Critical Perspectives on AI: Discourse, Culture, and Power A Deep Research Report

 

Critical Perspectives on AI: Discourse, Culture, and Power

Introduction

Artificial intelligence is often discussed in terms of technical capability, but a growing body of thinkers insists we understand AI as a discourse-producing system embedded in culture, power, and ideology. These scholars – spanning cultural studies, critical media theory, science and technology studies (STS), political economy, feminist theory, and critical geography – interrogate how AI shapes meaning and subjectivity rather than treating it as a neutral tool. They ask not just what AI does, but who it serves, which values it encodes, and how it restructures social relations. In contrast to the enchanted hype surrounding “magical” AI, critical perspectives reveal a technology deeply entangled with human biases, political economies, and historical power dynamicsdukeupress.wordpress.comdukeupress.wordpress.com. As Lauren Goodlad writes, AI’s recent boom is “powerful sleight of hand” whereby “the creative and communicative expression of the many… becomes a lucrative and exploitable resource for the few,” wrapped in monopolistic control, hidden labor, environmental cost, and “unaccountable biases and harms” obscured by hypedukeupress.wordpress.comdukeupress.wordpress.com. In what follows, we map the foundational and contemporary thinkers analyzing AI’s cultural and discursive dimensions, outline their key themes and frameworks, and examine how their approaches converge or conflict. We then group these thinkers into thematic clusters – from critiques of algorithmic bias and identity to analyses of surveillance capitalism and data colonialism – and highlight the intellectual traditions (e.g. Foucauldian, Frankfurt School, feminist, decolonial) they build upon or challenge. A visual schematic is included to illustrate how AI functions as a socio-technical system of power, knitting together the various threads of these critical analyses.

Key Thinkers and Their Contributions

  • Kate Crawford – A leading voice in critical AI studies, co-founder of the AI Now Institute, and author of Atlas of AI (2021). Crawford examines AI as a “power structure” entwined with politics, capital, and planetary resourcessobrief.com. She argues AI is not an abstract, neutral intelligence but a material industry of extraction – reliant on mining minerals, harvesting data, exploiting labor, and consolidating power in a few tech corporationstheverge.comsobrief.com. In Atlas of AI, she maps how seemingly magical AI services are underpinned by resource-intensive infrastructures and hierarchies of power. For example, training data are scraped from online communities without consent, treating human experience as raw material; “classification” by AI is shown to be a form of power that encodes the biases of its creatorssobrief.com. Crawford’s work highlights how AI systems magnify inequity by design, from biased facial recognition to military surveillance, and how the “Great Houses” of Big Tech exercise disproportionate controlsobrief.comsobrief.com. She calls for connecting social movements – data rights, labor rights, racial and environmental justice – to challenge the extractive logics of AI and imagine more equitable alternativessobrief.comsobrief.com.

  • Ruha Benjamin – A scholar of African American Studies and STS, known for Race After Technology: Abolitionist Tools for the New Jim Code (2019). Benjamin explores how modern “neutral” technologies perpetuate old forms of racial domination. She introduces the concept of the “New Jim Code,” defined as “the employment of new technologies that reflect and reproduce existing inequities but are promoted and perceived as more objective or progressive than the discriminatory systems of a previous era”aaihs.org. Through examples from automated soap dispensers to predictive policing, Benjamin reveals “engineered inequity” – design choices that encode white supremacist biases into algorithms and devicesaaihs.orgaaihs.org. Her work blends critical race theory with STS (an approach she terms “race critical code studies”) to show how technologies created in an unequal society will systematically disadvantage marginalized groupsaaihs.orgaaihs.org. Benjamin’s abolitionist perspective not only critiques biased algorithms but also imagines tools and practices to dismantle these discriminatory systems. She situates AI bias in a longer history of racial “coding” (from Jim Crow laws to “colorblind” policies) and calls for an actively anti-racist, justice-oriented approach to tech designaaihs.orgaaihs.org. Her phrase “the New Jim Code” has become a key framework for understanding algorithmic racism and the urgency of reforming or abolishing such systems.

  • Safiya Umoja Noble – Author of Algorithms of Oppression: How Search Engines Reinforce Racism (2018). Noble’s research focuses on how seemingly objective search algorithms reproduce societal prejudices. She famously demonstrated that a Google search for “Black girls” yielded pornographic or demeaning results – an illustration that search engines are far from neutral and in fact “replicate and reinforce racist and sexist beliefs that reverberate in society”blogs.lse.ac.uk. Noble shows that search rankings reflect commercial interests and existing hegemonies: marginalizing voices of women of color while amplifying harmful stereotypesblogs.lse.ac.ukblogs.lse.ac.uk. She argues that the ideology of tech-neutrality hides the reality that algorithms are created by humans with biases, and that companies like Google optimize for profit or ad revenue in ways that can privilege misinformation and bigotryblogs.lse.ac.ukblogs.lse.ac.uk. Noble places these practices in the context of neoliberalism and mass media, noting how corporate control of information online can distort knowledge and even radicalize users (as in the case of extremist propaganda found via search)blogs.lse.ac.ukblogs.lse.ac.uk. Her work calls for greater accountability for platform algorithms and suggests that oppressed communities must regain power in defining their own representations onlineblogs.lse.ac.uk. Algorithms of Oppression has become a foundational text in showing the real-world harms of algorithmic bias and the need to treat information platforms as sites of social struggle rather than neutral utilities.

  • Shoshana Zuboff – A scholar in social psychology and business, known for her monumental book The Age of Surveillance Capitalism (2019). Zuboff analyzes how global tech companies like Google and Facebook pioneered a new political economy built on harvesting personal data. She terms this system surveillance capitalism, defined as “a new economic order that claims human experience as free raw material for hidden commercial practices of extraction, prediction, and sales.”eastvillagemagazine.orgeastvillagemagazine.org In Zuboff’s account, firms extract a “behavioral surplus” – data about our behaviors, emotions, and interactions – which is then fed into machine intelligence to predict and shape future behavior for profiteastvillagemagazine.orgeastvillagemagazine.org. This goes beyond targeted ads; it amounts to a “rogue force” in which companies seek to modify human behavior at scale, eroding autonomy and democracyeastvillagemagazine.org. She draws an analogy to industrial capitalism’s exploitation of nature: just as industrialism ravaged the earth, surveillance capitalism “unilaterally claims human experience” and threatens to cost us our humanity if left uncheckedeastvillagemagazine.org. Zuboff’s key concepts include instrumentarian power (power achieved through instrumenting behavior via data and algorithms) and behavioral futures markets (markets trading in predictions of our future actions)eastvillagemagazine.orgeastvillagemagazine.org. Her work, grounded in a Frankfurt School-like critique of instrumental reason, highlights the asymmetry of knowledge (corporations know everything about individuals, individuals know next to nothing about the operations of corporate AI)eastvillagemagazine.orgeastvillagemagazine.org. By treating personal data as a free resource and prioritizing corporate surveillance over individual rights, surveillance capitalism, Zuboff warns, creates unprecedented inequalities of power and “a new collective order” that must be confronted through regulation and public awarenesseastvillagemagazine.orgeastvillagemagazine.org.

  • Nick Seaver – An anthropologist of technology who studies algorithms as cultural artifacts. In Computing Taste: Algorithms and the Makers of Music Recommendation (2023), Seaver provides an ethnography of the engineers behind recommender systems (like music streaming algorithms). He challenges the common image of algorithms as inhuman, purely technical entities, showing instead that algorithms are products of human culture and decision-making. Seaver notes that a prominent discourse imagines algorithms as “essentially inhuman, pursuing uniquely technical or mathematical goals” – a view that casts them as anti-cultural forces chopping human culture into bitscampanthropology.org. However, his research found that “humans are all over the place deciding how things work” in algorithm designcampanthropology.org. Engineers carry folk theories and cultural assumptions (about taste, genre, what users want) that profoundly shape how their systems functioncampanthropology.org. As Seaver writes, “the way an engineering team thinks about their work is as significant to the functioning of the system as the structure of the code itself… if we tried to look past all the people to find ‘the algorithm,’ we’d miss a crucial part of how and why these systems work as they do.”campanthropology.org This perspective aligns with broader STS insights that technology is socially constructed. Seaver’s work implies that algorithms carry the imprint of their creators’ values and the organizations’ culture – whether that is an ethos of maximizing user engagement, a notion of “taste” in music, or anxieties about the system’s social impact. By humanizing the algorithm, Seaver slightly complicates more deterministic critiques: he shows that while algorithms can reinforce power or bias, they are also contingent cultural products, open to change if the people and values behind them change. His work builds a bridge between cultural anthropology and AI, insisting that we study “algorithms in practice” – including the intentions, improvisations, and institutional pressures behind their developmentcampanthropology.orgcampanthropology.org.

  • Safiya Noble, Ruha Benjamin, and Joy Buolamwini – Critical race and gender perspectives on algorithmic bias. Noble and Benjamin have been discussed above; they form part of a wider movement of Black feminist scholars interrogating how algorithms can perpetuate racism and sexism. Joy Buolamwini, a computer scientist and activist, exposed racial bias in commercial facial recognition systems (in the Gender Shades project). Her finding – that these AI vision systems had significantly higher error rates for darker-skinned and female faces – dramatized the exclusionary default in AI training data (overrepresentation of lighter male faces) and helped spur industry and policy attention to algorithmic biasaaihs.orgaaihs.org. Buolamwini founded the Algorithmic Justice League to advocate for more equitable AI and has framed bias in AI as not just a technical flaw but a social injustice requiring activism. Similarly, Timnit Gebru, a former Google researcher, has been influential in highlighting bias and ethical issues in large language models and datasets – and her controversial firing from Google in 2020 for raising these issues underscores the tensions between Big Tech and critical researchers. Together, these contemporary thinkers push for “algorithmic accountability,” bringing concepts from critical race theory and intersectional feminism to bear on AI. They emphasize that issues like biased policing algorithms, discriminatory credit scoring, or misgendering by AI systems are connected to larger structures of oppression. Their work often aligns with abolitionist tech ethics – envisioning the reduction or removal of harmful automated systems (e.g. moratoria on facial recognition, as Buolamwini successfully campaigned for) and the development of community-driven alternatives.

  • Frank Pasquale – A legal scholar whose book The Black Box Society (2015) examines the opaque algorithms in finance, search, and social media that make consequential decisions about individuals. Pasquale’s work is grounded in a political economy and legal perspective: he critiques how corporate secrecy around algorithms (the “black boxes”) undermines transparency, accountability, and governance. He argues that when algorithms rank us, score us, or predict our behavior without oversight, they can reinforce inequalities (for example, search engine autocomplete spreading hateful stereotypes, or credit scoring penalizing the poor). Pasquale calls for “algorithmic accountability” through stronger regulation, auditing, and the right to explanation. He also writes about the political economy of AI more broadly, critiquing the notion that tech companies should self-regulate. In recent work, Pasquale contrasts “permissionless” innovation ideology with approaches that emphasize human rights and dignity (sometimes called “critical AI” vs. “ethical AI” debates). In the landscape of critical AI thinkers, Pasquale represents a juridical and policy-oriented strand, aligning with scholars like Julie Cohen or Virginia Eubanks who focus on law, surveillance, and inequality (Eubanks’ Automating Inequality documents how AI in welfare systems can deepen social stratification). His contributions underscore that power asymmetries in the age of AI are not inevitable; they are the result of policy choices and can be changed with the right legal frameworks.

  • Nick Couldry & Ulises A. Mejias – Media theorist and communication scholar (Couldry) and information scientist (Mejias) who together developed the concept of “data colonialism.” In their book The Costs of Connection (2019), they argue that the global data economy operates as a new colonial system, appropriating human life and knowledge as resources. Just as historical colonialism seized land, labor, and bodies, data colonialism “mirrors the historic appropriation of land and resources” by treating personal data and online behaviors as free for the takingsup.org. Apps, smart devices, and platforms capture every aspect of life as data, which is then funneled into capitalist enterprisessup.orgsup.org. Couldry and Mejias stress this is not mere metaphor: “data colonialism is not a metaphor. It is a process that expands many dark chapters of the past into our shiny new world of smartphones, smart TVs, and smart stores.”sup.org. The result is a “new social order” where human behavior is continuously monitored, and where knowledge and power are concentrated in those who control data. This framework draws on decolonial theory and political economy, linking the likes of Zuboff’s surveillance capitalism to an even broader historical arc of expropriation and control. Importantly, they highlight the lived impacts: how people’s “ways of knowing, means of production, and political participation” are being controlled by this regimesup.org. Their work aligns with other critiques of the geopolitics of AI – for example, how AI development relies on low-wage data labor in the Global South (data labeling farms, content moderation in places like the Philippines or Kenya), reminiscent of colonial labor structures. By conceptualizing data extraction as colonialism, Couldry and Mejias call for a “decolonization” of the digital sphere – reasserting data sovereignty for individuals and communities, and resisting the idea that continuous surveillance is a necessary price of connectionsup.orgsup.org. This perspective has been influential in connecting AI to global inequality and historical patterns of domination.

  • Michel Foucault (Foundational) – Although Foucault did not write about AI (he died in 1984 just as computing was accelerating), his theoretical frameworks are widely applied by contemporary scholars to make sense of AI’s discursive and power effects. Foucault’s concept of discourse – systems of knowledge that define what is true and normal – informs analyses of AI as producing new “regimes of truth.” For instance, algorithms decide what information is seen as relevant or which individuals are categorized as risky, thereby exercising a form of power/knowledge. Scholars have likened modern data-driven management to Foucault’s idea of governmentality: AI enables a mode of governance that operates through continuous monitoring and normalization rather than overt coercion. “Algorithmic governmentality,” a term coined by Antoinette Rouvroy, describes governance by algorithms that bypass the need for individuals’ active participation or subjectivity, acting directly on data traces to make decisionsmichel-foucault.commichel-foucault.com. In this scenario, power works through algorithmic predictions that shape people’s opportunities (credit, insurance, policing) without ever engaging them as thinking subjects – fulfilling Foucault’s insight that modern power often operates by organizing knowledge and “conducting conduct” invisibly. Foucault’s notion of biopower (power over populations) is also relevant: AI systems manage populations through statistical predictions (e.g. predictive policing allocating resources to neighborhoods, or health algorithms stratifying who gets care). These are exercises of control at the level of life chances, echoing biopolitics. Moreover, the Panopticon metaphor (from Foucault’s Discipline and Punish) frequently appears in AI discussions – with surveillance cameras, facial recognition, and digital tracking described as a “digital panopticon” that induces self-discipline under the gaze of AI watcherslilacdraccus.medium.com. Contemporary critical geographers like Louise Amoore extend Foucauldian analysis to AI in security and war, examining how risk-scoring algorithms enact a “politics of possibility” (governing by what might happen) that pre-emptively targets individuals in ways that can bypass legal accountability – an extension of Foucauldian preventive logics of power. In sum, Foucault’s emphasis that power is productive (not merely repressive) and that knowledge systems create subjects is a through-line: AI produces new subjectivities (e.g. the “profiled” subject scored by algorithms) and new discourses of what is normal or deviant. Scholars invoking Foucault encourage us to see AI systems not as neutral decision-makers, but as part of an apparatus of power defining reality and steering behavior in line with certain institutional interests.

