This article investigates how Artificial Intelligence Generated Content (AIGC) intervenes in memory and what this intervention means for qualitative knowledge production in an emergent “post-reality” condition. Drawing on new materialism theory, this article treats AI as an agential actor and examines how it co-constitutes reality through “intra-action” with human memory. The study employs an innovative experimental design, conducting in-depth interviews with 16 human participants and four AI participants role-played by large language models, based on their responses to AI-generated images. Empirically, the data show that AIGC can cue plausible recollections by matching culturally available memory templates, while disclosure of artificial provenance prompts renewed negotiation of authenticity. Comparison of the two deliberately non-equivalent material sets further reveals a limit to simulation: AI produces standardized recollections and statistical archetypes, whereas human accounts retain embodied, contingent details that we call an “authenticity core.” Our research indicates that the boundary between authentic and synthesized experience has become unstable and must be situated and negotiated. We define post-reality as this generative and contested condition and propose AI Anthropology as a reflexive agenda for qualitative research into experiential worlds co-constituted by humans and machines.
Platform algorithms can pose issues of legibility and exclusion for LGBTQ + people, yet queer perspectives remain largely absent from public debates on artificial intelligence and algorithmic systems. This paper examines how LGBTQ + actors interpret, challenge, and reconfigure algorithmic logics to foreground their critical role in engaging with platform governance. Our analysis brings together the concept of algorithmic imaginaries as shared visions of how algorithms operate and the act of queering as a way of destabilizing normative assumptions embedded in sociotechnical infrastructures. Our study draws on two focus groups conducted in Canada with 11 LGBTQ + participants, including comms professionals from queer non-profits and queer tech workers, during which media controversies were used as an elicitation strategy. Through critical discourse analysis, we highlight how participants queer dominant algorithmic imaginaries through their accounts, critiques, and translations. Findings show that participants unpack the cisheteronormative black box by exposing how bias is reproduced through training data, design defaults, and corporate narratives of neutrality. They also negotiate queer legibility by refusing imposed categories, pointing to the risks of algorithmic inference, and questioning the imagined user that platforms often preconfigure. In addition, they practice queer infrastructural refusal through everyday tactics that include deleting traces, blocking, diversifying platforms, and reimagining alternative models such as a “community's algorithm.” This paper contributes to critical algorithm studies and queer theory by amplifying the voices of LGBTQ + professionals, highlighting how queering algorithmic imaginaries can contribute to ongoing debates on platform governance and inclusivity.
I present a critical analysis of ‘Digital Public Infrastructure’ (DPI). Through the example of India's DPI, also called the India Stack, I argue that the promotion of DPI contributes to the expansion of technological dependence on Big Tech companies’ digital infrastructure, transforming not only citizens but also states into their customers. To understand why such a contradictory policy has become so popular, I trace the origin of the term to Bill Gates and illustrate how his foundation, the World Bank, the United Nations Development Programme and other related institutions describe DPI as foundational digital systems that play a key role in the fulfilment of the UN sustainable development goals and, ultimately, development itself. I situate this discourse within broader narratives that promote technological solutions to structural economic and political hardships, particularly underdevelopment. I then build on Gramsci's notion of cultural hegemony and the literature on philanthro-capitalism to deconstruct the meanings of ‘public’ and ‘infrastructure’ promoted by the Gates Foundations and its partnering international organizations. This commentary concludes by underscoring the consequences of the DPI rhetoric for peripheral and semi-peripheral states’ governance.
Artificial Intelligence (AI) is hegemonically framed through the language of efficiency and scale, conditioning futures through trajectories of profit and control. This framing obscures that AI is inscribed by regimes of power, rooted in extractive colonial and racialized histories. This article explores an alternative trajectory: approaching AI not as an inevitable technology of progress driven by corporate actors, but as a political formation whose groundings and sociotechnical imaginaries are open to challenge. We invoke Afrofuturism in conjunction with speculative design as a critical methodology to unsettle dominant techno-futures. This is based on a collaborative project involving university researchers and BLAST Fest, a Black-led community organisation in the UK. Through a series of co-designed group activities and participatory workshops, AI was interrogated with speculative tools to activate other possibilities. Moving away from common tropes of inclusion and fairness in existing systems of AI, the project developed four critical orientations: equity; epistemic justice; re-purposing technology and alternative timelines for reimagining AI. These orientations trace how project collaborators questioned AI's assumptions, contested its dominant imaginaries and explored what might be refused or rebuilt from minoritised standpoints. We propose that Rebooting AI is a collective counter-imaginary practice: a way of developing critical AI literacy as collective agency, opening space for imagining technological futures otherwise.
Seeking information via social media instead of search engines has become increasingly common. This study examines RedNote, a popular Chinese platform that has emerged as a mobile search tool. Arguing that RedNote’s abundance of useful content plays a key role, I propose the concept of “manufactured usefulness” to capture the constructedness of such content. Individual users, the platform, and, increasingly, advertisers have co-constructed a culture of usefulness, embodied in two prominent content genres: instructional and recommendation posts. Instructional posts teach users the skills they want to learn and help them solve immediate problems, while recommendation posts draw attention to things that people may want to use, experience, or own. Useful posts are expected to possess qualities such as authenticity and communicative efficiency, which further break down into factors (e.g., argumentation, discursive style, formatting, etc.) that can be consciously assembled and employed. Thus, usefulness can be manufactured. RedNote monetizes manufactured usefulness by selling “useful” posts to advertisers, but ubiquitous sponsorships raise users’ doubts about authenticity. Paradoxically, RedNote upholds the banner of authenticity to retain users and further monetize this value, with its underlying commercial logic being both restrictive and productive in shaping the culture of usefulness.
This paper introduces YSocial, a new-generation virtual twin designed to replicate an online social media platform. Digital and virtual twins are virtual replicas of physical systems that allow for advanced analyses, control, experimentation, and scenario simulations. In the case of social media, a virtual twin such as YSocial provides a powerful tool for researchers to simulate and understand complex online interactions. YSocial leverages state-of-the-art large language models (LLMs) to replicate sophisticated agent behaviors, enabling accurate simulations of user interactions, content dissemination, and network dynamics. By integrating these aspects, YSocial offers valuable insights into user engagement, information spread, and the impact of platform policies. Moreover, integrating LLMs allows YSocial to generate nuanced textual content and predict user responses, facilitating the study of emergent phenomena in online environments. To better characterize YSocial, we describe the rationale behind its implementation, provide examples of the analyses that can be performed on the data it generates, and discuss its relevance for multidisciplinary research.
The rapid advancement of Generative AI, particularly large language models (LLMs), has sparked an extensive debate regarding the use of synthetic data in social research. Beyond the profound epistemological implications of this possibility, the first prerequisite for its feasibility is to test if such models adjust their responses to capture the vast individual and social diversity found in human populations and, if so, to what extent their outputs are comparable to those observed in human samples. This study investigates whether LLM-generated synthetic personas can accurately express basic psychological needs and human values considering a wide range of distinct individual and social positions. Additionally, to assess whether the model can maintain such consistency beyond information available within the scientific community, we propose a novel task in which the model cannot rely on prior knowledge. The findings suggest that a language model, such as GPT-4o , embodies a set of implicit theories about how a human with specific characteristics would respond to a questionnaire on basic needs and values, and it can apply these theories with sufficient consistency both when responding to an established scale and to a newly developed one. We further examine how certain aspects of our results align with key limitations identified in the critical literature on the use of synthetic data.
Data-intensive health research requires robust infrastructure, encompassing the material, social, and institutional structures that facilitate large-scale data storage, sharing, and linkage. Such infrastructures are a central element of the research process, offering opportunities to integrate ethical considerations and promote responsible research practices. We analyze different ways of embedding ethics in infrastructures for data-intensive health research and their consequences for the allocation of responsibilities: guidelines and policies aimed at (i) individual actors like researchers; (ii) social arrangements, including committee review and data stewardship; and (iii) material elements and technological measures such as encryption. We argue that different approaches to embedding ethical considerations influence the behavior and latitude of involved actors, which moral competences are required, and how responsibilities are distributed among actors. Infrastructures play a crucial role in allocating responsibilities for addressing ethical concerns and preventing or mitigating undesirable outcomes, and thus in fostering responsible research practices. However, the benefits of specific ways of embedding ethics in infrastructures must be weighed against potential downsides, such as increased regulatory burdens and bureaucracy, and potential impacts on researchers’ moral competences.
Despite a growing interest in how prevailing influences of algorithmic systems are being resisted, debates about algorithmic resistance unfold within conceptually diverse but disconnected scholarly conversations. Adopting a new materialist ontology, this integrative review traces algorithmic resistance's multiple conceptualizations across a growing interdisciplinary landscape. Academic publications are understood and analyzed as the outputs of research-machines : assemblages of theories, methods, technologies, researchers, disciplinary norms, and institutional logics that collectively produce particular visions of algorithmic resistance (while muting others). Reviewing 106 items, this study analyzes how diverging understandings of algorithmic resistance and its properties are territorialized within particular problem spaces: configurations of agencies, problematizations, and locations. Seven contrasting clusters of research-machines are identified (e.g.,Algorithm Aversion, Mundane Opposition, or Epistemic and Ontological Refusal) according to their shared productions of algorithmic resistance. Properties of these outputs are understood along six essential axes (intentionality, scale, visibility, materiality, temporality, relationality) and arranged into a provisional topography of intensities, silences, overlaps, and tensions within current scholarship on algorithmic resistance. This review offers two principal contributions: First, it provides an integrative perspective on the fragmented landscape of interdisciplinary scholarship that reveals how algorithmic resistances are produced within specific problem spaces and unfold along a topography of properties. Second, it advances a conceptual understanding of research-machines that sensitizes toward knowledge practices as sites at which the conditions and capacities of resistance to algorithmic power are constituted and configured.
The regulation of artificial intelligence (AI) is a prominent issue in both the European Union (EU) and the United States of America, but with distinct approaches to the governance of this rapidly evolving field. The EU has developed a comprehensive regulatory framework, notably through its 2024 AI Act. Meanwhile, the United States has gradually developed a piecemeal and voluntary federal approach, often driven by principles of self-regulation and sector-specific non-binding guidelines. But, state-by-state, the United States has shown intense regulatory activity of a binding nature. We approach this variation by mobilizing the concepts of policy style and regulatory style. We find that the EU seeks to anticipate the evolution of AI risks and impose a comprehensive architecture of rules in ways that make adaptive regulation hard to develop. However, there are issues with the enforceability of the AI Act and its capacity to protect human rights and public values. The United States intervenes legislatively when specific risks are clearly present, or sector-by-sector. Overall, this approach is more stringent in practice, as it is more enforceable than the EU AI Act, also in relation to the public interest. The state-by-state and sector-by-sector evolution of rules provides more space for learning from experience, imitation, and diffusion, in line with the features of laboratory federalism. By contrast, the EU space for learning in adaptive ways is more limited.
This study examines the challenges that researchers faced while accessing, interpreting, and assessing data accessed via Meta's now-sunset CrowdTangle (CT) application programming interface (API). Drawing on interviews with 20 academic users, we introduce the framework of Seeing like an API to illustrate how CT's technical design, governance rules, and platform-mediated knowledge networks shaped what researchers could see, ask, and conclude about platform activity. Participants adapted to limitations such as follower-count thresholds, the exclusion of comment data, and opaque content removal through three recurring strategies: optimizing the API's capabilities, extending data visibility with supplementary methods, and verifying outputs against external sources. These adaptations were developed in response to ongoing uncertainty about data completeness and provenance, which constrained both the scope of feasible research questions and the reliability of resulting analyses. Our findings show how platform-controlled data infrastructures shape research design, collaboration, and epistemic norms, and they inform emerging models for independent, third-party-regulated data access, such as those envisioned in the European Union's Digital Services Act.
Digital video advertising has become the world's largest platform-economized form of visibility, yet it remains strikingly opaque to viewers and underexamined by social scientists. This paper examines how digital video advertising economies are produced and maintained across platforms such as TikTok, YouTube, Meta, and Netflix. Drawing on a multi-method study, it analyzes how impressions are assembled, stabilized, and monetized within the computational supply chains of digital video. The paper argues that video ad platforms are not mere multi-sided markets but stacked economization processes. At their base lies barter, through which viewers exchange attention, behavioral data, and cognitive labor for access and entertainment. To specify the objects exchanged in these asymmetric relations, the paper introduces the concept of the dyad: paired goods that supply the raw materials of digital video advertising markets. Building on this foundation, marketization processes transform fleeting attention into measurable, comparable, and tradable impressions through technical rules, visibility thresholds, and device-level constraints. The paper concludes that digital video advertising is best understood as a set of stacked-economization platforms organizing contemporary visibility, with significant regulatory and environmental consequences.
This article explores how the rise of Go AI systems has redefined the expertise and professional roles of human players in the Go community. Historically, professional Go players held jurisdiction over two key tasks: demonstrating ideal gameplay and interpreting games, which are all challenged by the arrival of Go AI since 2016. Drawing on Andrew Abbott's theory of jurisdiction and Gil Eyal's network model of expertise, this article argues that the integration of Go AI did not fully replace human Go experts, but reconfigured their role from knowledge authorities to interpreters of AI outcomes. The scope and content of Go professional's work have changed in response to Go AI's challenge. Meanwhile, the sociotechnical networks that sustain Go AI are selectively displayed to enable the performance of expertise regarding two tasks. The insights from Go community will serve as an early case for other realm where AI exceeds the capabilities of human experts.
In the context of unprecedented U.S. Department of Defense (DoD) budgets, this paper examines the recent history of DoD funding for academic research in algorithmically based warfighting. We draw from a corpus of DoD grant solicitations from 2007 to 2023, focusing on those addressed to researchers in the field of artificial intelligence (AI). Considering the implications of DoD funding for academic research, the paper proceeds through three analytic sections. In the first, we build on past scholarship to offer a critical examination of the distinction between basic and applied research, showing how funding calls framed as basic research nonetheless enlist researchers in a warfighting agenda. In the second, we offer a diachronic analysis of the corpus, showing how a ‘one small problem’ caveat, in which affirmation of progress in military technologies is qualified by acknowledgement of outstanding problems, becomes justification for additional investments in research. We close with an analysis of DoD aspirations based on a subset of Defense Advanced Research Projects Agency grant solicitations for the use of AI in battlefield applications. Taken together, we argue that grant solicitations work as a vehicle for the mutual enlistment of DoD funding agencies and the academic AI research community in setting research agendas. The trope of basic research in this context offers shelter from significant moral questions that military applications of one's research would raise, by obscuring the connections that implicate researchers in U.S. militarism despite the pervasive presence of those connections in funding documents.
This article develops the concept of the algorithmic other to explain how artificial intelligence systems operationalize structural dehumanization through infrastructures of classification. Moving beyond models of bias, fairness, or representation, the article argues that algorithmic governance produces differential personhood by translating historically constituted social hierarchies into computational logics of legibility, prediction, and control. Drawing on Black feminist science studies, racial capitalism, Indigenous data sovereignty, and crip theory, the article reconceptualizes dehumanization not as an affective failure or design flaw but as an infrastructural condition embedded within data capitalism. Using a critical theoretical synthesis and comparative case analysis of facial recognition, predictive policing, and credit scoring, the article identifies three interlocking mechanisms—misclassification, hypervisibility, and erasure—through which algorithmic systems fabricate and manage social difference. These mechanisms operate relationally: misclassification produces predictive illegibility; hypervisibility territorializes exposure as governance; and erasure structures differential inclusion through systems of ranking, visibility, and exclusion. Together, they constitute a form of computational hierarchy that organizes recognition, risk, and disposability at scale. By theorizing dehumanization as procedural and infrastructural rather than psychological or interpersonal, the framework extends existing approaches to algorithmic bias, surveillance capitalism, and data colonialism. The article concludes by outlining a research agenda for studying algorithmic power through ethnography, transnational comparison, and participatory design while foregrounding the need to reconstruct data infrastructures toward epistemic justice rather than merely reform their outputs.
A major challenge for sustainable agriculture is finding alternatives to herbicides, which contribute to biodiversity loss and health risks. While digitalization is often presented as the solution to "the weeding problem," this article untangles how digitalization is changing human-machine-nature interactions in the case of digital weeding technology. Building on critical theory of technology and Science and Technology Studies (STS), we analyze the histories of these technologies and their grounded interactions with nature and humans. Our findings challenge dominant narratives of "Big AgTech" by revealing that digital weeding is largely driven by mid-sized family corporations and specialized startups, creating a distinct mechanical-digital hybridity. We demonstrate that "precision" is not a stable technical property but an emergent outcome of interactions between mechanical tools, nature's materiality, and local farming contexts. Furthermore, while these technologies may create new dependencies on proprietary software and mapping services, they simultaneously challenge findings on "digital Taylorism" by potentially improving work quality for tech-savvy employees. Moreover, we uncover affective dimensions of human-robot relations-farmers naming robots and treating them as animal-like co-workers-that shape technology adoption in ways overlooked by current literature. Ultimately, we argue that understanding agricultural digitalization requires moving beyond data-extractive models to account for the messy, affective, and ecological entanglements of working farms.
Understood as the practices that seek to centre and empower individuals and communities in the collection, usage and sharing of data, the concept of participatory data stewardship (PDS) is often praised for its potential to challenge datafication and facilitate public involvement in decision-making processes mediated by and about data. However, little is known about the role of data literacies – the skills, knowledge and practices involved in accessing and using data both practically and, more critically, with a view to civic action and social change – within PDS. Based on semi-structured interviews with community researchers (CRs) taking part in a PDS project conducted in Widnes, a town in the UK, this article examines the importance of developing CRs’ data literacies in the context of their involvement in the project. Key findings suggest that, whilst CRs recognised significant gaps in their data literacies, with data often being referred to as an abstract and obscure concept, they had both strong motivations to better understand data and expectations for how this may be used to improve their community. Bridging media literacy research on critical data literacies with PDS research, this paper argues that, if we are to expect PDS to potentially empower communities in a datafied society, then members of these communities need to be supported to develop their data literacies. Implications for research, practice and policy are discussed.
Scholars increasingly warn that commercial AI products reproduce narrow, stereotypical gender identities, but far less is known about how those identities are made in practice. This article addresses that gap through the case of AI streamers in China's expanding live-commerce sector, where generative-AI streamers are built to appear hyper-feminine and sell products as stand-ins for human streamers. Drawing on 120 h of behind-the-scenes ethnography in two Chinese AI startups and 48 interviews with engineers, designers, and brand marketers, I show that an AI streamer's gender is produced, not merely reflected, through an optimization loop : a recursive, metric-driven cycle in which developers generate, refine, and scale the variant that meets commercial goals. Pre-launch, teams translate abstract brand ideals into parameters across voice, face, gaze, gesture, and script. Post-launch, continuous A/B tests link these parameters to performance metrics (retention, click-through rate, sales per minute). Exposure is reallocated to higher-performing variants, and the winners are written back into the product as defaults. Across cycles, data do not simply register a pre-given persona. They select and lock in a gendered one, yielding a soft-spoken femininity optimized for sales. This article extends bias-reproduction accounts by unpacking the production pipeline of AI products and showing that user feedback is not a mirror of preexisting bias but a design lever teams use to reverse-engineer persona traits. This reframing shifts accountability from “bad data” to human choices and makes clear that identities in AI are engineered, traceable, and therefore contestable.
There is an ongoing shift in the Swedish welfare state, characterised by an increased datafication. The increasing reliance on data as a means of predicting citizens’ behaviour not only transforms the way the welfare state is organised, but also the very notion of welfare. As a response to a growing discussion on inaccurate welfare payments in Sweden, the Swedish Payments Agency (UBM) was established. The authority will make use of system-wide data analysis in order to prevent, forestall and detect inaccurate welfare payments. As such, this paper approaches the authority as an automated welfare surveillance scheme. Researchers have repeatedly emphasised how the use of computational methodologies in welfare surveillance negatively affects marginalised citizens in a disproportional manner. Consequently, effort has been put into achieving more ethical and just outcomes. However, it has also been argued that instead, more emphasis should be placed on the underlying politics of welfare surveillance technologies. In order to make the implicit problem representations underlying the establishment of the Swedish Payments Agency visible, this paper contributes insights to the implicit assumptions made within and about the datafied welfare state. This paper draws on the What's the Problem Represented to be(WPR) approach. By examining policy documents related to the establishment of the UBM, this paper identifies three representations of the ‘problem’, namely (a) inaccurate welfare payments, (b) lack of control and (c) data silos. