The Rules of Sociological Method (1966) were problematic. There he confidently declared that: ‘Social phenomena were things and ought to be treated as things’ (Durkheim, 1964: 27). Latour was right to object to this kind of claim. However, a closer and more sympathetic reading of Durkheim suggested that he was not reifying society but only emphasising the centrality of collective factors in social existence. In Durkheim’s later work and, of course, in The Elementary Forms (1964), he clarified his position. These works have become the key reference point for many sociologists today. In The Elementary Forms, Durkheim sought to identify the origins of religion in order elucidate the universal social process of group formation. He sought to explain how society came into being. Famously, he argued that when the members of an aboriginal clan came together for religious rites, they believed they were in the presence of their totemic deity. Durkheim fully accepted the reality of the experience. However, he argued that in fact, what they took for their God was in fact, their social group. The effervescence which they felt so viscerally could be traced to the ecstatic interactions with their fellow clans-people. As a result of these intense emotions, the aborigines committed themselves to each other and to their clan, symbolised by their totem, morally, emotionally, and intellectually: ‘Since religious force is nothing other than the collective and anonymous forces of the clan, and since this can be represented in the mind only in the form of the totem, the totemic emblem is like the visible body of god’ (Durkheim, 1964: 221).

Because the emotional commitment cathected in the ritual faded, clans, and indeed, all social groups need to engage recurrently in collective rites of unification: ‘A society can neither create itself nor recreate itself without at the same time creating an ideal. This creation is not a sort of work of supererogation for it, by which it would complete itself, being already formed; it is the act by which it is periodically made and remade’ (Durkheim, 1964: 422). Societies and social groups exist only insofar as their members come together and to develop a common understanding of themselves as a special community, sharing special bonds with each other and committed to particular joint enterprises. This work of unification involves both an act of imagination and intense emotional expression, which are embodied for group members by the shared symbols they all recognise. Scholars have been impressed by Durkheim’s explanation of how humans form groups in the course of their interactions, uniting conceptually and morally around common totems, which they collectively invest with significance. The Elementary Forms has become a major reference for many contemporary scholars.

For instance, two of the most important schools of sociological thought in the US have actively embraced later Durkheim as their inspiration. In the 1980s, Jeffrey Alexander committed himself to a form of multi-dimensional functionalism which was significantly indebted to Talcott Parsons (Alexander, 1998). Yet, he subsequently moved towards Durkheim (Alexander, 1988, 2003). In his most important work of his Durkheimian period, The Civil Sphere, Alexander (2008) has sought to identify the collective, social underpinnings of political debates, especially in the US. For him, the civil sphere constitutes a collective consciousness, prescribing the limits of acceptable political debate and policy. His more recent work on objects and drama similarly seek to apply a Durkheimian methodology. Physical objects and drama are constituted by collective processes (Alexander, 2014, 2020). With Philip Smith and Ron Eyerman, Alexander has established the Centre for Cultural Sociology at Yale. This Centre has been deeply influenced by Durkheim’s later work and especially by his discussions of classifications systems. In a series of important studies, the Yale ‘Strong Program’ has sought to show how shared cultural categories, especially ‘binaries’, have been drawn upon to structure debates and action in a variety of spheres (Alexander, 2008, 2009, 2010, 2011; Kurakin, 2019; Mast, 2012; McCormick, 2015; Smith, 2014, 2020). For Alexander and his colleagues, collective performance is critical to the creation and sustainment of these binaries. They exist only insofar as they are activated in practice. Above all, the binaries are empowered in ecstatic public performances. Philip Smith, for instance, has discussed how categories of ‘near’ pollution are collectively organised (Smith, 2014). In a fascinating recent paper in this tradition, Kurakin (2019) analysed the bizarre 1959 Dyatlov pass tragedy, when a group of ski-tourers disappeared in a very remote area in Russia. Their bodies were eventually found, half naked over two separate sites. Various conspiracy theories emerged attributing the deaths to murderers, animals, radioactivity, UFOs and Zombies. Kuratin shows that these explanations were not the objective product of the facts, but reflected the identities and interests of identifiable groups.

Independently of the Yale Strong Program, Randall Collins, at the University of Pennsylvania, has demonstrated a similar trajectory to Alexander. Max Weber was a major influence in Collins’s early work on conflict sociology in the 1970s (Collins, 1979, 1986, 2009). However, in the last twenty-years, as Collins has articulated his Interaction Ritual Chain Theory, the debt to Durkheim has become ever clearer – and Weber less so. Collins has attempted to apply Durkheim’s micro-analysis of social solidarity across a range of social activities including sport, philosophy, and violence (Collins, 2000, 2004, 2008). In each case, Collins shows how interactions generate emotional energy which determine the outcomes of the situation. The Durkheimian tradition is vibrant and growing in the US.

It is equally vibrant elsewhere. A former member of the Edinburgh School, Don Mackenzie’s work on technology and his most recent work on banking might be situated broadly in this tradition (Mackenzie, 2022). In that work on the digitisation of large-volume financial trading, Mackenzie fully recognises the potency of the algorithms which banks now use to buy and sell stock ‘at the speed of light’. They seem to be powerful actants. Yet, his work actually highlights the regulatory and organisational changes which have facilitated this transformation. Governments, regulators, banks and investors – social groups – finally determined whether and how algorithms were used. It was very noticeable that the ‘flash crash’ of 2014 was quickly stabilised by a human intervention when bankers realised something was wrong. Other sociologists have also been deeply influenced by the Durkheimian tradition (Ablitt and Smith, 2019; Bowring, 2016; Smith, 2021). In his work on stand-up comedy, Daniel Smith has shown how comedy is in fact a product of the collective interaction of the audience and the performer (Smith, 2021).

For all these scholars, the later Durkheim provides a decisive theoretical framework. They try to explain a variety events by reference to the interactional patterns, and the groups involved or arising from them, which are constitutive of them. Neither groups, still less their interactions, are pre-established. On the contrary, neo-Durkheimian scholars attempt to show how, through their interactions, actors actively unify themselves to create common beliefs and emotions. This aim of this paper is to affirm and extend these Durkheimian traditions by showing the flaws in Actor Network Theory and, above all, by highlighting the contradictory character of the Actant. In every case, when Latour appeals to an Actant, it is possible to identify collective human agency – a social group formed through interaction – at work to constitute that object.

The actant

Bruno Latour developed his Actor Network Theory early in his career. The outlines of the theory might be seen in his 1979 book Laboratory Life (Latour and Woolgar 1986). By the 1980s, with the publication of The Pasteurisation of France and Science in Action, ANT was fully developed. Latour designed ANT for STS. In order to understand ANT, it is therefore worth returning to its origins and to debates in STS at that time. Latour’s work was a reaction to a number of strands in sciences studies at that time, but its relationship to the Edinburgh School is particularly revealing. In the early 1970s, the Edinburgh School, led by Barry Barnes, David Bloor, and Don Mackenzie, had developed what they called the ‘Strong Programme’, drawing on Wittgenstein, Hume and Durkheim. The Strong Programme was not relativist: it never disputed the successes of science. The natural world had an objective existence. Scientific discoveries could be tested against that reality and shown to work. Nevertheless, in order to understand why certain theories were established, it was not enough, according to the Edinburgh School, to adopt a naïve realist view; scientists did not simple discover a single, objective reality. Faced with complex and often ambiguous empirical findings, scientists might adopt a variety of responses, any one of which might be fruitful. In the end, the tribunal of validity was not objective reality on its own but the scientific community. The collective interests, orientation, and goals of scientists finally determined what constituted a valid, successful, and useful finding. Scientists provided the decisive frame, not nature. Latour was impressed by this radical epistemological move: ‘It is the glory of the Edinburgh school of the social studies of science to have attempted a forbidden crossover. They used the critical repertoire that was reserved for ‘soft’ parts of nature to debunk the ‘harder’ parts, science themselves!’ (Latour, 1993: 54). The Edinburgh School undermined the pristine objectivity of scientific theories by identifying their social constitution.

Nevertheless, Latour was unconvinced by the Edinburgh School’s Strong Programme. In the end, in Latour’s view, the Edinburgh School reified society over nature; they over-invested in the social interests of scientists. They believed (according to Latour) that the pre-formed, exogenous social interests of scientific communities determined their results: ‘Society had to produce everything arbitrarily including the cosmic order, biology, chemistry and the law of physics! The implausibility of this claim was so blatant for the ‘hard’ parts of nature that we suddenly realised how implausible it was for ‘soft’ ones as well’ (Latour, 1993: 54–55). For Latour, the Edinburgh School wrongly dismissed things. According to Latour’s reading, the Strong Programme asserted that social interests drove scientists to adopt certain theories, whatever natural reality – things – might be telling them. Scientists were determined by their pre-formed, extraneous interests, be they political, financial or institutional, which they imposed onto the evidence. The Edinburgh School were social determinists. Scientists constructed natural reality.

Latour’s dispute with the Edinburgh School is very useful for understanding his reaction to sociology in general. According to Latour, the entire tradition of classical sociology was flawed. The fundamental problem of sociology is that since its origins in the late 19th century it ontologised the ‘social’. The pioneers of sociology invoked the ‘social’ as a master causal variable. Human agency was to be explained by the force exerted on individuals by pre-existing social reality: social forces, social facts, social interests, social density.

Latour wanted to unmake the ‘social’. He wanted to undo easy and lazy sociological presumptions about pre-existing social categories. Instead of presuming that social reality is always already formed, ANT was an attempt to trace its actual formation. Latour was interested in process, not society or structure. He was highly sensitive to the contingent, temporary, and situational character of social life. Consequently, against a sociology of society, he called for the study of associations: ‘It is possible to remain faithful not to the original institution of the social sciences by redefining sociology not as ‘a science of society’ but as the tracing of associations’ (Latour, 2005: 5). Latour rejected the concept of pre-formed structures or collectivities: ‘For ANT, if you stop making and remaking the groups, you stop having groups. No reservoir of forces flowing from ‘social forces’ will help you’ (Latour, 2005: 35). ANT, therefore, forbade sociological deus ex machina from his system; it was impossible to argue that social reality was the automatic and inevitable product of pre-existing (but unexplained) economic, ideological, cultural or political facts; ‘In ANT, it is not permitted to say: No one mentions it. I have no proof but I know there is some hidden actors at work behind the scenes’ (Latour, 2005: 53). To explain the existence of social groups, it was necessary to plot the way their members continually mobilise themselves, enjoining each other into their common ventures. ANT was an anti-essentialist programme (Latour, 2005: 5). Whether Latour was right about the Edinburgh School and all of the other sociologists he rejected is questionable. However, with ANT he laid out an invigorating manifesto.

Actants are a crucial, if not a decisive, element in this endeavour to rid sociology of the social. But what is an actant? What work does it do in Latour’s system? There are many useful passages in Latour’s work articulating this. At one level, Latour’s definition of an actant is unproblematic; it is just a non-human physical object which is crucial for any social action to take place. An actant ‘can literally be anything provided it is granted to be the source of an action’ (Latour, 1996: 7). Alternatively, an actant was ‘any entity that modifies another entity in a trial’ (Latour, 1988: 237). For Latour, actants had agency. A carpenter’s hammer provides the simplest examples of an actant’s agency. Without a hammer, a carpenter cannot drive a nail into a piece of wood. Carpentry is impossible without the hammer. The hammer makes that social practice what it is: ‘If you can, with a straight face, maintain that hitting a nail with and without a hammer, boiling water with and without a kettle. . . are exactly the same activities. . .then you are ready to transmigrate to the Far Land of the Social and disappear from this lowly one’ (Latour, 2005: 71). Because a hammer transforms the capabilities of the human carpenter, Latour claims that the hammer itself can be said to exert agency. That agency has to be included in sociological accounts.

Indeed, Latour accorded the agency of actants equal status to human actors (Latour, 1988). Thus, Latour repeatedly called for ‘a principle of irreduction’, an ontology of irreducibility; nothing can be ignored or reduced. Each entity has its own status and agency. As he states in We have never been Modern: ‘Objects are not the shapeless receptables of social categories’ (Latour, 1993: 55). Consequently, networks do not consist just of individual human actors allying with each other, but of non-human actants too. Indeed, actants play an equal role in the formation of networks: ‘We should consider symmetrically the efforts to enrol human and non-human resources’ (Latour, 1987: 258).

The actant was not just empirically important in social life. The actant played a critical theoretical role in Latour’s system, holding ANT together. The philosophical challenge presented by ANT was prodigious for Latour. With ANT, Latour wanted to refute any appeal to collective entities; but he still wanted to explain the existence of extended social orders which bound human actors. This presented a problem for him. The danger was that his denial of the social group could reduce his sociology to individualism. Indeed, if it consisted of an ontology of autonomous, independent individuals, his sociology might collapse into a form of rational choice theory, game theory, or existentialism. Plainly, that was never his intention. Those programmes failed to explain how social order extended over time and space. As Talcott Parsons famously noted with his Utilitarian Dilemma, unless rational, independent individuals were determined by something outside themselves, interaction between them could not generate social order (Parsons, 1937). Individualist (utilitarian) theories always collapsed on the free rider problem; if individuals were rational and independent, cooperation was impossible. It was obvious to Latour that individuals could not simply do as they pleased. However, if Latour allowed individuals to form groups and communities, whose members mutually and collectively influenced, constrained and empowered each other, he was back to Durkheim. He had to choose another way. It was here that the actant became significant. The actant was a way of avoiding either collectivism or individualism. The actant had an independent, non-human agency. It was able to constrain humans to a degree, and, therefore, to stabilise networks, without appealing to collective processes. In all of Latour’s ANT analyses, the actant, therefore, plays a prominent role.

Latour uses a number of examples to explain the agency of the actant. For instance, he discusses sleeping policemen – speed bumps (Latour, 1999a: 190). For Latour, speed bumps were not just raised slabs of tarmac. They exercised actual agency; they could physically harm cars which did not slow down. They forced drivers to change their behaviour. Consequently, although not conscious, they were not just passive objectives; they actively enforced speed limits. They were active participants in urban traffic flows. They were as much a part of the traffic as drivers, pedestrians and cyclists; they were not just road features. Therefore, on this basis, Latour claimed that actants could be described as having goals.

There are other examples. For instance, rejecting Heidegger’s concept of technology, Latour contemplated the case of hand-gun used to shoot someone. He examines the debate in the US which is split between the gun-controllers who claimed that ‘Guns kills people and the NRA which asserts that ‘Guns don’t kill people: people kill people’ (Latour, 1999a: 176). Latour disagreed with both – and therefore with many sociologists. Most sociologists would maintain that the gun was just a weapon; it was a necessary resource to perform a murder. The pistol was used to kill the victim; the ballistic properties of the pistol explained the injuries of the victim. However, it was not legally implicated in the murder; the shooter pulled the trigger. Without the gunman, the pistol was just a cold piece of metal. Latour demurred. He claimed that to understand this action properly, it was necessary to recognise the interaction between the gun and the gunman. The act was co-created by the gun and the gunman; they exerted joint agency. Together, they created ‘a new goal that corresponds to neither agents’ programme of action’ (Latour, 1999a: 178). For Latour, actants preferred certain outcomes for which they were designed and, whether consciously or not, pushed the network towards their ends. In this way, actants might be said to have intentionality. They have a purpose which they actively pursue (like Liebniz’s monads), even if they do not have consciousness in the human sense. A network is a contingent aggregate of the goals and the agency of all the actants and all the actors in the network.

It is a distinctive ontology. It seems very strange to imbue objects like speed bumps or hammers with agency; can they really have ‘goals’? Are they really intentional? For all the qualifications, it does seem to commit Latour to some form of determinism. Many scholars think he is guilty of determinism. For instance, in their recent criticism of Latour, Krarup and Blok have rejected to Latour’s denial of human agency: ‘Our concerns start from the suspicion that Latour may not be symmetrical enough in his dealings with (quasi-) objects and subjects, paying much more attention to the former than the latter’ (Krarup and Blok, 2011: 42). For Krarup and Blok, Latour wrongly disaggregates individual human agents: ‘Latour seems in practice to be ‘heterogenizing’ human subjectivities onto a background of materially stabilised, and technologically shaped, assemblages’ (Krarup and Blok, 2011: 57). While humans are derogated in ANT, objects retain their unity. His analysis of hammers and guns ‘leaves us short of understanding, let alone describing, how bodies, symbols, and subjective desires simultaneously contribute to the process of forging socio-technical effects’ (Krarup and Blok, 2011).