  • Stuart Hall (Foundational) – Hall’s cultural theory, especially his work on media and representation, provides a basis for understanding AI as part of the cultural circuit of meaning. Hall argued that media representations are not neutral reflections of reality but are constructed within regimes of power, often reinforcing stereotypes and dominant ideologiesmediatheory.net. This insight is clearly echoed in works like Noble’s (on search engine bias) – essentially applying Hall’s notion that what we see in the media (or in search results, or AI-generated content) is the product of social biases and serves specific interests. Hall’s concept of encoding/decoding is also relevant: AI systems encode certain preferred meanings (e.g. a platform encodes what is “trending” as what should be valued), but audiences/users might decode them differently. In algorithmic content feeds, users can resist or reinterpret the information given – though the algorithm strongly frames what is visible (echoing Hall’s theory of hegemonic vs. oppositional readings). Hall also introduced the idea of articulation and cultural hegemony: how disparate elements (like technology and racial ideology) can be articulated together to strengthen a hegemonic worldview. We see this when AI is articulated as a tool of progress while quietly perpetuating racial bias – the ideology of tech-neutrality masks the underlying power structure, much as Hall analyzed how news or TV could mask class/race biases behind an appearance of “natural” reality. Hall’s emphasis on identity and representation being fluid and contested can inform AI critiques too. For example, facial recognition systems impose rigid identity categories (often failing to represent non-normative identities), and critical scholars argue for more inclusive or pluralistic designs – a stance in line with Hall’s advocacy for multicultural, diverse representationsmediatheory.netmediatheory.net. In sum, Hall’s legacy in cultural studies – that media/tech must be studied in terms of how they produce meaning and whose interests that meaning serves – is foundational for many contemporary AI critics. They treat AI systems as part of the broader media landscape that can either challenge or cement social power.

  • Judith Butler (Foundational) – Butler’s theories of performativity and the social construction of identity have become a touchstone in analyzing AI’s impact on gender and identity. Butler argued that gender is “not a noun, but a verb” – an ongoing performative act rather than a fixed essencephilosopheasy.comphilosopheasy.com. This idea, applied to AI, raises questions about systems that classify gender or make decisions based on gendered assumptions. For instance, algorithmic hiring tools or ad delivery systems often reinforce binary gender roles (showing STEM job ads more to men, for example). A Butlerian lens would note that these algorithms are citing existing gender norms and in doing so, reinforcing them – effectively performing gender by how they sort and treat people. Some scholars explicitly call for a performative approach to AI bias, suggesting that rather than thinking of gender or race bias as attributes that an AI has, we consider how the AI enacts or performs social categories in its operationresearchgate.net. Butler’s insight that identity categories gain power through repetition (and can be subverted through parody or refusal) suggests strategies for resistance in the AI realm: for example, “algorithmic self-defense” tactics like altering one’s digital trace or using adversarial inputs to prevent an AI from correctly gendering you could be seen as a form of gender performance that confounds the system. In digital and social media spaces, Butler’s work helps explain phenomena like people experimenting with identity via avatars or pseudonyms – highlighting how online platforms can both liberate and constrain performative identityphilosopheasy.comphilosopheasy.com. Moreover, Butler’s focus on how power regulates which performances are deemed legitimate intersects with concerns about AI content moderation and community standards – algorithms often encode what kind of speech or self-presentation is acceptable, sometimes in ways that silence non-normative identities. In summary, Butler provides a framework to critique AI systems that treat identity as static or purely data-driven: these systems often fail to grasp the fluidity and contextuality of identity, and thus can impose harms (like misgendering or reinforcing stereotypes). By drawing on Butler, theorists encourage designs that recognize user identities as dynamic and plural, and caution against AI applications (like emotion recognition or gender detection algorithms) that claim to read inner truths from outer appearance – a notion fundamentally challenged by poststructuralist gender theory.

  • Donna Haraway (Foundational) – A prominent feminist theorist of science and technology, Haraway’s work (e.g. the “Cyborg Manifesto” 1985) has heavily influenced critical thinking about AI. Haraway famously embraced the metaphor of the cyborg – a hybrid of human and machine – to challenge essentialist boundaries and imagine more egalitarian futures. Today’s AI assistants and wearable tech make cyborg metaphors literal, but Haraway’s deeper point was to situate knowledge and expose how technology is never neutral. She advocated “situated knowledges”, arguing that all knowledge (even scientific/AI outputs) comes from particular embodied perspectives. Critical AI scholars channel Haraway when they point out that AI’s “view from nowhere” – claiming objective omniscience – is a fallacy; in truth, AI reflects the standpoints (often white, male, Western, capitalist) of its creators or training data. Haraway also presaged today’s debates on AI ethics by critiquing the god-trick of objectivity and calling for accountability and partial perspectives in tech. Additionally, her cyborg imagery has been taken up in discussions of AI and the self: rather than fear cyborgification, Haraway suggested using it to break down binaries (human/machine, male/female, etc.). Some contemporary thinkers like Aubrey Anable or Katherine Hayles extend this to AI, seeing opportunities for new forms of subjectivity that AI might enable, even as they warn of the militarized, patriarchal uses of these technologies. Haraway’s influence is evident in feminist data science (e.g. “data feminism” by Catherine D’Ignazio and Lauren Klein, which explicitly cites Haraway’s approach to embrace plural, community-driven ways of knowing with AI) and in critical robotics (the idea of “companion species” she developed, applied to how we might relate to AI assistants or robots in non-dominating ways). In sum, Haraway’s work contributes a decolonizing, feminist lens that many critical AI scholars use: emphasizing coalition (cyborg as coalitional figure), challenging the myth of neutral mastery, and encouraging imaginative reconfigurations of human-machine relationships away from domination and toward mutuality.

