The Return of the Nonhuman: The Reproduction of Anthropocentrism in Social Theory

Separation of Worlds: How Analytical Distinctions Gain Relevance

Distinctions such as that between nature and society often enter analysis as ready-made ordering devices, even in approaches that seek to move beyond them. In the example above, this appears in how the surgery is retrospectively assembled around a recognisable figure of human accomplishment. The distinction does not operate through the exclusion of nonhumans, but through a descriptive organisation in which certain elements more readily appear as action or authorship while others become conditions or support.

This descriptive tendency resonates with a much longer history in which nature and society have been treated as separate domains of reality. Human societies have frequently been understood as existing independently of the rest of the natural world, positioning humans as the primary locus of recognisable action while relegating nonhumans to the status of environment or condition (Descola, 2013). This separation is also visible in the organisation of scientific knowledge, where natural sciences and social sciences have historically operated in isolation from each other (Latour, 2000). Consequently, distinctions between nature and culture often come to appear as given rather than intertwined and co-produced.

A related point appears in work on justification, where distinctions are treated as situational resources rather than fixed structures (Boltanski & Thévenot, 1999). Similarly, engagements with political ecology show how different ‘worths’ of nature are enacted in specific situations, further exposing the contingency of such divisions (Blok & Jensen, 2019). The development of the study of the human species – the ‘anthropos’ – in relation to nature, culture and society has largely been shaped around this ontological divide.

Over the past two decades, scholars have increasingly challenged such separations by arguing for the need to take multiple ontologies seriously (Viveiros de Castro, 2004, 2015) and by examining how realities are enacted through situated practices (Kohn, 2013; Law & Lien, 2013; Mol, 2002; Woolgar & Lezaun, 2013). Yet, discussions continue over whether there is a singular reality with multiple interpretations (interpretations being beliefs that have no ‘real’ connection to ‘the world’), specifically in relation to cultures other than non-Western ones, or if there exist many realities (Latour, 2005). Many traditions within social sciences continue to work with versions of what Latour (1993) calls the ‘Great Divide’ – the separation of the perceived abilities of humans and nonhumans.

These discussions draw attention to an ontological politics that broadens or narrows what can be registered analytically. As Mol (1999, p. 75) argues, combining ontologies with politics proposes that ‘the conditions of possibility are not given’. Likewise, Latour's (2013) discussion of modes of existence suggests that different ways of assembling reality can coexist without being unified into a single ontological scheme. In this sense, ontology multiplies through practice. Not in the sense of generating a variety of ‘truths’, but in the sense of how worlds are enacted in practice: reality is made differently in different practices (e.g. Mol, 2002). It could be otherwise (Woolgar & Lezaun, 2015).

In line with this ontological politics, social life is rarely composed of a single type of actor and we do not live in isolation from each other (van Dooren & Bird Rose, 2016). Ontological distinctions matter not only for how worlds are analytically ordered, but for what kinds of relations and forms of action become thinkable in the first place. Anthropocentric ordering is reproduced through descriptive tendencies that sort worlds in advance. Certain forms of action become more readily recognisable as interventions, decisions, or authorship. Anthropocentrism thus persists not only through who or what is included in analysis, but through how situations are rendered intelligible in the first place.

Entities: How They are Characterised in Theoretical Accounts

It is analytically limiting to assume entities as given. This assumption not only obscures the practical work through which entities are produced, bounded, and maintained, but also leaves unquestioned the criteria by which something comes to count as an entity in the first place. In many analyses, entities appear as finished units, while the processes through which they are composed recede from view. Wildlife management in California offers one example. In Los Angeles, the mountain lion P–22 became widely known for living in Griffith Park, surrounded by highways and urban development. The lion was photographed by camera traps, monitored through GPS collars, and repeatedly described in reports tracking its movements, encounters with humans and other animals, and involvement in livestock predation. The lion also became a cultural figure, sometimes described as the ‘Brad Pitt of cougars’. The mountain lion appears as an identifiable entity moving across a landscape and becoming the subject of intervention. At the same time, what counts as ‘the mountain lion’ is assembled through a range of heterogeneous practices. GPS signals are translated into movement patterns, camera traps produce images that are classified, genetic samples distinguish individuals, and statistics estimate populations. These practices are entangled with changing prey availability, roads and urban expansion that shape where and how lions move. Rather than simply describing the animal, these elements participate in making it present as an object of knowledge. And yet, when these practices are gathered together, the mountain lion appears as something that can be located, counted, and governed. The lion becomes the object that policies address and decisions concern. The processes through which it is composed – data practices, ecological relations, and infrastructural arrangements – remain positioned as conditions, while only certain elements are stabilised into the figure of ‘the mountain lion’. In this way, the object does not precede the practices that bring it into being. What remains less visible are the relations, transformations, and exchanges through which this apparent entity is continuously composed.

The capacities of entities cannot be assumed in advance. Rather, we encounter what Haraway (2008) would call compounds and what Latour (1996) describes as heterogeneous collectives: arrangements of humans and nonhumans together, mixed and composed of various materials, organic and otherwise. Entangled and deeply involved with one another in various ways, sharing histories. Histories that are embedded in language, making us turn to concepts like human and nonhuman in the first place (which is itself part of the problem, since entities come as collectives, already entangled). Perhaps this is what Pickering (1995) refers to when, after Nietzsche, he speaks of the prison house of language. From this perspective, the appearance of entities is an effect of descriptive and analytical practices rather than a property of the world itself. In such a world, there is no ‘anthropos’ to put at the forefront of sociological inquiry – only an ongoing, brewing ontological soup. What this suggests is not the absence of entities, but their contingent stabilisation. And yet, description retrospectively presents these stabilisations as if they were given. What is temporary and contingent is rendered as fixed and self-contained.

This tendency is also visible in much of our mundane discourse that presupposes and exemplifies the assumption that the characteristics of an entity can be associated with a particular bounded space. This is part of what Wright (1990a) calls the object hypothesis. He suggests that one of the biggest questions which philosophers beg is whether there are objects or whether there is a continuum, a ‘flux of varying viscosity’ (Wright, 1990b). What is not sufficiently recognised is that the object hypothesis is a convenient fiction – a method presumed necessary for ordering, representing, describing, and accounting for the world (Woolgar, 1994, p. 226).

Unpolluted entities neither enter nor come out of the ontological soup. From time to time, and often in unexpected and dramatic ways, elements of the soup coalesce. On these occasions, and sometimes for a period thereafter, we then apprehend the mixture as containing some fixed, factual and knowable entities. What is crucial is that this stabilisation is temporary and contingent, even if description retrospectively presents it as given. What goes in and comes out of the soup are metamorphoses of bastards, what Haraway (1991) would call tricksters and shapeshifters. Anthropocentric forms of analysis often struggle to register such entities except by stabilising them into more recognisable forms.

Multispecies studies are one branch attempting to bridge the social and the natural worlds (Kirksey & Helmreich, 2010), promoting a ‘cultivation of attentiveness’ (Rose & van Dooren, 2017; van Dooren et al., 2016). This attentiveness has been directed foremost towards biotic life. Multispecies scholarship has focused on microbes (Schrader, 2012), mushrooms (Tsing, 2015), plants (Choy, 2009; Hustak & Myers, 2012; Miller, 2019; Myers, 2015), and animals (Despret, 2016; Fudge, 2019; Mullin & Cassidy, 2007; Porcher, 2014) – an agenda of undeniable importance in a time of mass extinction (Tsing et al., 2017; van Dooren, 2014). The issue here is one of scope. A biotic orientation leaves abiotic action largely unexplored. Even recent calls for rethinking the nonhuman tend to reproduce this limitation (e.g. Fair & McMullen, 2023). The word ‘species’ also remains tightly associated with living organisms, and not all actors can easily be accommodated within this analytical orientation. While the concept of ‘species’ has been stretched, the underlying assumption of bounded entities often remains intact.

What comes into view is the limit of vocabularies tied to specific types of beings, rather than to wider processes of world-making – where entities transform one another, exchange properties, and metamorphose through encounters. Anthropocentrism persists here not through the exclusion of nonhumans, but through a selective visibility – where some forms of existence readily appear as entities, while others remain difficult to register as such. As with the problem of separation of worlds, entities do not precede relations but are stabilised through them.

