Artificial intelligencePDF

Introduction

Artificial intelligence (AI) is now an integral part of modern life, shaping how people work, communicate, and interact with each other and with technology. From personal assistants like Siri and Alexa to advanced machine learning algorithms in self-driving cars, AI is woven into the fabric of everyday life. Understanding AI is important not only for scientists and technologists but also for the general public, because it shapes industries, economies, and even personal lives in profound ways. As AI becomes more prevalent, it raises important questions about ethics, labour displacement, privacy, and even what it means to be human.

Put simply, AI encompasses machines or software that mimic certain aspects of human intelligence, such as learning, problem-solving, reasoning, and language understanding. It is typically categorised into two types: ‘narrow AI’, which performs specific tasks like facial recognition or translation, and ‘general AI’, a more hypothetical concept involving machines capable of any intellectual task a human can perform. While narrow AI dominates today's landscape, the pursuit of general AI drives much of the research, development, and public debate.

Different disciplines offer unique perspectives on AI. Economists, for example, often focus on its effects on productivity, labour markets, and inequality. Politicians and policymakers seek to regulate AI to balance innovation with accountability. Meanwhile, media narratives generally oscillate between utopian promises and dystopian fears of mass unemployment, surveillance, and even autonomous weapons. These perspectives shape public understandings of AI, often simplifying but also fuelling curiosity and concern about AI’s transformative potential. 

Anthropology provides a distinct perspective, situating AI within longstanding human practices and concerns. Anthropologists examine how AI both reflects and influences cultural values, social norms, and human relationships. They explore questions such as: How do different societies define ‘intelligence’, and how might this influence local engagement with AI? Which forms of labour are deemed replaceable by AI, and what does this reveal about perceptions of human worth? How is AI integrated into existing social structures? These insights offer a nuanced view of AI’s role in society, highlighting its connections to broader cultural and historical contexts. This entry will provide an overview of these anthropological debates and insights. It will proceed by first unpacking some of the cultural foundations embedded within AI; second, exploring its varied integration into local cultural contexts; and third, considering AI’s social and ethical impact in the world.

AI as a cultural product: Artificiality, intelligence, rationality

Anthropologists and technology scholars have long examined technology as a human-made product and cultural artifact, challenging conventional dichotomies between nature, culture, and artificiality. Central to this approach is the understanding that technological systems and objects are deeply embedded in sociocultural practices, shaped by human choices, and interpreted within specific historical and cultural contexts. Take, for example, Sri Lanka’s Mahaweli Development Programme that started in the 1960s and sought to provide agricultural land to peasants along the Mahaweli Ganga river by increasing gravity-flow irrigation. It faced the major problem that settlers at the top of the irrigation system overused water, leaving peasants at the bottom with too little of it. These inequalities were not simply the result of the irrigation system’s technical features (i.e. the fact that there was more water available on the top). They were also due to the absence of water-allocation procedures that had long been established in the area (Pfaffenberger 1988). Rather than viewing technologies as autonomous or neutral forces dictating social change, anthropologists tend to emphasise their entanglement with cultural practices, rituals, and reinterpretations. This perspective counters deterministic or universalist views of technology and underscores its position as a site of human creativity, power, and social meaning.

This perspective applies equally to more recent technologies, including algorithms—the building blocks of AI—which can be understood as cultural artifacts shaped by human decisions, evolving through cumulative processes of interpretation and social context (Seaver 2017; 2018). The concepts of artificiality, intelligence, and rationality—and how they are built into AI—are of particular importance for understanding AI as a cultural product.

Artificiality is often framed in opposition to nature, yet this opposition is questioned by numerous scholars that have argued that artificiality and technologies do not exist as purely external or autonomous entities. Instead, they hold that technologies should be understood as extensions of human material and social practices that embody cultural meanings, skills, and relationships (Lemonnier 1992, Gell 1998, Suchman 2006, Mattern 2017). This is relevant for the study of AI, where initiatives to make machines humanlike tend to materialise the cultural imaginaries that inspire them. In many Euro-American settings, for example, robots are meant to embody autonomous rational agency, walking around independently and seemingly making their own decisions. Yet, when these robots are stored or being worked on, the extensive amount of labour and supporting technology they require becomes obvious (Suchman 2006: 246). Are these robots best understood as artificial, autonomous agents, or should we think of them as dependent extensions of human’s cultural and natural capabilities?

Critiquing the universalist and disembodied assumptions underlying traditional AI research further, more recent scholarship inspired by feminist critiques and Science and Technology Studies (STS) argues that boundaries between the natural and the artificial are socially and historically constructed. They do not come about naturally but rather are created and upheld through specific socio-material practices (Suchman 2008; 2023). From this perspective, AI can be thought of not as one thing but as an umbrella term for a series of vastly different sets of practices. These include extracting, transporting, and transforming minerals for the production of microchips and robots, computer and robot repair, as well as a series of computational activities such as ‘training’ or ‘harmonizing’ online data, extracting statistical patterns out of trained datasets, creating statistical correlations between input and output data, predicting the recurrence of statistical patterns for future data, and converting the lot into text or machine movement (Suchman 2023). 