(The above list is not exhaustive, but covers many of the influential figures and ideas at the intersection of AI and critical theory. Other notable thinkers include Lucy Suchman, who as early as 1987 critiqued AI’s assumptions of rationality with her Plans and Situated Actions (arguing that human action is too context-bound to be captured by formal AI plans); Wendy Hui Kyong Chun, whose recent work Discriminating Data (2021) explores how big data and predictive algorithms reinforce segregation and call for a new understanding of “habit” and connection; Tarleton Gillespie, who examines the custodians of algorithms (like content moderators and platform policies) and how they shape public discourse; Benjamin Bratton, who in The Stack (2015) conceptualizes the planetary-scale computing infrastructure (from user layer to cloud to address the geopolitical and architectural power of AI infrastructures); and Hannah Arendt-inspired commentators who worry about the loss of human agency and judgment in an age of automated decision-making. Each brings their disciplinary lens – be it political theory, media anthropology, or philosophy – to bear on the entanglement of AI with society.)

Thematic Clusters and Frameworks

While each thinker above has a distinct focus, their ideas converge into several thematic clusters. These clusters represent key problematics and frameworks for analyzing AI’s role in culture and power:

1. Algorithmic Bias, Inequality, and Social Justice

One major theme is that AI systems can encode and amplify biases, disproportionately harming marginalized communities. Scholars in this cluster (e.g. Ruha Benjamin, Safiya Noble, Joy Buolamwini, Virginia Eubanks, Cathy O’Neil) document examples of racist, sexist, or classist outcomes in AI – from hiring algorithms that reject women, to risk assessment tools that profile Black defendants as higher risk, to health algorithms that underserve poor patients. The framework they employ often draws on critical race theory, Black feminist thought, and intersectionality, seeing AI as the latest arena in which structural inequalities play out. Key concepts include Benjamin’s “New Jim Code” (the continuity of racism through coded systemsaaihs.org), Noble’s description of technological redlining (the way search engines or ad systems can exclude or exploit communities of color), and Buolamwini’s term “coded gaze” (the biased gaze of the algorithm’s creators reflecting in its outputs). These theorists link back to Stuart Hall’s point that media/tech are not depicting an objective reality but are actively constructing a social reality that often privileges the already powerfulmediatheory.net. A shared argument is that bias is not merely a technical glitch to be fixed with better training data; rather, it stems from and perpetuates systemic injustice. Thus, technical solutions (like debiasing algorithms) must be coupled with social and institutional change. We also see influence of Foucauldian discourse analysis here: the very discourse that paints AI as objective masks what Benjamin calls the “gloss of objectivity” over racismaaihs.org. Scholars in this cluster often advocate for algorithmic transparency, community oversight, and participatory design as ways to inject justice into AI. They sometimes diverge, however, on how radical the response should be – with some pushing for abolition of certain AI uses (e.g. predictive policing) and others for reforms (more diversity in AI development, better regulations). Nonetheless, they collectively ensure that questions of race, gender, and equity remain central in any conversation about AI’s impact.

2. AI, Capitalism, and Data Colonialism

Another cluster centers on political economy: the marriage of AI with capitalism, surveillance, and global exploitation. Shoshana Zuboff’s concept of surveillance capitalism anchors one part of this discussion – framing AI-driven data extraction as a new stage of capitalism where human behavior itself is commodifiedeastvillagemagazine.orgeastvillagemagazine.org. In parallel, Nick Couldry and Ulises Mejias’s concept of data colonialism extends this critique historically and globally, arguing that the appropriation of data is akin to colonial appropriation of land and bodiessup.orgsup.org. Both frameworks describe a totalizing system: surveillance capitalism turns every user interaction into a behavioral data point to be monetized, while data colonialism posits that no aspect of life is outside the reach of data extraction. These thinkers align with traditions of the Frankfurt School and Marxist critique, insofar as they see advanced technology reinforcing a total system of domination (where human freedom and unpredictability are threatened by regimes of measurement and control). They also invoke imperial and world-systems analysis – e.g., pointing out that AI’s benefits and harms are unevenly distributed, often following colonial-era North/South divides (with the Global South providing raw materials and ghost labor for the AI of the wealthy North). The geopolitics of AI enters here too: some critical theorists note how AI development is a site of international struggle for power (e.g. US and China competition), raising questions about digital sovereignty and the risk of new forms of imperialism (for instance, authoritarian uses of AI for population control, or Western tech firms imposing AI systems on other societies). Within this cluster, key themes include: the surveillance state (e.g. extensive use of AI for citizen monitoring in China’s social credit system or the West’s intelligence agencies – analyzed through lenses of Orwell, Foucault, and Agamben’s states of exception), consumer manipulation (the Facebook/Cambridge Analytica scandal as a case of AI-driven political propagandatheverge.com), and the economics of automation (who gains from AI-driven productivity and who loses jobs or agency). There is some tension within this camp: Zuboff tends to emphasize the novelty of surveillance capitalism as sui generis, whereas others like Couldry/Mejias or Christian Fuchs might stress continuity with existing capitalism (arguing it’s capitalism adapting rather than a wholly new logic). But all agree that AI is entangled with capitalism’s drive for profit and control, and they often call for robust democratic governance – from antitrust measures on Big Tech to data protection laws – to reclaim personal and collective autonomy from these systems.

3. Knowledge, Discourse, and the Power of AI Narratives

AI doesn’t just do things in the world; it also transforms how we make meaning and what we consider knowledge. A number of thinkers focus on AI as a discursive and epistemic agent. They might ask: How is the idea of AI used to shape policy and public imagination? (For instance, the narrative of AI as inevitable progress can discourage critique – a concern Goodlad raises about hypedukeupress.wordpress.com.) Or, how do AI systems themselves produce or organize knowledge? (Think of YouTube’s recommendation algorithm as a cultural curator, or large language models that generate texts – effectively participating in discourse formation.) Here we see explicit use of Foucault’s theories: AI can be seen as producing a new “regime of truth,” where certain things count as true or valuable because algorithms say so. The concept of algorithmic bias ties in, but beyond bias, it’s about how the structure of algorithmic information favors certain interpretations of reality. For example, data-driven predictive policing implicitly treats crime as a phenomenon divorced from social context – a discourse of criminality as a mappable risk density, which critics argue ignores structural causes and thus perpetuates a punitive view of certain neighborhoods. This cluster also includes media theorists who examine AI in the context of information ecosystems: how do recommendation systems, trending algorithms, search autocompletes, etc., shape public discourse and what we talk about? Tarleton Gillespie (who studied the curation functions of platforms) and Zeynep Tufekci (who wrote on how YouTube’s algorithm can radicalize by leading viewers to ever more extreme content) are relevant names. They build on concepts like agenda-setting and ideological state apparatuses (à la Althusser) – essentially updating them: now algorithms, rather than TV networks or schools, are key apparatuses influencing ideology. Another important concept is “algorithmic mediation” – the idea that our communication is increasingly intermediated by AI (from dating apps to news feeds), which raises questions about power and agency in those mediations. Are we freely forming opinions, or are algorithms nudging us in particular directions? Some, like Jaron Lanier or Frank Pasquale, warn of a “faceless algorithmic editor” controlling information flows in ways we don’t see. On the other hand, optimistic media scholars might see potential for new discourses (e.g., AI highlighting voices that mainstream media ignored, though evidence for this is mixed). Within this cluster, thinkers also analyze the rhetoric around AI – e.g., how terms like “artificial intelligence” themselves carry cultural meanings (often inflated or misrepresentative, as Goodlad notes the science-fiction connotationsdukeupress.wordpress.com). They might point out that calling something “AI” can black-box it socially, making people less likely to scrutinize it because it’s seen as advanced and inscrutable. Thus, a critical task is demystifying AI – replacing the enchanted discourse with one that reveals AI’s mundane reality (statistics, human work behind the scenes, etc.), which is exactly what many of the aforementioned scholars do. In summary, this thematic cluster treats AI as a phenomenon in language and knowledge, probing how it influences what society considers true, normal, or important. It leans on communications theory, epistemology, and linguistics, intersecting with both Foucault (knowledge-power) and Hall (media framing), and is practically important for understanding phenomena like misinformation, public opinion manipulation, and the cultural narratives that either legitimize or question AI’s expanding role.

4. Subjectivity, Identity, and Human-Machine Relations

A further cluster of analysis looks at how AI shapes our sense of self and our social identities, as well as the nature of human agency in an AI-pervaded world. Here, Judith Butler’s influence is evident in questions of how rigid algorithmic categories clash with fluid identities. For instance, facial recognition systems that attempt to classify gender not only often fail non-binary people, but in doing so they perform and enforce a binary concept of gender. Similarly, recommendation algorithms create “algorithmic identities” – profiles of users’ tastes and traits (as discussed by John Cheney-Lippold in We Are Data). These algorithmic identities can become self-fulfilling: you start seeing yourself as the kind of person that the algorithm assumes you are, because it consistently feeds you certain content (a process akin to performativity where the repetition of categorization produces an identity). Some theorists raise alarms that AI can narrow one’s subjectivity – e.g., if a music app decides you are a “rock listener” and never exposes you to jazz, does your taste ossify? Nick Seaver’s work, however, suggests there is still play and negotiation here: users aren’t passive, and engineers sometimes deliberately design to broaden horizons (though often they struggle, since the business models favor keeping people predictable)campanthropology.org. Another perspective in this cluster comes from philosophy of mind and posthumanism: thinkers like Katherine Hayles or Luciano Floridi ask how human self-understanding changes when AI entities are part of our everyday life or when we delegate cognitive tasks to machines. Hayles’s idea of “cognitive assemblages” sees human and AI as interlinked actors in meaning-making networks. Meanwhile, existential and phenomenological angles (some draw on Heidegger or Arendt) consider whether relying on AI for decision-making could erode our capacity for critical thought and moral judgment (Arendt’s concerns about thoughtlessness in bureaucratic systems resonate here). On identity, critical feminist and queer scholars note that AI applications often reflect a narrow understanding of identity – e.g., Instagram’s algorithms privileging certain beauty standards, or voice assistants being given female names and submissive personas, reinforcing gender stereotypes. This cluster is also where ethics of AI conversations often land: what does it do to human dignity or agency if an AI is always scoring us (credit scores, social media influence scores, etc.)? Drawing on Habermas or Kant, some argue this treats people as means to an end (violating autonomy). Thinkers like Evgeny Morozov and Langdon Winner earlier warned about technological systems “programming” humans in subtle ways (Winner’s famous question “do artifacts have politics?” can be applied: yes, AI artifacts often have politics that shape subjects). A fascinating point of debate is whether AI can create new subjectivities or subcultures – for example, the emergence of “AI influencers” (completely synthetic social media personas) or the idea of people identifying with certain algorithm-driven communities (like how TikTok’s algorithm fosters niche subcultures). Optimists might cite Haraway’s cyborg as a positive new subjectivity – mixed and liberated from old categories – whereas pessimists might cite Foucault, saying these are just new dividing practices to categorize and control us. In any case, this theme keeps the human element at the center: it insists that beyond metrics of accuracy or efficiency, we ask how AI alters our lived experience, our interactions, and even our interior life (e.g., when you know a “score” or algorithm is judging you, you may start to self-censor or alter behavior – a kind of internalized surveillance which echoes Bentham/Foucault’s panoptic effect).