So, some scholars have worried about the determinism implied by the concept of the actant. If actants do not at least constrain what humans can do, it seems impossible to explain how a network can remain stable. While human agents are potentially volatile and malleable, physical objects are more secure. Latour never wanted to go this far. Consequently, the ontological status of the actant and its agency remained ambiguous in Latour’s work: ‘We cannot say that an actant follows rules, laws, or structures, but neither can we say that it acts without these’ (Latour, 1988: 160, 174). In the end, he avoided a final definition. Actants influenced humans– but they did not determine them. They did not determine because the effect of actants in any network was dependent own their contingent configuration. No single actant ever operated alone; it could not determine simplistically. Nevertheless, networks or associations of actants might exert substantial force on their human partners. Together actants could provide the skeletal structure around which a network can form and stabilise.

Unscrewing the actant

Officially, no collectives are at work in ANT. Durkheim has been exorcised. In order to abjure the human social group as an explanatory variable, Latour invested agency in things instead. Objects – actants – played a key role in influencing human behaviour and therefore in generating temporarily stable association. Yet, is this really what happens in ANT? Are social groups, and the interactions which constitute them, really as irrelevant as Latour claimed?

Many critics believe that they are not. They have claimed that in fact Latour covertly relies on precisely the collective entities whose significance he denies. David Bloor became one of Latour’s strongest critics in this regard. Bloor wrote a long refutation of ANT in 1999. The article focused on Latour’s writing on science and particularly, on works like his celebrated book, The Pasteurisation of France. There, Bloor observed that Latour’s ‘aim is to produce some manner of non-sociological, non-reductionist analysis of knowledge, one that neither reduces nature to society nor society to nature’ (Bloor, 1999: 86-7). To do this, ‘the entire Durkheimian tradition is dismissed’ (Bloor, 1999: 87). Latour not only rejected the standard view that it was possible to explain Pasteur’s discoveries by reference to the great man alone, but also rejected conventional sociological explanations which examined the human political, social, and scientific institutions which provided a context for Pasteur’s discoveries. According to Latour, microbes were also a crucial part of the research endeavour because their existence provided a grounding and validation of Pasteur’s theories: ‘The Pasteurian hagiography is what makes the real work of Pasteur and his followers incomprehensible since it conceals their own work in a larger whole that includes what others did for them and in their place’ (Latour, 1988: 60). Without microbes, Pasteur’s triumphs would have been impossible. It is an intriguing argument.

Yet, in the end in Pasteurisation and other writings, Latour cannot ultimately avoid an appeal to human social groups. As Bloor notes, ‘after baffling talk about “quasi-objects”’, Latour finally has to resort to something more recognisable and concrete. The ‘co-production’ of nature and society, in fact, ‘resides in “common practices”’ (Bloor, 1999: 98). By common practices, Latour refers to the collective actions taken by a community of scientists who understand themselves to be part of the same research programme. They recognise themselves as a scientific group, with special commitments and obligations – and privileges. For all his discussion of actants, in the end, Latour needs to appeal to a concrete social group, whose members decide together how things will be understood and used by them: ‘He [Latour] is brought back to the same starting point as a sociologist of knowledge’ (Bloor, 1999: 98). In short, ‘he identifies social groups’ (Bloor, 1999: 99). It is noticeable Bloor’s reaction to ANT was shared by other scholars who also criticised Latour for both denying social groups and the processes of interaction which generate them; but then relying on them (Amsterdanska, 1990; Schaffer, 1988; Shapin, 1988; Sturdy, 1991). For instance, the celebrated analysis of the dispute between Renault and the EDF about the development of the electric car, ‘Unscrewing the Big Leviathan’, which he wrote with Michel Callon (Latour and Callon 1981), makes no sense without the existence of those corporate bodies. For all these critics, the actant was a mirage. In the end, Latour relied on the very social categories – human groups, communities, institutions, collective practices, and the interactional processes from which they arose– which he denied.

Bloor and other critics focused on Latour’s writing in the 1980s and 1990s. In particular, they concentrated on The Pasteurisation of France, Science in Action, We have never been modern, and Aramis, in which Latour described how science and technology studies should be done. To refute the actant, it is useful to take a later work, The Making of Law, which does not look at science at all. This analysis will show whether the problem with the concept of the actant was general to his theory, or just specific to his work in science. Latour himself believed that his analysis of the Law was compatible with his work on science (Latour, 2010: 202–203). Latour’s success has often relied on his skilful use of rhetoric which distracts and beguiles the reader. The best way to assess the validity of the actant is to conduct a systematic analysis of a specific case. It is an optimal way of identifying the actant at work. A detailed analysis of one empirical study will highlight the plausibility of Latour’s theory of the actant most clearly.

The Making of Law analysed the Conseil d’Etat [Council of State] which is the highest civil court of appeal in France. Latour explored how the judges on the Council of State decided on the various litigations brought before them. Latour’s analysis in this book is brilliant. It is remarkable how he was able to turn the dry workings of a civil bureaucracy into a thrilling story. The work is a wonderful read; rich, engaging and empirically fascinating throughout.