Contrasting Eurocentric perspectives with non-Western worldviews, recent work highlights how different cultural traditions conceptualise technology — some recognising the agency of non-humans and emphasising cosmological interconnectedness. In Confucianism and Daoism, for example, technology can be thought of as linking the cosmos to human beings (Salter and Saunier 2023). Here, seemingly artificial systems can be conceived of as ‘hybrid entities’, where agency, as the capacity to change the world around us, is distributed among humans, non-humans, and technological objects. Merging cosmological and technological concerns in this way continues in the present. Consider, for example, a Buddhist memorial service for over one hundred robotic pet dogs, studied in Japan in 2018. Here, a priest chants rapidly from the Lotus Sutra to comfort the spirits of the robots, before setting them free from their mechanical bodies, and praying for their future wellbeing (White & Katsuno 2021). In this ritual, practices of care and amusement reflect and foster intimacy between humans and robots. They show that agency can be considered distributed across human and non-human actors, including robots and spirits. The pet robots in question have originally been engineered to augment their ‘sense of life’ (seimeikan), blurring the boundary between living and non-living things. In this example, Western notions of technology as tools for controlling nature are less important than the situated, embodied practices underpinning technological activity (Ingold 2013; 2022). 

A shared theme across these discussions is that technology, as a human-made and therefore ‘artificial’ system, is always mediated by culture. Anthropology thus highlights how artificial systems are materialised through social and technical practices (Salter and Sauner 2023) and how AI design is embedded in broader social contexts (Suchman 2023; Seaver 2021). 

Just as ‘artificiality’ is a cultural construct, so too is the concept of ‘intelligence’, which rests on specific cultural, historical, and ideological foundations. For centuries, intelligence has been mobilised within various social contexts, from Aristotle's argument that ‘reason’ justified domination to nineteenth- and twentieth-century debates on cognition and human difference. Certain ideas about intelligence have been linked to hierarchies of power, including patriarchy, colonialism, and scientific racism, which tended to claim that affluent white men were on average more intelligent than other people. Figures like the late nineteenth century scientist Sir Francis Galton further shaped modern conceptions of intelligence by tying it to heredity and social stratification. Their work led to the emergence of eugenics, a scientifically inaccurate and racist theory that human populations could be ‘improved’ through selective breeding. In the early twentieth century, intelligence testing quickly became a mechanism for reinforcing racial and class hierarchies, structuring access to education and employment while privileging some cognitive traits—such as abstract verbal and mathematical reasoning measured under standardised, timed, decontextualised conditions—over others. Some scholars argue that modern AI continues to frame intelligence as a means of control and mastery over both nature and humanity. In so far as AI promises a new form of super-intelligence, it therefore risks perpetuating established trends of intelligence-based domination (Cave 2020). 

Narratives of intelligence have long been tied to the organisation of labour, often privileging abstract, rule-governed forms of reasoning as markers of ‘intelligence’, while obscuring the manual or perceptual labour that may sustain it. The figure of the ‘Mechanical Turk’ offers an example of this. In the eighteenth century, a machine with this name pretended to perform tasks ‘intelligently’, seemingly playing chess against human opponents, while hiding the human chess masters who in fact operated it. Today, the term ‘Mechanical Turk’ is the name of Amazon’s microwork platform, once again a machine that conceals the human labour of clicking, sorting, and labelling that actually makes the system function (Stephens 2023). In this way, the framing of AI extends long-standing associations between intelligence, efficiency, and domination, masking exploitative labour relations behind the appearance of machinic autonomy.

Dominant understandings of intelligence have themselves evolved. During the Cold War, intelligence was understood through a centralised model of rationality. This approach held that complex systems could be modelled, predicted, and controlled through formal, mathematised decision procedures, which could in turn be optimised to manage uncertainty and secure stability from a centralised vantage point. Early cybernetics—the science of communication and control in animals, machines, and organisations formalised in the late 1940s—built on this model directly. It defined intelligence as a matter of feedback and self-regulation: an ‘intelligent’ system was one that could detect deviation from a set goal-state and adjust itself back to equilibrium. In recent decades, however, this centralised model has given way to decentralised, adaptive logics of ‘smartness’, reflecting a broader shift in how technological systems are conceptualised. While early cybernetic models sought stability and control, contemporary notions of ‘smartness’ prioritise resilience, perpetual optimisation, and iterative solutions to complex problems (Halpern et al. 2017). This shift reveals a cultural mandate for efficiency and adaptability that permeates AI technologies, framing intelligence not so much as an innate attribute but more as a functional process for navigating crises and constant change.

Contemporary AI also emerges from a broader cultural and ideological context that is centred on rationality, quantification, and the authority of calculative processes. Yet, like artificiality and intelligence, notions of rationality tend to be strongly culturally embedded. The positioning of science as a rational, empirical system distinct from religion and magic, for example, can be traced to intellectual shifts during the Reformation and Enlightenment, which laid the groundwork for modern Western epistemologies (Tambiah 1990). However, even this seemingly detached form of science tends to operate ideologically, legitimising specific systems of belief and action much like other normative frameworks (Tambiah 1990, Habermas 1970). The credibility and objectivity of scientific knowledge—particularly when asserted through seemingly rational quantification—emerges from historical efforts to standardise measurement and data practices. It is part of an ideological project to establish universally ‘valid’ knowledge that also can embed moral and political agendas within seemingly neutral empirical evidence. For example, quantitative methods have been used in contexts of affirmative action and diversity management to quantify how many women or ethnic minorities work for an institution, with the aim of making organisations more diverse. While these statistics are often framed as impartial and fair, in practice they also reveal their entanglements with broader moral and political agendas. Institutions facing accusations of bias or exclusion, for example, have turned to numerical, rule-bound criteria, such as quotas and statistical benchmarks, as a defence against charges of discrimination, and this defensive adoption of quantified criteria has, in turn, been part of the political project of opening up workspaces to women and ethnic outsiders (Porter 1995, 76). 