5. Camps and Crossroads of Thought

Finally, it’s helpful to recognize how these thinkers and themes cluster into a few intellectual camps or lineages, even as many of them engage in cross-disciplinary borrowing:

  • Foucauldian / Post-Structuralist Camp: Emphasizes discourse, power, governmentality. Includes scholars explicitly citing Foucault (Rouvroy on algorithmic governmentality; some critical data studies folks examining how AI defines what is “normal” or “risky”). They often focus on how AI systems classify and order knowledge, producing new power relations without explicit coercion. They may conflict with more Marxian folks by downplaying economic exploitation in favor of subtle power dynamics, but in practice many integrate both (e.g., a Foucauldian reading of surveillance capitalism sees it as a new dispositif of power/knowledge). They resist technological determinism, seeing AI not as an independent force but as embedded in social practices and discourses that can be contested.

  • Frankfurt School / Critical Theory Camp: Draws on Marx, Adorno, Horkheimer, Marcuse, etc. It views AI as part of the culture industry or as an instrument of instrumental rationality that could lead to a more administered, one-dimensional society. Zuboff, while not explicitly Frankfurt School, echoes this tradition’s alarm at how enlightenment (here in the form of high-tech AI) can become myth and dominationeastvillagemagazine.org. These thinkers align on critiquing the commodification of human life and the alienating effects of AI (e.g., workers reduced to data points, creativity reduced to patterns). They sometimes clash with Foucauldians over method (Frankfurt tends to moral philosophy and macro-economics, Foucauldians to micro-power analysis), but both share a skeptical stance toward claims of neutrality and progress. A subset of this camp is the political economy of communication tradition (like Dallas Smythe, Herbert Schiller) updated for AI, examining how control of data and algorithms by corporations is a form of capitalist power that the public must democratically check.

  • STS / Anthropological Camp: This includes Seaver, Suchman, and others like Bruno Latour or Madeleine Clare Elish (who wrote about “moral crumple zones” in automated systems). They study how humans and machines interact in practice, often via case studies and ethnography. Their frameworks (actor-network theory, ethnomethodology, etc.) highlight that AI’s effects are not universal; they depend on local contexts, implementations, and the actors involved. They sometimes criticize the sweeping generalizations of the other camps – for instance, an STS ethnographer might say, “Not all AI is the same; look at how a medical diagnostic AI is used by doctors in one hospital, it might empower them rather than control them.” This can clash with, say, a data colonialism theorist who is painting with a broad brush. Yet there’s also synergy: STS findings of bias or misalignment on the ground feed critical theorists’ broader claims. This camp tends to push for situated, bottom-up understanding and often engages directly with technologists, sometimes making them more palatable to industry than the more radical critics.

  • Decolonial / Critical Race Feminist Camp: Centering authors like Benjamin, Noble, Couldry & Mejias, Sasha Costanza-Chock (who wrote Design Justice), and many others, this camp foregrounds how AI intersects with race, gender, and global power asymmetries. They draw from Black studies, indigenous studies (e.g., some discuss how AI can be a form of “digital extraction” from indigenous lands and knowledge), and intersectional feminism. They align with the Frankfurt camp in seeing AI as tied to capitalism, but add that it is specifically the racial capitalism or patriarchal capitalism that must be named. They also align with Foucauldians in seeing the continuation of colonial power/knowledge in AI’s classifications (e.g., skin tone databases reflecting colonial era’s typologies). This camp often advocates not just critique but also designing AI for social justice – community-led AI projects, bias audits, ethical guidelines informed by marginalized groups, etc. Within this group, there can be differences: some work from within institutions (e.g., advising tech companies on fairness) while others are more activist and oppositional (calling for abolishing some tech altogether).

  • Humanist / Philosophical Ethics Camp: Thinkers here (some overlap with others) include those doing AI ethics from philosophical perspectives – e.g. drawing on Aristotle, or virtue ethics, or phenomenology. They might be less visible in our discussion but exist in the broader discourse (like Floridi’s information ethics, or Sven Nyholm’s ethics of AI relationships). They sometimes clash with the critical theory folks by focusing on abstract principles (e.g., “Is AI conscious? Can it have rights?” or “How do we ensure AI is fair?”) rather than systemic power. However, recently there’s convergence in “critical AI ethics” that merges these, as suggested by philosophical arguments that AI ethics must be about changing society (an idea put forth by some as “AI ethics is a critical theory” in that it seeks not just to understand but to liberatelink.springer.com). This camp tends to seek dialogue with policy (e.g., advising governments on ethical frameworks), and sometimes is critiqued by others as focusing too much on future hypotheticals (like superintelligence) instead of current injustices – a divide often noted between “ethics of AI” vs “politics of AI” approaches.