Latour examples are peculiar, even a little kooky; Mr Delavallade wants to sue the commune of La Rochefaucault for the damage done by the town’s pigeons to his crops. Mr Delavallade claims that the mayor failed in his responsibility to sterilise these avian vermin. Other examples include a drug dealer (falsely) claiming asylum; Latour describes him as ‘toad’ (2010: 220). There is a fussy nit-picker, an incompetent Minister with a decree which is obviously a ‘tissue of absurdities’ (Latour, 2010: 220), and a fatal ski accident. How do the judges make coherent, consistent decisions in the face of such diverse cases? For Latour, this is complex. The judges had to weigh the evidence, classify the case, work out the legal precedents for it, and, finally, make a decision. If that were not hard enough, the judges on the Council of State were not trained lawyers and judges. They were appointed to the Council from politics or government. Although each was permanent, the membership of the Council was therefore fluid (Latour, 2010: 111).

According to Latour, the Council’s files played a key role in its legal decision-making. They were decisive actants. The files contained all the evidence of the case in question, as well as from previous cases. The job of the judges was precise: ‘The entire work consists of establishing the relationship between the two collections of writings. The counsellor must decide. . .by linking the first to the second’. The judges had to align the evidence of a particular case with the law, on the basis of the precedent set by past cases (held in their files). In this process, the files themselves exerted non-human agency on the judges: ‘The judges do not reason; they grapple with a file which acts upon them, which pushes them, and which makes them do something’ (Latour, 2010: 192). The files did not quite compel the judges, but they heavily informed the judges’ decisions, constraining what they could do; the judges’ role was to act as a link between these non-human actants. The files, like Latour’s speed bumps or microbes, have goals; they exerted an agency so that their inherent intentions were fulfilled. For Latour, the Council of State was not to be understood as a collective judicial body, but as a network of actor-actant-actor-actant, or judge-file-judge-file. Only once actants and judges had been aligned was a decision possible. When they had done this, a judgement was possible: ‘The bridge has now been established by fitting elements of the claim into texts and by weaving the means of bringing them closer and closer to the laws and decrees’ (Latour, 2010: 88).

Yet, if we actually look at what the files do and how they are maintained, Latour tells us a rather different story. For instance, Latour writes: ‘Therefore, the nature of the Council does not depend on its equipment, but on the homogeneity of the world of files that are kept, ordered, archived and processed, and upon the homogeneity of a staff that is renewed, maintained and disciplined’ (Latour, 2010: 203). Here Latour tries to avoid determinism. The Council cannot be reduced to its files; it is not determined by its equipment. Nevertheless, decisions depend upon ‘the world of files’. However, as the sentence proceeds, Latour retreats to a different position. The world of the files does not just exist sui generis; it is kept. The question is who actually keeps the files? How is the homogeneity of the files sustained? In the second half of the sentence, we find out. Latour is forced to smuggle a collective human body into his account; he begins to admit an alternative form of agency. Thus, Latour notes that the files – and their homogeneity – depend upon the homogeneity of staff. In the end, and quite against his intentions, Latour admits that the human staff into his account. Precisely because they are homogeneous (an expert, cohesive group, united in goals and outlook), the counsellors and the staff are able to maintain the files, as a useful, organised resource. It is possible to reinterpret this sentence then. The files are important enablers for the Council but the world of the files is finally dependent upon collective efforts (the ‘homogeneity’) of the staff, its judges, secretaries and administrators. The judges plainly were crucial here. To decide, they have to unify themselves around a single interpretation. This seems a long way from ANT and rather closer to conventional sociology.

Let us explore the process of making law more. In an important passage, Latour describes the constituent elements of how the Council makes the law in detail. He lists key attributes which are necessary. They include:

This is classic Latour. It is ingenious. It is funny. It is fascinating. By listing all the things the Council used to make a decision, Latour seems to disaggregate the agency of the Council, back into its micro human and, above all, non-human elements. The corporate body of the Council is dissected into its constituent parts, to show that the locus of the decision-making was diffuse, not centralised; it was an assemblage, not a collective.

Can Latour really deconstruct the Council back into its individual actants and actors? Are the interactions between judges really irrelevant to their collective decisions? Crucially, at decisive points in this passage, Latour lets his ANT account slip. For instance, in the list, he documents the ‘variable authority of the members’, ‘the organisation of the flow’, ‘the modification of interest’ and ‘quality control’; these processes are central to the law. In each case, however, these attributes originate in the Council as a collective social body – not in the files. Every single one of Latour’s six points can be reduced to the judges and the Council. For instance, Latour explicitly notes, in point (1), that the judges – not the files – decide on the authority of individual members. Similarly, the progress of the cases depends on the ‘continuous action of lawyers, reporters, revisers and judges’ (2). The presidents of the sub-sections organise the files (3); the interests of the members determine their judgement of the cases (4); the weight which is placed on case-law depends upon a decision taken by the judges (5); the ‘members of the Council exercise quality control’ (6). For all the talk of actants, ANT reverts back quite banally to humans in an identifiable social group. Above all the judges as a collective body are decisive.

Of course, Latour tries to cover his tracks with rhetorical devices. He tries to conceal that he has ever appealed to the judges as a social body. For instance, in points 2 and 5, he repeatedly employs passive constructions which obscure the human subjects. Instead of simply saying in point (2) that the progress of a case depends upon whether lawyers, judges etc. accept or reject legal arguments, the legal arguments ‘prosper or are rejected’. Legal arguments become the subject of the sentence. In point (5), instead of saying that the judges decide on the evidential significance of case law in each case, the ‘weight of case law’ depends on whether ‘a decision is taken up’. This enables him to avoid identifying a definite subject. It conveniently suggests that the decision was taken by a more nebulous, contingent assemblage, rather than an identifiable human group. Agency is diffused. The judges do not decide. Decisions are made by a human-non-human assemblage. Nevertheless, despite this rhetoric, the fact remains that in the end, the agency of the judges as a collective body is crucial to his account of the making of law.