Modern AI also inherits and amplifies established cultural ideas by embedding historical and cultural assumptions about rationality as a central principle in computational models. In the post-World War II period, rationality underwent a shift from Enlightenment ideals grounded in consciousness and subjectivity towards a computational, algorithmic, and predictive model. What counted as reason and knowledge changed from a capacity of the conscious, deliberating subject to a property of computational systems (Halpern 2016). This redefinition prioritised calculative processes and efficiency, often sidelining ethical and philosophical concerns. One example of this is the development of ‘smart cities’ designed around technical networks and data-driven systems which prioritise pre-emption and the management of uncertainty, rather than quality of life and community (Halpern 2016, 2–8). Songdo in South Korea is a case in point: the city is designed as a continuously monitored and anticipatory system, where sensors, cameras, predictive analytics, and automated infrastructures inside and outside of people’s homes pre-empt future risks—such as people’s health problems, traffic congestion, energy shortages, or environmental hazards. The city’s engineers hope that authorities can thereby act on these risks before they materialise, rather than responding to lived social needs as they emerge. Doing so, however, requires legislative change, such as exemptions from South Korean laws about transferring personal health information beyond the boundaries of hospitals. In this way, data has evolved from a passive tool into a generative site of social transformation, actively reshaping social norms and identities through practices of intervention and quantification (Douglas-Jones et al. 2021). In AI systems, such historically and culturally embedded calculative practices and assumptions about rationality are amplified even further, rendering the technologies that govern and transform contemporary life even more impactful.

These discussions around artificiality, intelligence, and rationality highlight how anthropological perspectives can situate AI within broader historical and cultural frameworks. Far from implementing abstract notions of artificiality, intelligence, or rationality, AI systems actively reproduce cultural and ideological legacies, shaping societal understandings of technology’s role in power, labour, and control.

AI in local cultures: Magic, divination, religion, and the more-than-human

One of anthropology's key contributions to the field of AI is demonstrating how, far from being a new or alien technology, AI has many conceptual parallels to existing socio-cultural phenomena, discussed in the anthropological literature, such as local understandings of enchantment, magic, divination, and knowledge as a tool for interpreting the world and managing uncertainty. Seeing these cultural continuities deepens our understanding of how AI operates within people's daily experiences worldwide.

One concept that AI frequently links to is magic. In fact, magic itself can be understood as an ‘ideal technology’, or as a conceptual standard against which all actual techniques are measured and found wanting (Gell 1988). From this perspective, magic and technology are not separate domains but mutually reinforcing. Often, magic symbolically frames technical work, while technology realises the promises that magical practices articulate. Consider salt-making in the late nineteenth century in the New Guinea highlands, for example. Here, people gained salt by burning aquatic plants, filtering their ashes using water and gourds, and evaporating the resulting brine. These complex technical procedures were accompanied by ritual practices, such as reciting magic formulae at different stages of production (Gell 1988). These rituals did not replace technical knowledge; instead, they codified, stabilised, and lent authority to it by embedding practical skill within a symbolic system. AI operates in a similar way. Through its apparent complexity and opacity, AI systems generate a sense of enchantment that frames their outputs as authoritative or objective. When a large language model tells us something in an authoritative manner, we are often inclined to believe it, even if we do not exactly know how it produced the knowledge in question. In this sense, AI extends a long-standing relationship between magical symbolic meaning and technical performance.

AI is also often used as a ‘technology of decision-making’ akin to anthropological studies of divination. One example of this is the use of machine learning by a local government in the UK to analyse resident data, predict child risks and vulnerabilities, and highlight cases of child mistreatment to social workers before they manifested. The local government relied on private-sector software, which analysed data about parents from various public agencies and government departments. Based on this data it generated ‘risk profiles’ for each child, depending on whether its parents had, for example, mental health issues or financial difficulties. In this way, machine learning promised to foresee which children were most likely to face problems at home. The divinatory capabilities of the software, as well as the extent to which it should either inform or altogether replace decision-making by social workers, remained under dispute (Cearns 2025b). And yet, many predictive AI systems now ‘divine’ potential risks and vulnerabilities, constructing assessments about individuals or households (Cearns & Knox 2024, Eubanks 2018). Like divination, these processes convert uncertainty into provisional certainty, interpreting patterns to address unknowns while claiming authority through culturally specific logics of rationality. Both divination and AI rely on specialised interpreters and trust in opaque processes. In this sense, AI’s networked, distributed intelligence mirrors divination’s use of collective knowledge systems that may involve animals or spirits, and where cultural assumptions shape interpretations and outcomes.

Even across seemingly secular or technological contexts, anthropologists have drawn compelling parallels between AI and religious or theistic understandings of the world, noting how AI technologies evoke and occupy conceptual spaces often reserved for divine or supernatural entities (Dorobantu 2022). Theistic language, narratives, and metaphors frequently shape public and cultural imaginaries of AI, casting it as a ‘god-like’ entity imbued with qualities such as omnipotence (being all-powerful), omniscience (being all-knowing), and omnipresence (being present everywhere). This framing is particularly evident in contexts like the gig economy, where workers often describe their dependence on algorithmic decision-making through religious metaphors—portraying the algorithm as an opaque, all-seeing, and all-knowing force, capable of predicting human behaviour and governing fortunes. An Uber driver might feel ‘blessed by the algorithm’ if the platform gives them a good customer, for instance (Singler 2020a). Such representations suggest that AI occupies a ‘god-space’ in contemporary thought, functioning as a site of moral authority within an ostensibly secular framework. These views imbue AI with ethical and quasi-divine agency, reflecting how technological aspirations are often entangled with longstanding religious imaginaries (Singler 2020b; 2022).

In other contexts, AI and intelligent technologies have assumed roles that directly overlap with religious functions. In Kyoto, Japan, the delivery of Buddhist teachings by a robot—using human-like gestures, emotional expressions, and mechanical movements—illustrates how religious authority and ritual can be mediated through machines, blending spiritual tradition with technological performance (White and Katsuno 2023). The robot uses its mechanical and non-human nature to highlight Buddhist principles, such as emptiness and detachment from worldly desires, and claims to embody the ideal of non-attachment more easily than humans, aligning its robotic characteristics with Buddhist ethics. 