The camps above are not rigid; many thinkers move between them or combine elements. For example, Kate Crawford combines feminist/STS (Haraway-influenced) with political economy (talking about labor and power) and even Foucauldian notes (the idea of “anatomy” of AI resonates with archaeology of knowledge)anatomyof.aianatomyof.ai. Tensions do arise: e.g., an STS researcher might accuse a Frankfurt-style critic of technological determinism or lack of empirical grounding, whereas the critic might retort that STS descriptions can be apolitical or too incremental. Yet, a shared mission unites these perspectives: to situate AI in its social, cultural, and historical context, and to challenge the notion that AI’s impacts are either entirely positive or beyond human control.

Visualizing AI’s Socio-Technical Anatomy

Figure: Conceptual map of an AI system’s extractive processes, adapted from Crawford & Joler’s “Anatomy of an AI System.” This diagram illustrates how an everyday AI device (the Amazon Echo) is interwoven with human labor, data, and capital. Far from a neutral gadget, the device links the user’s interactions (voice commands) to vast industrial and data networks. As shown, the consumer is simultaneously positioned as a source of data, an unpaid worker providing feedback, and a target for monetization – embodying Crawford’s point that AI turns human experience into free raw materialtheverge.com. The map also hints at hidden layers: supply chains of mined minerals, global assembly labor, cloud infrastructures, and algorithmic models – each layer governed by corporate power. Such visuals help synthesize the critical insights: AI systems operate through extraction (of resources, labor, data) and concentration (of knowledge, power), raising questions of justice that the thinkers discussed above collectively address.theverge.comtheverge.com

Conclusion

In surveying the landscape of critical thought on AI, we find a rich dialogue across disciplines. Cultural theorists and STS scholars reveal AI as a cultural artifact – shaped by and shaping the meanings, values, and identities in society. Critical race and feminist scholars expose how AI can reinforce structural oppressions, yet also articulate visions for more inclusive, emancipatory technologies. Political economists and geographers situate AI within global flows of capital and power, connecting silico to soil and code to colonialism. And through it all, foundational frameworks – from Foucault’s power/knowledge to Hall’s representation to Butler’s performativity – provide language to decipher how something as technical-sounding as “algorithm” is also profoundly social. These thinkers do not always agree: they variously emphasize discourse or economics, human agency or structural forces, reform or abolition. However, their approaches converge in rejecting the myth of AI as autonomous and value-free. Instead, AI is understood as “a technology that is being packaged as if it were beneficial to most people,” when in fact it often reflects the goals of those who design and deploy itdukeupress.wordpress.com. By aligning and contrasting these perspectives, we see the emergence of a field sometimes called Critical AI Studies, which insists that debates about AI must include questions of meaning, power, and justice – not just efficiency or accuracy.

In practice, this means examining who gets to define the problems AI is used to solve, whose knowledge is prioritized, and who bears the risks. Are we headed toward a future where AI extends existing inequalities and surveillance (a “shiny new” data colony built on old injusticessup.org)? Or can these critical insights guide us to redirect AI toward empowering the marginalized, increasing transparency, and democratizing decision-making? The thinkers profiled here lay the groundwork for the latter, whether by mapping AI’s hidden supply chains (Crawford), coding fairness into technical design (Buolamwini), or reimagining governance structures (Pasquale, Zuboff). As AI increasingly mediates our world, their analyses – sometimes complementary, sometimes in tension – equip us to understand and shape AI as participants in a discourse, not just passive subjects of technology. To delve deeper, readers may explore the cited works (e.g. Crawford’s “Atlas of AI,” Benjamin’s “Race After Technology,” Zuboff’s “Surveillance Capitalism,” Noble’s “Algorithms of Oppression,” and the essays in the new Critical AI journal). Each offers a pathway to critically engaging with AI, ensuring that as this powerful technology evolves, humanistic and social perspectives remain in the loopdukeupress.wordpress.com, interrogating AI’s role in producing our shared future.

Sources and Further Reading: Critical AI Studies is an interdisciplinary field. Key texts and articles include “The Atlas of AI” by Kate Crawfordsobrief.comsobrief.com, “Race After Technology” by Ruha Benjaminaaihs.org, “Algorithms of Oppression” by Safiya Nobleblogs.lse.ac.uk, “The Age of Surveillance Capitalism” by Shoshana Zuboffeastvillagemagazine.org, “Computing Taste” by Nick Seavercampanthropology.org, “The Black Box Society” by Frank Pasquale, “The Costs of Connection: Data Colonialism” by Couldry & Mejiassup.orgsup.org, and articles like “Algorithmic Governmentality” by Antoinette Rouvroymichel-foucault.com. Foundational theories referenced include Foucault’s writings on discourse and governmentality, Hall’s essays on media representationmediatheory.net, and Butler’s Gender Trouble (for performativity)philosopheasy.com. These works, among others, provide a strong starting point for understanding the cultural, discursive, and political-economic dimensions of AI.

Comments

Popular posts from this blog

Trump as the Generational Destroyer (but not the Final Word)

Influence of African English Varieties on LLM Language a Deep Research Report