In his criticisms of Latour, David Bloor makes a subtle Durkheimian point about the relationship between humans and objects which is immediately pertinent to The Making of Law. Bloor states: ‘the systems of beliefs are the medium through which people coordinate their shared interactions with non-social nature’ (Bloor, 1999: 88). Bloor was not advocating nominalism. The world and the objects in it are real. However, to achieve anything human beings must cooperate. In order to cooperate with each other, individuals must finally agree to some shared understandings about the world and how to use the things in it to their collective advantage. If they want to cooperate the members of a social group have to assign a common status to objects (Barnes, 1974). They look at those things and use them in the same way so that they can work together. Even the most technical object has a quasi-totem like status then; it is imbued with a common significance. In the case of the Council of State, the judges similarly ascribed a common status to the files and the evidence in them in each case so that they could cooperate; they could make a decision. Their deliberations were precisely the process of coming to a collective agreement about the status of these objects.

A Durkheimian reading of the law

Surprisingly, Latour gives us no evidence about the Council’s deliberations. Perhaps, he focused on the files rather than the discussions and interactions of the judges because he wanted to maintain the Council’s anonymity and confidentiality (Belorgey, 2004: 113). It seems unlikely. The decision to focus on the files, and not the judges, seem to be a deliberate ANT strategy. Latour was able to sustain his ANT interpretation, only by suppressing key pieces of evidence about the Council’s operations. Had Latour focused on the deliberations between the judges themselves, it seems likely he would have come to a different interpretation. As a result of his detour into bureaucratic apparatus – into files, mail-boxes, and folders – the process of how the judges actually discussed cases together, how they weighed evidence, how they applied the law so that they could come to a consensus, is completely absent. As Krarup and Blok have noted, ANT analysis seems convincing, only because the yawning gap in sociological evidence, allows Latour to denigrate the agency of human actors.

Latour did not provide any empirical evidence about how judges came to their shared agreements in the Council of State. Fortunately, many examples are available which might provide some cognate evidence of how this happens. They might allow a Durkheimian reading of the law. Indeed, there has recently been a notorious example of judicial decision-making in the U.S. which is very useful in this regard: the Dobbs v. Jackson Women’s Health Organization, when the Supreme Court overturned the Roe v. Wade decision on 24 June 2022. This episode is a nearly perfect case study for assessing the validity of Latour’s claim that files themselves exert legal agency. Whatever we might think of the decision ethically or politically, the reversal of Roe v. Wade is deeply interesting sociologically. In 1973 and 2022, the US Supreme Court made two radically opposing decisions. In 1973, it legislated that abortion was a constitutional right for women; in 2022, it withdrew that right. The Constitution was a key resource in both cases. Yet, the Constitution, on the basis of which both decisions were justified, was exactly the same. The same document legitimated two diametrically opposed decisions made 49 years apart. Manifestly, the Constitution cannot have compelled the judges to do anything in these cases; the same words facilitated two completely contradictory decisions. To explain the differential application of the Constitution in 2022, we have to look to the Supreme Court and the judges themselves.

The Supreme Court has been moving rightwards for decades. However, as a result of the death of Ruth Ginsburg in 2020, and President Trump’s appointment of Amy Coney Barrett, the Supreme Court’s political orientation changed decisively. From 2020, five of the nine judges declared themselves as ‘Originalists’: Amy Coney Barrett, Clarence Thomas, Samuel Alito, Neil Gorsuch, and Brett Kavanaugh. In addition, Chief Justice John Roberts is sometimes an originalist. With the appointment of Barrett, the originalists held a majority in the Supreme Court for the first time; it requires six judges for a quorum (Cole, 2022a). The Originalist judges argue that the Constitution should be applied in 2022, as it was intended by the Founding Fathers when they wrote it in 1776, and when subsequent Amendments were added. In the case of abortion, the five judges ruled that when the 14th Amendment (utilised by Roe in 1973) was adopted in 1868, abortion was a crime in most states (Cole, 2022b: 18). Consequently, since abortion was not recognised legally by the Constitution, the Originalists argued it could not be used to defend it now.

Of course, it would be wrong to deny the significance of the Constitution as an artefact in this case. Latour was right that objects are vital to social life. The Constitution, as a physical document, was very important to the Originalists. It was their prime reference point. It was invested with a sacred status by them, by the Supreme Court and, indeed, by all American citizens. Precisely because of its place in American society and politics, it provided the judicial legitimation for their decision. But note: it could not have exerted independent agency itself. The Originalist judges – not a document written in 1776 and its later Amendments – arrived at a shared interpretation of the Constitution. For instance, Erwin Chemerinsky, the legal scholar, has demonstrated that the Constitution can have no intrinsic ‘original’ meaning: ‘for most constitutional provisions, there is “no original meaning” to be discovered. Instead, there is a range of possibilities that allows for exactly the judicial discretion that originalism seeks to eliminate’ (Chemerinsky, 2022). The Constitution has no intrinsic goals. Rather, in the Dobbs case, the judges made a collective decision on the basis of their shared reading of the Constitution. Their decision relied not on the words in the Constitution (which could be variously interpreted), but rather on their own political orientation. That orientation provided the interpretative framework investing the Constitution with specific meaning. Of course, the Originalists drew on wide-ranging support from the right-wing of the Republican Party and its radical constituency especially in the South. They ultimately embodied its will. The Dobbs v. Jackson Women’s Health Organization is a very good example of Durkheimianism at work. A political group in the US, represented by Supreme Court judges, invested an object – the Constitution – with shared collective significance in order to unite themselves around a common political project. The Constitution was not an actant.