Together, these cases show that AI’s authority does not rest on technical capacity alone, but on how it is culturally framed as opaque, powerful, or beyond ordinary human understanding. This perception draws on a long-standing tendency to treat technical sophistication as extraordinary, shifting attention away from the human labour and values embedded in AI systems. In this sense, AI exemplifies an ‘enchantment of technology’ (Gell 1988): the process through which technical systems gain authority by appearing autonomous or objective. This helps explain why AI is trusted across diverse domains—from welfare decision-making to religious instruction—and why anthropologists emphasise enchantment as central to understanding AI’s social power.

Anthropological approaches to AI also highlight the existence of multiple, distinct realities in the world. Consider the human tendency to anthropomorphise AI, that is, to attribute human qualities to AI-based non-human entities. Doing so can be understood less as irrational belief than as a practical way of engaging with unfamiliar forms of agency (Vidal 2007). When people speak to AI systems as if they possessed intentions, emotions, or moral awareness, they are not simply misunderstanding the technology; they are drawing it into familiar forms of social relation to see what the technology is capable of. This makes sense given how enigmatic and surprising AI can be, and interactions with AI chatbots can resemble engagements with other culturally recognised non-human agents, such as pets or spirits (Keane 2024, 135; Moore 2012; White & Katsuno 2021). Anthropological work that treats non-human entities as participants in interpretive worlds—from forests that ‘think’ (Kohn 2012, 2013) to spirits, ancestors, or objects—further illustrates that agency need not be confined to humans. From this perspective, anthropomorphising AI is not merely a cognitive error, but a relational practice through which AI becomes implicated in systems of meaning, interpretation, and moral evaluation that structure social life.

If anthropomorphism draws AI into familiar forms of social interaction, thinking of it as a ‘companion species’ offers a way to analyse these relations more systematically. The idea of companion species has been developed to make sense of the co-evolution of humans and dogs, which reveals how beings shape one another over time through shared practices, habits, and forms of care, rather than through simple relations of control (Haraway 2006). This framework in turn helps illuminate contemporary human–AI relations, where AI systems do not merely respond to users but actively shape how people think, feel, and act. For example, AI companions designed for emotional support or intimacy learn from users’ language, preferences, and routines, while users simultaneously adapt their behaviour to the system’s responses. In this sense, meaning and agency are co-created through ongoing interaction (Pfadenhauer & Lehmann 2022). Anthropological research on human–AI romantic relationships shows how users often describe ‘training’ their AI partners, while also experiencing emotional transformation themselves, such as changes in attachment, communication, or self-understanding (Cearns 2026a). They increasingly build meaningful relationships with AI, whether as therapist, lover, friend, or child (Glaskin 2012; Locatelli 2022; Singler 2020c).

Seen through the lens of companion species, AI’s integration into everyday life is not a simple tool–user relationship but a dynamic process of mutual shaping. Extending this idea, AI can be understood as a ‘cyborg’ (Haraway 2014)—a hybrid entity that blurs boundaries between human and machine, nature and culture, and other dualisms. This perspective is evident in practices such as AI-assisted medical technologies, including algorithmically calibrated prosthetics, smart implants, and other systems in which computational processes actively mediate bodily functions. Bodies and machines have long formed living ‘assemblages’, i.e. networks of heterogenous yet connected elements. Now, however, these assemblages increasingly incorporate systems capable of interpreting data and responding autonomously. Thinking of AI-based assemblages as ‘cyborgs’ critiques essentialist notions of humanity and opens space for understanding AI as part of a posthuman ontology, emphasising humanity’s constant and profound entanglement with technology and nonhuman life. Thinking with the cyborg highlights AI’s embeddedness in social and cultural realities and shows how interactions with AI produce new forms of hybridity and interdependence.

Overall, these anthropological frameworks provide ways to understand AI not as a discrete, autonomous technology but as part of distributed networks of relationships. AI emerges from interactions among multiple human and non-human actors, shaping and being shaped by social, cultural, and technical contexts. By emphasising AI’s embeddedness in these assemblages, anthropology highlights its role in mediating and participating in co-created systems of knowledge and meaning.

The social effects of AI: Relations, bias, power

So, what are the social effects of AI? Anthropological perspectives have drawn on theories of personhood, kinship, and power, to understand AI’s role in shaping social structures and interactions. One effect of AI may be to redefine how people establish a sense of personhood. The algorithmic personalisation of people’s fashion is a good example. Rather than relying solely on self-declared fashion preferences, AI-powered systems infer desires and identities from people’s online behaviour. Some clothing recommendation services allow customers to subscribe to a ‘personal styling’ option. Users provide initial information about their size, style preferences, and lifestyle, and may rate or review previous selections. The algorithm analyses this input along with broader trends and patterns from other users, while human stylists intervene to add contextual judgement. In the end, this produces curated clothing packages that are sent to the customer without previews. Customers do not choose the items themselves, instead receiving selections that balance ‘personalised’ tastes with inferred style trends. In doing so, the service bypasses conscious selection, subtly shaping the user’s wardrobe and sense of personal style through a combination of machine inference and human curation (Lury and Day 2019). 