Abilities: How Capabilities and Attributes are Assigned in Descriptions

As with entities and actions, the attribution of abilities is not given but achieved in description.

However, capabilities commonly appear as already in place, attached to particular kinds of entities, rather than treated as something that is achieved, attributed, or sustained in practice. What is at work is a descriptive tendency to assign abilities as inherent properties of entities, rather than as effects of relations and arrangements. This tendency resonates with a longer tradition of human exceptionalism, in which particular capacities – such as reasoning, communication, or complex social organisation – are treated as uniquely human. As environmental philosopher Eileen Crist (1999) reminds us, the use of some concepts and abilities simply seems to be off-limits when it comes to other than humans. Even where such claims are empirically challenged, social science explanations often return to them – suggesting how much is at stake in preserving human uniqueness. This issue continues to spark debate (i.e., Brenneis & Strier, 2017; Harms, 2022) and human exceptionalism still finds staunch defenders (Bråten, 2022; Hornborg, 2021). This implies that any displacement of human uniqueness also unsettles the social order organised around it.

When scholars point to what is ‘uniquely human’ – such as complex societies and communication – they often run into empirical difficulty as shown by Chapman and Huffman (2018). Take, for example, the idea that humans are commonly seen as the only ones that build complex (enough) societies and social structures. Such abilities and behaviours can for example be found in animal settings as well (Chapman & Huffman, 2018; Despret, 2016; Lien & Pálsson, 2021). In studies of animal behaviour, chimpanzees may be said to use tools, and crows to solve problems and plan ahead. Yet, such descriptions are frequently accompanied by hesitations, caveats, or methodological caution. A distinction emerges between behaviours that resemble human abilities and those that are granted the status of such abilities in their own right. The same capacity – such as planning or reasoning – does not travel easily across species but must be argued for, demonstrated, and made visible as such. What appears as an inherent ability in humans becomes, in nonhumans, a matter of empirical negotiation. The distribution of abilities is not simply observed but accomplished through particular forms of description. That such demonstrations continue to be necessary suggests how persistently capacities such as reasoning remain implicitly human-centred. Recognising nonhuman capabilities requires work – work that reveals how unevenly the capacity to act is distributed in description.

Despite sustained challenges to the idea that what is taken to be ‘uniquely human’ is unique after all (Despret, 2005; Fudge, 2019; Porcher, 2014), many strands of social scientific explanation continue to return to these arguments. We should remember that the worst kind of anthropology portrays monolithic identities inhabiting clearly consensual societal rules and cultural values. But humans are themselves in constant dispute over what counts as distinctly human, and this has varied historically. Human exceptionalism therefore remains in social scientific explanation.

Given the strength of anthropocentrism, it is not surprising that scholars are expressing a worry over, being met with concerns regarding anthropomorphism (cf. Gibbs, 2020; Puig de la Bellacasa, 2017), which furthers this tendency. Anthropomorphism, the practice of treating nonhumans as if they had what is commonly seen as uniquely human abilities – or God forbid, feelings or desires! This supposedly dangerous practice of trying to attribute agency where there should not be any (Latour, 2017a) is indeed still a minefield in sciences and clearly a risky business. Attributing agency, intentions, or feeling to nonhumans is often treated as a methodological error, a warning against overextending human categories. This caution also limits what can be recognised as action in the first place. In this way, the attribution of abilities is carefully policed: certain capacities are protected as properly human, while others are rendered suspect when extended beyond it. The idea that humans have ontologically unique abilities comes with assumptions that ontology is something static, rather than achieved in practice. Just because scholars accept the proposition that there could be multiple ontologies, doesn’t necessarily mean that they are willing to accept that ontology is situated.