At the end of The Making of Law, Latour observes that the ‘ethnography of law resembles the ethnography of science’ (Latour, 2010: 248). For Latour, both demonstrated actor-networks at work. Actants, not just actors, were critical. Yet, it is possible to make a different claim. The ethnography of law may not so much resemble the ethnography of science, as Latour conceived it, but rather the ethnography of religion which Durkheim pioneered (Durkheim, 1964). There Durkheim described how, in the course of their worship, aboriginal clans-people united themselves around a common sacred totem, which they invested with a shared meaning. Like Durkheim’s aboriginal clans-people, the judges of the Council interacted with each other until they eventually unified themselves around concrete artefacts like the files, which they imbued with special, collective significance. From Latour’s description, it seems unlikely that there was ever much effervescence in this august office of state. Nevertheless, the unification of clans-people and judges around a common shared belief, embodied in physical artefacts which come to represent the group, share a close family resemblance. For all the discussion of actants, it has no explanatory purchase on what is actually going on. Indeed, there is an irony to Latour’s actant. Latour repeatedly criticised sociologists for appealing to pre-formed social forces which could not be specified (but which determined individual action). With his actant, Latour falls into a similar error. In order to account for the existence and particular configuration of a network, Latour invokes actants, which operate as black boxes separate from human agents. They are rather like the social forces he disparages.

ANT and AI

Latour may have failed to convince everyone about ANT in his lifetime. His work relied on an appeal to the actant which was simply not sustainable. Ironically, and in stark contrast to Tarde, Latour may be vindicated posthumously though. Today, sociologists are facing a new challenge: the rise of artificial intelligence. In the last two decades, there have been remarkable developments in machine learning. There have been a series of major breakthroughs. For instance, in March 2016, Demis Hassabis’s AlphaGo beat the world champion Go player, Lee Seedol, 4-1. Most commentators thought it would be impossible for a computer ever to master Go. Even more strikingly, in game three, AlphaGo made an extraordinary move: ‘It’s not a human move’ (du Sautoy, 2019: 31). It won the game. It was not just that a computer programme had learnt to play Go; it seemed to have developed an intelligence and creativity which exceeded even the best human player. Consequently, many predict the rise of genuinely independent, general artificial intelligence in the near future – the singularity (Bostrom, 2017; Kurzweil, 2005). In May 2023, Geoffrey Hinton, the renowned computer scientist and AI pioneer, left Google, claiming that AI now posed an ‘existential threat’ to humanity. Digital actants with neural networks, the workings of which their own engineers do not fully understand, seem to be emerging.

How can sociology possibly come to terms with the rise of machines which can think and act? They seem to defy Durkheimian sociology with its focus on the human social group arising from visceral interaction. ANT may be ideally placed here. Latour presciently recognised the power of actants decades before machine learning developed. The algorithms of AI represent precisely the kind of actant he has in mind, which had its own agency, its own goals. Algorithms are already directing, assisting, and predicting human behaviour; they will only become more adept. Surely a form of ANT, with actants at its heart, must be the future of sociology?

Sadly, for its advocates, the reality is more mundane. The development of machine learning algorithms is remarkable. Some of their capabilities are breathtaking. Yet, second generation artificial intelligence operates by statistical probability (Lee, 2021: 13; Cantwell Smith, 2019; Cukier et al., 2021; Marcus and Davis, 2019; Mayer-Schonberger and Cukier, 2014). On the basis of a mass of data, algorithms recognise statistical patterns and therefore make probabilistic predictions: ‘Modern AI is based on statistical analysis and so relies on an inductive framework’ (Larson, 2021: 118). AI understands nothing. It does not have consciousness, still less general human intelligence. That would require the ability to interpret, to understand. AI cannot abduct; it is incapable of inferring a cause or an intention from a single fact or event and then building a model of how the world might be on the basis of that interpretation. This is something humans do every day (Cantwell Smith, 2019; Larson, 2021: 160). Second generation AI does not do that. It is brittle and narrow; it only calculates probabilities. However, impressive its achievement, it does not have any intentionality (Larson, 2021). It is just a programme. It is a very capable tool for processing digital information. Like many other machines, it can do things which humans cannot, or would find very difficult and time-consuming to do. Yet, it is not an independent actor.

Moreover, even the greatest successes of AI have often been mis-represented. The successes of AI have been imputed to their software programming, to algorithms and to neural networks alone. It is presumed that because it can process data independently of immediate human intervention, AI is genuinely autonomous, rather than just automated. This interpretation systematically ignores the vital human capital which is required to develop, maintain and refine AIs. For instance, after AlphaGo was defeated in Game three, Demis Hassabis and his chief programme announced they ‘would need to go back and analyse why it had made such a lousy move’ (du Sautoy, 2019: 37). They refined its programme. It was noticeable that when Lee Seedol was defeated in 2016, the team of engineers from the company Deep Mind, which programmed AlphaGo, celebrated: ‘Hassabis punched the air. Team members hugged and high fived’ in the box above the arena (du Sautoy, 2019: 36). The team was unified in a moment of ecstasy. The incident showed something very important: AlphaGo did not really defeat Lee Seedol on its own. Lee Seedol was defeated by a team of brilliant computer scientists, data engineers and software engineers who had been working together on this project for years. They programmed AlphaGo and continually refined AlphaGo’s software in the course of its competitions. For instance, when it played Fan Hui a year before the Seedol game, the match exposed weaknesses in the software which the Deep Mind engineers resolved only by locking down the code (du Sautoy, 2019: 28). It is wrong to claim that AlphaGo won autonomously. Rather, the team at DeepMind used AlphaGo to defeat Lee Seedol. A human team of engineers, programmers and scientists used software – a machine – to help them beat another human. Their interactions in this process were invisible when AlphaGo eventually triumphed. They took place months and years before in Cambridge and at Google’s headquarters in Silicon Valley as the team worked on the software together, discussing, testing and affirming how to improve the system. We return to familiar Durkheimian territory; groups uniting around collective endeavours in the course of their interactions, generating common signs and symbols so that they can work together.