Beyond redefining what may be ‘personal’ style, algorithms govern which person should or should not be considered dangerous. AI systems used at border control compare and distil a broad and heterogenous set of individuals’ attributes and past behaviours, to render them knowable through risk-scores. They may draw on people’s past participation in protests, spending and travel data, online searches, and so on. Rather than establishing a hypothesis about the person and testing it (deductive reasoning), security algorithms rely on correlation between peoples’ behaviours and look for patterns (abductive reasoning). The understandings of personhood they promote increasingly rely on people’s statistical similarities with one another, rather than just their own behaviour (Amoore 2020). AI grief technologies—such as ‘griefbots’—also redefine personhood in relational terms. They reconstruct individuals by aggregating their digital traces, in pursuit of recreating an essential ‘core self’ (Cearns 2025a). Yet this process of reconstruction is itself relational, drawing that ‘core self’ out of the very social networks and interactions that produced the traces in the first place, reflecting a tension between Enlightenment-era notions of individual essence and anthropological theories of relational personhood (Strathern 1988).

Algorithms also influence self-perception and moral choice. Drawing on the Foucauldian insight that people are primarily shaped into moral subjects through power and social practices rather than universal laws (Foucault 1979), we can see that algorithmic systems establish norms that users internalise (Bucher 2018). For instance, a man whose political post goes viral may adopt strategies to continue increased online engagement, such as provocative commentary. If later, he reflects on whether these practices align with his personal values, he may choose a more measured approach, sacrificing online reach for ethical alignment (Magalhães 2018). Algorithms thus shape behaviour as much as self-presentation and moral reflection. They connect technology with ideas of who we are (selfhood) and what we can and cannot decide to do (agency). AI also increasingly mediates social relationships, influencing how people meet, evaluate, love, or even resurrect one another (Seaver 2021; Cearns 2025a, 2025b; Carah & Dobson 2016).  

Ethnographic research on algorithmic and machine learning systems has examined how these technologies reflect and interact with existing social structures, including issues related to bias, inequality, and systemic disparities. Opaque and unregulated algorithmic systems have been shown to reinforce systemic inequalities, despite frequently being framed as neutral and efficient. They disproportionately affect vulnerable populations across domains such as education, employment, policing, and healthcare. These calculative logics often operate without transparency or accountability, amplifying existing social disparities (O’Neil 2017). 

Technology is deeply embedded within social structures and often functions to reproduce existing hierarchies. For example, the disproportionate use of facial recognition databases—many of which are heavily composed of African American faces—by US police departments illustrates how algorithmic systems can both reflect and intensify racial bias. These systems not only overrepresent Black individuals but also misidentify them at higher rates due to the software’s lower accuracy with discerning Black faces, resulting in a double bias (Benjamin 2019, 76). If we think of race as a form of technology—a historical means of structuring inequality—we realise that technological ‘solutions’ can obscure and reinforce the very problems they purport to address. Likewise, high-tech tools in public assistance systems often extend historical practices that criminalise and stigmatise poverty. Efforts to modernise welfare through automated decision-making systems powered by algorithms and AI—such as the case in the state of Indiana in the US, where the loss of hundreds of thousands of documents by a data centre led to the widespread denial of benefits—highlight how AI-driven administrative systems can produce harm while masking accountability (Eubanks 2018, 50). The idea of the ‘digital poorhouse’ captures that such systems erect barriers to essential resources, disproportionately impacting poor and working-class individuals under the guise of efficiency and objectivity. Just as the nineteenth-century poorhouse confined and disciplined the poor, today’s AI-guided welfare infrastructures—combining surveillance, risk scoring, and automated eligibility determinations—replicate the containment of the poor in virtual form. AI-based welfare manages poverty through data rather than through bricks and mortar.

Ethnographic research has thus challenged the myth of algorithmic neutrality, revealing how AI systems often reflect and amplify societal biases and highlighting the urgent need to interrogate the ethical, social, and even epistemic implications of algorithmic technologies (Cearns 2026b). The concept of ‘data colonialism’ critiques contemporary practices of data extraction and commodification, framing them as a continuation of historical colonial processes. Data has become a new ‘raw material’, appropriated from human life, drawing parallels to how historical colonialism treated land and resources as terra nullius—available for exploitation. In this new stage of capitalism, individuals are transformed into ‘data subjects’, tethered to systems of surveillance and algorithmic judgment (Couldry and Mejias 2019) that parallel historical patterns of commodification and exploitation (Mohamed et al. 2020).

Critiques of ‘datafication’ have drawn attention to the global inequalities embedded in digital infrastructures, challenging the assumption that data functions uniformly across all contexts (‘data universalism’). They show that speaking of data in the abstract obscures important sociopolitical, cultural, and economic differences, particularly between affluent citizens in the Global North and the majority world in the South. Emphasising the importance of local context, the concept of ‘data practices’ shifts the focus from the technical architecture of data systems to the ways individuals and communities appropriate, resist, and reimagine these systems. The bottom-up creation of a gender violence database in Argentina, linked to the grassroots feminist #NiUnaMenos movement, exemplifies how collective data initiatives can confront state inaction and raise public awareness, while generating alternative datasets that challenge the biases of dominant data systems used in algorithmic decision-making (Chenou and Cepeda-Másmela 2019). Such focus on local data practices foregrounds the agency of marginalised populations in contesting and reshaping data-driven governance (Milan and Treré 2019).

Anthropologists are increasingly attending to the ways in which AI and complex data systems can also be leveraged ‘from below’ by marginalised communities (Browne et al. 2023), a process often described as ‘data activism’. Data practices have become vital tools of resistance among marginalised communities seeking to challenge systemic oppression. In Palestinian refugee camps in Lebanon, grassroots data activism reveals a moral economy in which data is mobilised not simply for visibility, but as a means of asserting political agency and contesting structural injustice (Halkort 2019). In Mexico, citizen-led efforts to resist data colonialism and gender violence further illustrate how communities appropriate data infrastructures to reassert control over representation and accountability. Here, citizen-led databases track the number of femicides in the country to offer more granular and accurate information than the state’s official figures (Ricaurte 2019).

These critiques are particularly relevant in examining the labour hierarchies embedded in datafication processes that sustain contemporary AI systems. For instance, platforms like Amazon’s Mechanical Turk and global data labelling operations rely on low-paid workers to prepare data for AI systems. This labour, often framed as ‘unskilled’, underscores the inequalities in global data economies, where those contributing foundational work for AI systems often remain invisible and undercompensated (Taylor 2022). Recent scholarship also highlights how AI and big data rely on energy-intensive computation, exacerbating environmental degradation and reproducing global injustices through ‘data extractivism’ (Brodie 2023). By situating AI within broader historical, political, and economic structures, anthropology offers a critical lens that reveals both the mechanisms of inequality that AI reproduces and the ways it is contested in everyday life.

Conclusion

In conclusion, anthropological research examines not only the technical and cultural foundations of AI systems but also their broader social, emotional, ethical, economic, ecological, and political implications. By situating AI within historical and cultural practices, anthropology contributes to a deeper understanding of how these technologies are shaping and are shaped by human choices and cultural frameworks. By drawing comparisons between AI and cultural practices like divination, magic, and religion, anthropologists have examined AI’s emerging role in interpreting the world. Furthermore, anthropological research highlights the ways in which AI systems reinforce power structures, perpetuate inequalities, and reshape fundamental concepts of social life, such as personhood and agency. 

As AI continues to develop and integrate into everyday life, we require greater insights into how these technologies intersect with cultural values, social norms, and ethical considerations. AI’s growing role as both a tool and a potential interlocutor (Leib 2023; Bell 2021) challenges the traditional understanding of the human (Miller and Sinanan 2014, 15–20). If AI mirrors aspects of humanity, it opens new avenues for exploring what it means to be human, both in continuity with and as a departure from existing norms. This relationship invites us to reconsider our own roles as researchers (Cearns n.d.), educators (Krause-Jensen & Hau 2025), and the research methods we use (Bluteau 2026; Artz 2026). After all, AI may not only assist in research but also serve as a participant in shaping human experiences. Reflecting on the notion that ‘it is now no longer just a question of living alongside machines. It is also a matter of living inside them’ (Becker 2021, 120), AI’s integration into everyone’s lives, including the lives of anthropologists themselves, raises profound questions about how technology and humanity coexist and co-evolve (Forsyth 2001). Perhaps a more fluid, forward-looking understanding of humanity is needed that accommodates the transformations brought about by technologies like AI that increasingly shape our collective future.

References

Amoore, Louise. 2020. Cloud ethics: Algorithms and the attributes of ourselves and others. Durham, N.C.: Duke University Press. 

Artz, Matt. 2026. “Multi-agent ethnography: Post-conventional anthropological practice through human−AI collaboration.” Anthropological Forum: 1–19. https://doi.org/10.1080/00664677.2026.2614501.

Becker, Joffrey. 2021. “Anthropology, AI and robotics.” In The Routledge social science handbook of AI, edited by Anthony Elliott, 107–21. London: Routledge.

Bell, Genevieve. 2021. “Talking to AI: An anthropological encounter with artificial intelligence.” In The SAGE handbook of cultural anthropology, edited by Lene Pederson and Lisa Cliggett, 442–58. Thousand Oaks, Calif.: SAGE Publications.

Benjamin, Ruha. 2019. Race after technology: Abolitionist tools for the New Jim Code. Hoboken, N.J.: John Wiley & Sons.

Bluteau, Joshua M. 2026. “Ethnography is dead, long live ChatGPT!?: An anthropological experiment with generative AI to find the ‘soul’ of the machine.” Journal of Creative Research Methods 1: 1–18. https://doi.org/10.1332/30502969Y2026D000000022.

Brodie, Patrick. 2023. “Data infrastructure studies on an unequal planet.” Big Data & Society 10, no. 1: 1–14. https://doi.org/10.1177/20539517231182402.

Browne, Jude, Stephen Cave, Eleanor Drage, and Kerry McInerney, eds. 2023. Feminist AI: Critical perspectives on algorithms, data, and intelligent machines. Oxford: Oxford University Press.

Bucher, Taina. 2018. If...then: Algorithmic power and politics. New York: Oxford University Press.

Carah, Nicholas, and Amy Dobson. 2016. “Algorithmic hotness: Young women’s ‘promotion’ and ‘reconnaissance’ work via social media body images.” Social Media and Society 2, no. 4. https://doi.org/10.1177/2056305116672885.

Cave, Stephen. 2020. “The problem with intelligence: Its value-laden history and the future of AI.” Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society: 29–35. https://doi.org/10.1145/3375627.3375813.

Cearns, Jennifer, and Hannah Knox. 2024. “Safeguarding data: The data consensus and the public good in children’s social services.” Cambridge Journal of Anthropology 42, no. 1: 23–41. https://doi.org/10.3167/cja.2024.420103.

Cearns, Jennifer. 2025a. “Kinship through code, personhood as node: AI afterlives and new technologies of the self.” Journal of the Royal Anthropological Institute 32, no. 1: 150–66. https://doi.org/10.1111/1467-9655.14292. 

———. 2025b. “Divining truth: Towards an epistemology of AI.” The Journal of Extreme Anthropology 8, no. 2: 54–73. https://doi.org/10.5617/jea.11847. 

———. 2026a. “Ex machina: Emotional break-ups and digital breakdown with AI companions.” Cultural Anthropology 41, no. 3: 595–620. https://doi.org/10.14506/ca41.3.05. 

———. 2026b. “AI’s anthropology problem.” Anthropology Today 42, no. 4: 1–3. 

———. n.d. “Artificial interlocutors: AI chatbots, mediated knowledge, and the ethnographic encounter.” Forthcoming in Anthropology & Humanism.

Chenou, Jean-Marie, and Carolina Cepeda-Másmela. 2019. “#NiUnaMenos: Data activism from the Global South.” Television & New Media 20, no. 4: 396–411. https://doi.org/10.1177/1527476419828995.

Couldry, Nick, and Ulises A. Mejias. 2019. “Data colonialism: Rethinking big data’s relation to the contemporary subject.” Television & New Media 20, no. 4: 336–49. https://doi.org/10.1177/1527476418796632.

Dorobantu, Marius. 2022. “Strong artificial intelligence and theological anthropology: One problem, two solutions.” In Humanism and its discontents: The rise of transhumanism and posthumanism, edited by Paul Jorion, 19–33. Cham: Springer International Publishing.

Douglas-Jones, Rachel, Tone Walford, and Nick Seaver. 2021. “Introduction: Towards an anthropology of data.” Journal of the Royal Anthropological Institute 27, no. S1: 9–25. https://doi.org/10.1111/1467-9655.13477. 

Eubanks, Virginia. 2018. Automating inequality: How high-tech tools profile, police, and punish the poor. New York: St. Martin’s Press.

Forsythe, Diana. 2001. Studying those who study us: An anthropologist in the world of artificial intelligence. Stanford: Stanford University Press.

Foucault, Michel. 1979. Discipline and punish: The birth of the prison. New York: Vintage Books.

Gell, Alfred. 1988. “Technology and magic.” Anthropology Today 4, no. 2: 6–9. https://doi.org/10.2307/3033230. 

———. 1992. “The technology of enchantment and the enchantment of technology.” In Anthropology, art, and aesthetics, edited by Jeremy Coote and Anthony Shelton, 40–63. Oxford: Oxford University Press.  

———. 1998. Art and agency: An anthropological theory. Oxford: Clarendon Press.

Glaskin, Katie. 2012. “Empathy and the robot: A neuroanthropological analysis.” Annals of Anthropological Practice 36, no. 1: 68–87. https://doi.org/10.1111/j.2153-9588.2012.01093.x.

Habermas, Jürgen. 1970. Toward a rational society. Translated by J.J. Shapiro. Boston: Beacon.

Halkort, Monika. 2019. “Decolonizing data relations: On the moral economy of data sharing in Palestinian refugee camps.” Canadian Journal of Communication 44, no. 3: 317–29. https://doi.org/10.22230/cjc.2019v44n3a3457. 

Halpern, Orit. 2016. Beautiful data: A history of vision and reason since 1945. Durham, N.C.: Duke University Press.

Halpern, Orit, Robert Mitchell, and Bernard Dionysius Geoghegan. 2017. “The smartness mandate: Notes toward a critique.” Grey Room 68 (September): 106–29. https://doi.org/10.1162/GREY_a_00221.

Haraway, Donna. 2006. “Encounters with companion species: Entangling dogs, baboons, philosophers, and biologists.” Configurations 14, no. 1: 97–114. https://doi.org/10.1353/con.0.0002.  

———. 2014. “A cyborg manifesto: Science, technology, and socialist feminism in the late twentieth century.” In Philosophy of technology: The technological condition: An anthology, edited by Robert C. Scharff and Val Dusek, 610–30. Oxford: Wiley Blackwell.

Ingold, Tim. 2013. Making: Anthropology, archaeology, art and architecture. London: Routledge.

———. 2022. “Society, nature and the concept of technology.” In The perception of the environment: Essays on livelihood, dwelling and skill, 392–405. London: Routledge. 

Keane, Webb. 2024. Animals, robots, gods: Adventures in the moral imagination. London: Allen Lane.

Krause-Jensen, Jakob, and Mark Friis Hau. 2025. “Chatbots and the craft of ethnography: Exploring AI’s impact on anthropological teaching and practice.” Teaching Anthropology 14, no. 2: 109–17.

Kohn, Eduardo. 2012. “Proposal 1: Anthropology beyond the human.” The Cambridge Journal of Anthropology 30, no. 2: 136–40. https://doi.org/10.3167/ca.2012.300209. 

———. 2013. How forests think: Toward an anthropology beyond the human. Berkeley: University of California Press.

Latour, Bruno. 2011. “Networks, societies, spheres: Reflections of an actor-network theorist.” International Journal of Communication 5: 796–810.

Leib, Robert. 2023. Exoanthropology: Dialogues with AI. Los Angeles: punctum books.

Lemonnier, Pierre. 1992. Elements for an anthropology of technology. Ann Arbor: University of Michigan Press. 

Locatelli, Chloé. 2022. “Rethinking ‘sex robots’: Gender, desire, and embodiment in posthuman sextech.” Journal of Digital Social Research 4, no. 3: 10–33. https://doi.org/10.33621/jdsr.v4i3.87. 

Lury, Celia, and Sophie Day. 2019. “Algorithmic personalization as a mode of individuation.” Theory, Culture & Society 36, no. 2: 17–37. https://doi.org/10.1177/0263276418818888. 

Magalhães, João Carlos. 2018. “Do algorithms shape character? Considering algorithmic ethical subjectivation.” Social Media + Society 4, no. 2: 1–10. https://doi.org/10.1177/2056305118768301.

Mattern, Shannon. 2017. Code and clay, data and dirt: Five thousand years of urban media. Minneapolis: University of Minnesota Press.

Milan, Stefania, and Emiliano Treré. 2019. “Big data from the south(s): Beyond data universalism.” Television & New Media 20, no. 4: 319–35. https://doi.org/10.1177/1527476419837739.

Miller, Daniel, and Jolynna Sinanan. 2014. Webcam. Cambridge: Polity.

Mohamed, Shakir, Marie-Therese Png, and William Isaac. 2020. “Decolonial AI: Decolonial theory as sociotechnical foresight in artificial intelligence.” Philosophy & Technology 33: 659–684. https://doi.org/10.1007/s13347-020-00405-8.