When deciding who can be counted as an actor or not, one thing that commonly guides scholars’ explanations is if there is intention or purpose (or in ideal cases, moral) lurking behind actions. As noted by Despret and Chrulew (2018), a common argument is that objects ‘lack’ intention and therefore they cannot act. There is a tradition of using the word acting as connected to a motivation of some sort, like will, desire, intention, or self-consciousness (Haraway, 1992). Such explanations of actors have served the purpose of effectively separating ‘human agency’ from ‘the effects of objects’, showing the strong position of human exceptionalism in social theory. It seems to be a criterion for being an actor, that there is ‘will’ behind that act for it to count. When looking for actors, look for where there is will. However, it would be a mistake to assume that desire, will, or intentions, necessarily drive action. One way this operates is through the association of action with intention, will, or purpose. Entities that lack such qualities are often treated as incapable of acting. If instead we refrain from approaches that imply some sort of self-consciousness behind acts, an actor can be anyone or anything that is granted to be the source of action (Latour, 2017b). We could use the word actor to simply mean something that acts – something that affects what happens. An actor can be understood as anything that makes a difference to what happens. Then the range of possible actors expands considerably. What counts as action shifts with how it is described.

The assumptions we make shape the explanations and knowledge we generate, and these are not innocent practices. We tend to, perhaps unintentionally, sometimes dismiss nonhumans as less complex, less interesting than humans. Engaging in such practices, reduces the sense of how actively the world is brought about. Human exceptionalism nevertheless retains its attraction. This keeps other forms of action analytically marginal. Where entities are approached as relationally made (Haraway, 2008; Tsing, 2015) (equally true for all entities, technology, plants, animals and humans), the exclusion of nonhumans from social theory becomes harder to sustain.

Human exceptionalism tends to organise explanation in ways that make certain forms of world-making more difficult to register. This is why imagining the time we live in as ‘Anthropocene’ (Crutzen, 2002; Crutzen & Stoermer, 2000; Zalasiewicz et al., 2008) is not without its problem. As Crist (2018) points out, the ‘Anthropocene’ does not necessarily promote new worldviews. Rather than displacing human exceptionalism, the concept often recentres human agency at the very moment it claims to unsettle it (Crist, 2013; Haraway, 2015; Povinelli, 2017). Here again, capacities are distributed in ways that keep humans as primary agents of change. In this sense, the Anthropocene can be read as a contemporary reformulation of anthropocentrism, rather than its undoing.

Keeping humans as the core point of reference reproduces the notion that humans are the most important actors. This keeps in place an analytical privileging of the human as the primary point of departure. The risk of ascribing certain abilities to entities as inherent or unchanging is that it obscures the ways in which relations also contribute to shaping action. Taken together, these descriptive moves do not only organise analytical distinctions; they also shape what becomes possible to think, say, and do. This shows a persistent descriptive tendency to attribute certain capacities to humans as inherent, while rendering them absent elsewhere. Anthropocentrism persists not through the exclusion of nonhumans, but through the terms on which abilities are made visible and allowed to count. As with entities, abilities are not given but attributed.

Symmetry: How It Is Invoked – and Undone – in Theoretical Reasoning

Although approaches might call for treating humans and nonhumans with analytical symmetry, descriptions often reintroduce asymmetries by presupposing different kinds of relevance or effect for different actors. In practice, symmetry is often invoked at the level of principle but undone in description. A system may be said to ‘participate’ in a process, yet when its outputs diverge, it is repositioned as an object of evaluation rather than a contributor. The language shifts, even when the arrangement does not. What changes is not the role of the system, but the terms under which its contribution is allowed to count as action. What is at work here is a descriptive tendency to grant symmetry in principle while selectively withdrawing it in practice.

Symmetry, in this sense, is not simply denied but conditionally granted and withdrawn in practice. It operates less as a stable commitment than as a reversible and situational achievement within description. Anthropocentrism is reproduced in social theory through the reintroduction of asymmetry in description. ANT proposed analytical symmetry as a methodological starting point, suspending prior assumptions about who or what can act (Callon, 1986; Latour, 1987; Law, 1986). Yet, this core contribution of ANT has often been misunderstood. This did not imply that actors are equal in effect, but that their roles and capacities should be examined as outcomes of situated relations rather than assumed in advance. It was never about abandoning distinctions; but about being careful with the distinctions we make that we commonly assume are there. Instead of starting off with pre-existing abilities of any kind, exploring how entities might achieve their abilities should be made part of the study. This position resists prior commitment about how entities exist or relate. Abilities are not part of entities’ nature that can simply be put into play in various encounters with other entities. This shouldn’t be confused with saying that nonhumans and humans are alike or equal in effect. Analytical symmetry should not be interpreted as evenness, but as a methodological suspension of prior assumptions about who or what can act (Lezaun, 2017). What ANT argued for was that all entities (humans and nonhumans) should be approached with ‘analytical symmetry’. No assumptions about entities’ abilities and qualities should be assumed or taken for granted but instead examined in practice. The difficulty is that this commitment is rarely sustained in description. It is not rejected in theory but eroded in practice.

Critical posthumanist critiques of flat ontology sometimes interpret analytical symmetry as ontological equalness (Braidotti, 2019) arguing that it prevents any analysis of power relations. The ANT argument was not that relations in any way are symmetrical. The interpretation of evenness is closely connected to debates about scale and actors in social theory (Strathern, 1992). ANT was a promise to explain society in ‘which the things, facts and artifacts, are major components’ (Callon & Latour, 1992, p. 348). When other theories considered agency as a pre-existing characteristic belonging to a certain entity (commonly human), ANT tried to make the point that agency is a result of action. On a similar note, Barad (2007) talks of intra-action in order to push the notion that entities are not pre-established but change each other in and through the intra-action in which they involve each other.

These observations resonate with long-standing discussions in social theory about how best to describe action, materiality and the ‘social’. Various relational and practice-focused approaches have already shown that familiar distinctions do not always hold when examined closely. Our intention here is to draw attention to some of the analytic moves through which anthropocentric assumptions may be reinstated, even in work that seeks to unsettle them. There is a descriptive tendency to reintroduce asymmetry through the very accounts that claim to treat humans and nonhumans alike. Anthropocentrism is reproduced here through the terms on which symmetry is enacted – and subsequently undone – in description.

Agency: How Humans are Foregrounded

In discussions of agency, similar tendencies appear, where human actions continue to be foregrounded as decisive in how action is accounted for (Delanty & Mota, 2017). What is at work here is a descriptive tendency to acknowledge nonhuman influence while reserving agency for humans. Consider the use of an algorithmic system. A case is processed, data entered, and an output is produced: a recommendation. The output shapes what is attended to, which options appear viable, and how the situation is understood. Certain possibilities are foregrounded, others recede. In the moment of use, the system does not simply present information but actively configures the field of action. What can be seen, considered, and done is reorganised through its operation. And yet, when the situation is described, a different distribution settles in. The system is said to assist, support, or inform, while the human actor is described as deciding, judging, and acting. The output becomes an input to human reasoning; its effects are translated into something that contributes to, but does not itself count as, action. What is notable is not that the system fails to make a difference. Its contribution is central to how the situation unfolds. But this contribution is not stabilised as action. Instead, agency is gathered on the side of the human actor, while the system's role is rendered as a mechanism that shapes the conditions under which action takes place. Here, the distinction does not lie in whether something produces effects, but in how those effects are described. Some are recognised as actions, others as assistance. In this way, agency does not follow from what entities do, but from how their contributions are accounted for.

A further way in which anthropocentrism is reproduced concerns the unwillingness to acknowledge nonhumans as fully fledged actors in their own right. Why is it that we tend to think that the human ultimately is the driver of action? How can the pandemic be understood without accounting for how aerosols and viruses shape and drive action? Can the climate catastrophe be described without recognising that carbon dioxide participates in organising social life? These situations are often described as emblematic of the Anthropocene. What they practically foreground, however, is not a geological epoch as such, but the everyday difficulty of sustaining human-centred accounts of how action unfolds. The question here is whose action comes to matter – who is able to act, and whose action counts. In technoscience, questions about agency are frequently debated, even though scholars have shown both how technology guides and steers human action (Akrich, 1992) and how humans and technology transform and shape each other (Ihde, 2001).

In many descriptions, nonhumans do not enjoy the same privileges as humans do when it comes to the question of agency. The potential agency of nonhumans is often explained away through various analytical moves. While they may be acknowledged as having effects or consequences, these are rarely treated as agency in their own right. The ability of humans to ‘have’ agency, to act and affect others, is seldom brought into question; instead, it is typically assumed. If they’re lucky, nonhumans may sometimes be acknowledged as having effects or consequences. This is seen as something radically different from having agency. Additionally, nonhumans can sometimes be infused with human agency. Human agency is then re-enacted through, embedded in, delegated to, or attributed to, an object. How this magic comes about is often left unspecified. It is the ‘upshot of active work’ (Woolgar & Lezaun, 2013). In much social theoretical reasoning, agency is commonly talked about as just that – a property belonging to humans. These are restrictive logics of world-making. Current analytical practices often manage nonhuman participation in highly restrictive ways. What changes when human and nonhuman (abiotic as well as biotic) action are approached in the same terms is precisely what remains difficult to register.

Following in the wake of ANT scholars, scholarly interest in bringing the nonhuman back has since increased, not the least in the wake of the climate catastrophe. In response to calls for thinking differently about nonhumans, scholars in the social sciences have variously turned to new materialism (Coole & Frost, 2010), object-oriented ontology (Harman, 2018), and Eco-Marxism (Burkett, 2006; Foster, 2000), attempting to advance ideas of ontology. For example, scholars have tried to nuance the discussions of human/nonhuman arguing that different actors have different possibilities to exercise different degrees of agency, wanting to move away from separations of human/nonhuman to animate/inanimate (see, e.g. scholars from the ontological turn in anthropology such as Ingold, 1997; Kohn, 2013). For scholarly debates seeking to rethink nonhumans in the social sciences, tensions arise when entities are approached as having fixed properties or capacities prior to situated relations. While these approaches seek to nuance the understanding of nonhumans and critique dualism, they may nevertheless risk reinforcing nonhumans as a predefined category. In these accounts, nonhumans can appear as if endowed with autonomy and internal agency. This view is apparent in the use of phrases like ‘thing-power’ (Bennett, 2010), ‘things themselves’ (Henare et al., 2006), ‘the power of matter’ (Coole & Frost, 2010), and ‘autonomous objects’ (Harman, 2018). Bennett (2010, p. 10), for example, argues that material is itself ‘self-organising’. These approaches are sometimes positioned as an advance on ANT but, quite against ANT's central critical move, adopt an ontological stance which suggests that entities exist before practice, and that entities act on their own. New materialism places strong emphasis on the ‘power’ of entities, which risks losing sight of relationality. While the attempts to rethink agency beyond the human have taken many forms, these approaches have expanded the range of entities considered capable of acting. And yet, they often do so by attributing agency as a property of entities. In this way, the problem is displaced rather than resolved.

Suggestions that nonhumans ‘do things’, collaborate with humans, and are part of practices are difficult to think, as Despret (2016) suggests. Such situations are often described in only one register. This is beautifully exemplified, for example, in Porcher and Schmitt's (2012) text on cows. At first, when asked about the cows’ work, the farmers argued that only humans ‘work’. The cows’ work was invisible to them. Then, when Porcher and Schmitt started paying attention to how the cows resisted the farmer they could see that the cows understood what was expected of them and that they actively took part, or not, in the work. The attitudes we go in with and our notions of what ‘matters’ affect not only how we ask questions, but also our knowledge production. The assumptions we make about various actors and their abilities shape how entities are rendered and accounted for. By changing the proposition, a different account of who was acting – and how – became possible. As Despret (2004) argues, widening who is allowed to act also shifts what can register as action. If we consider objects, technologies, and ‘things’ as passive and uninteresting, they will indeed become so. Where entity encounters are described as transformative, entities appear as changing one another rather than merely confronting one another as fixed forms. What remains marginal are accounts that register how entities affect and transform one another in practice. Other ways of describing such relations remain largely marginal. The question here is how the unknown is rendered knowable. What remains less examined is how agency is assigned, withdrawn, or redistributed in description.

Agency is not only denied to nonhumans; it is also redistributed in restricted ways. Nonhumans may be granted effects but not recognised as acting. Agency may be delegated from humans to objects or attributed back to them as a derivative of human intention. In other cases, particularly in new materialist and object-oriented accounts, agency is treated as a property of entities themselves. Across these different moves, what remains underexamined is how agency is assigned, withdrawn, or redistributed in description. Anthropocentrism persists not simply through the exclusion of nonhumans, but through the terms on which agency is made to count.