Many social scientists interested in big data and AI in the last decade have been heavily influenced by Latour’s work. Latour’s lexicon of associations, assemblages, devices and infrastuctures suffuses discussions in this field (Amicelle, 2022; Aradau and Blanke, 2015; Valdivia et al., 2022). The implication is that AIs and their algorithms are actants which configure emergent tech networks. For instance, in their work on EU borders, Ana Valdivia et al. have examined the increasing role of datafication and AI ( Valdivia et al., 2022). AI programmes have played an important role in processing data. It might be feared that AI is taking over immigration control; it is becoming a powerful and nefarious actant. In fact, although these scholars defer to ANT, they describe a different reality. In order to use data, border agencies have contracted private companies to support them on lucrative projects. We are seeing not the genuine autonomization of border security; but its privatisation. Private tech companies, employing programmers and engineers, are playing an increasing role in migration control; they curate the data and write the codes which monitor the borders. The locus of datafication is not AI, but the private security companies like Atos or Thales which own the data and the codes. Crucially, in order to win these contracts, these companies are forming cooperative oligopolies around border security with very significant political implications. Valdivia et al. plot the emergence of alliances between these companies to form discrete clusters. These clusters have become powerful interest groups able to influence immigration policy.

Similarly, in his work on the use of AI to regulate financial services, Anthony Amicelle has stressed the human dimension of this work. It is impossible for banks to monitor all their transactions manually; the volume is vast. In order to trace fraud and criminality, they have therefore employed algorithms which alert them to risky transactions. AI has become crucial to financial regulation. Yet, creating and sustaining these algorithms requires human labour. As one of Amicelle’s informants noted: ‘It’s a lot of work, it took a lot of time because you want to make sure that at the end of the day everything has been mapped’ (Amicelle, 2022). Amicelle observes: ‘This quote sheds light on the invisible and hard work that makes big data surveillance possible’ (Amicelle, 2022). That work is human. It is done by expert groups in the banks, under the direction of management. Financial security is a collective product of these interactions between programmers themselves and programmers and managers, as they identify what assets need to be protected and how best to protect those assets. AI has become very potent, but only insofar as humans have designed and re-designed it in the light of changing needs.

AI is remarkable technical achievement. Perhaps, in the future as Large Language Models develop, genuinely autonomous programmes might appear. At that point, perhaps Latour’s theory of the actant might be valid. It is unlikely though. At present and in any foreseeable future, AI is better explained by reference to the scientists, engineers, programmers, and organisations which have developed it, not by any ghost actant in the machine. AI seems to defy ANT. The fact is that, AI, as a genuinely revolutionary technology, remains susceptible to traditional sociological methodology, which prioritises human groups, their interrelations, their practices, their shared beliefs, their interests and their collective goals. Of course, scholars do not have to employ an explicitly Durkheimian approach to analyse AI. A Marxist, Weberian, or feminist approach might be fertile. Yet, the work of Valdivia et al and Amicelle demonstrates that a Durkheimian approach, focussing on the interactions between participants and the subsequent emergence of groups, united by shared beliefs and interests, remains entirely pertinent even to the sociology of AI.

Conclusion

The death of Bruno Latour is a sad moment for sociology. It marks the end of one of the most creative sociological projects in the last four decades. Latour made a major contribution to the discipline. Nevertheless, the contradictions at the heart of Latour’s work are evident. Latour consistently denied the existence of the social group and the collective. Rejecting sociological reification, he tried to eliminate social groups and collective actors from his work. In place of groups, he appealed to actants. He placed things on the equal footing with individual agents. Together, the actants and actors in any network mutually generated the actions. This solution seemed, on the face of it, an ingenious way of avoiding any appeal to human social groups, without falling into determinism. Yet, in fact, throughout his work, Latour found that he had to revert to surreptitious appeals to collective entities. In the end, ANT relied on the existence of social groups. Latour needed laboratories, scientific communities, major industrial enterprises like Renault, public companies like EDF, and councils, whose collective agency he systematically denied, by appealing to the actant. It was impossible for him to explain these lifeworlds without an appeal to human social groups of which they were comprised. On its own terms, ANT simply cannot explain the existence of the very phenomenon for which it had been created: networks.

After his death in 1904, Gabriel Tarde fell precipitously into obscurity. His work became- and has remained – of bare antiquarian interest. In the light of his many adherents, Latour’s work is likely to have a different destiny. However, ANT should endure not because it is correct, but because of its faults. ANT articulates a theoretical position, which sociologists should reject, precisely because it is so plainly flawed. Indeed, especially with the re-emergence of Durkheimian traditions in the US and the UK, many sociologist have already turned away from it. Even sociologists working in Weberian, Marxist, or feminist schools, typically refer to recognisable social groups as the key variables in their work: status groups, classes, genders, patriarchy. The irony is that even as Latour promoted Tarde over Durkheimian sociology, his work is actually one of the most eloquent vindications of that conventional sociology and of the contemporary neo-Durkheimian programme. ANT might, therefore, be read as one of the most ingenious and fertile errors in contemporary sociological theory.