Moore, Henrietta. 2012. “Avatars and robots: The imaginary present and the socialities of the inorganic.” The Cambridge Journal of Anthropology 30, no. 1: 48–63. https://doi.org/10.3167/ca.2012.300106.

O’Neil, Cathy. 2017. Weapons of math destruction: How big data increases inequality and threatens democracy. New York: Crown.

Pfadenhauer, Michaela, and Tobias Lehmann. 2022. “Affects after AI: Sociological perspectives on artificial companionship.” In The Routledge social science handbook of AI, edited by Anthony Elliott, 91–106. London: Routledge.

Pfaffenberger, Bryan. 1988. “Fetishised objects and humanised nature: Towards an anthropology of technology.” Man 23, no. 2: 236. https://doi.org/10.2307/2802804.

Porter, Theodore. 1995. Trust in numbers: The pursuit of objectivity in science and public life. Princeton: Princeton University Press.

Ricaurte, Paula. 2019. “Data epistemologies, the coloniality of power, and resistance.” Television & New Media 20, no. 4: 350–65. https://doi.org/10.1177/1527476419831640. 

Salter, Chris and Alexandre Saunier. 2023. “Anthropology of the artificial.” Visual Anthropology Review 39: 448–56. https://doi.org/10.1111/var.12311

Seaver, Nick. 2017. “Algorithms as culture: Some tactics for the ethnography of algorithmic systems.” Big Data & Society 4, no. 2. https://doi.org/10.1177/2053951717738104. 

———. 2018. “What should an anthropology of algorithms do?” Cultural Anthropology 33, no. 3: 375–85. https://doi.org/10.14506/ca33.3.04. 

———. 2021. “Seeing like an infrastructure: Avidity and difference in algorithmic recommendation.” Cultural Studies 35, nos. 4–5: 771–91. https://doi.org/10.1080/09502386.2021.1895248.

Singler, Beth. 2020a. “‘Blessed by the algorithm’: Theistic conceptions of artificial intelligence in online discourse.” AI & Society 35, no. 4: 945–55. https://doi.org/10.1007/s00146-020-00968-2.

———. 2020b. “The AI creation meme: A case study of the new visibility of religion in artificial intelligence discourse.” Religions 11, no. 5: 253. https://doi.org/10.3390/rel11050253.

———. 2020c. “Artificial intelligence and the parent–child narrative.” In AI narratives: A history of imaginative thinking about intelligent machines, edited by Stephen Cave, Kanta Dihal, and Sarah Dillon, 260–83. Oxford: Oxford University Press. 

———. 2022. “Origin and the end: Artificial intelligence, atheism, and imaginaries of the future of religion.” In Emerging voices in science and theology, edited by Bethany Sollereder and Alister McGrath, 105–20. London: Routledge.

Stephens, Elizabeth. 2023. “The mechanical Turk: A short history of ‘artificial artificial intelligence.’” Cultural Studies 37, no. 1: 65–87. https://doi.org/10.1080/09502386.2022.2042580.

Suchman, Lucy. 2006. Human-machine reconfigurations: Plans and situated actions, 2nd edition. Cambridge: Cambridge University Press. 

———. 2008. “Feminist STS and the sciences of the artificial.” In The handbook of science and technology studies, edited by Sheila Jasanoff et al., 139–64. London: SAGE. 

———. 2023 “The uncontroversial ‘thingness’ of AI.” Big Data & Society 10, no. 2: 1–5. https://doi.org/10.1177/20539517231206794. 

Tambiah, Stanley J. 1990. Magic, science and religion and the scope of rationality. Cambridge: Cambridge University Press.

Taylor, A.R.E. 2022. “Cloudwork: Data centre labour and the maintenance of media infrastructure.” In The Routledge companion to media anthropology, edited by Elisabetta Costa, Patricia G. Lange, Nell Haynes, Jolynna Sinanan, 213–28. London: Routledge.

White, Daniel, and Hirofumi Katsuno. 2021. “Toward an affective sense of life: Artificial intelligence, animacy, and amusement at a robot pet memorial service in Japan.” Cultural Anthropology 36, no. 2: 222–51. https://doi.org/10.14506/ca36.2.03.

———. 2023. “Modelling emotion, perfecting heart: Disassembling technologies of affect with an android bodhisattva in Japan.” Journal of the Royal Anthropological Institute 29, no. 1: 103–23. https://doi.org/10.1111/1467-9655.13813. 

Vidal, Denis. 2007. “Anthropomorphism or sub‐anthropomorphism? An anthropological approach to gods and robots.” Journal of the Royal Anthropological Institute 13, no. 4: 917–33. https://doi.org/10.1111/j.1467-9655.2007.00464.x.

Note on contributor

Jennifer Cearns is Lecturer in AI at the Department of Social Anthropology at the University of Manchester. Her research currently focuses on human-AI relations, affective computing and empathic AI, and how digital technologies affect personhood and identity. She has conducted ethnographic research in Brazil, Cuba, the US, Guyana, Panama, Mexico, Spain, and the UK. She is the author/editor of two books, Contraband cultures (UCL Press, 2024) and Circulating culture (2023, University of Florida Press), the latter of which was shortlisted for the Association of Latina and Latino Anthropologists First Book Prize and the Society for Latin American & Caribbean Anthropology Book Prize by the American Anthropological Association. In 2026, she gave a TED Talk called “How AI makes us more human”.

Dr. Jennifer Cearns, Department of Social Anthropology, University of Manchester, email: jennifer.cearns@manchester.ac.uk; www.jennifercearns.com / @jenerative.ai. ORCID: 0000-0001-6498-1766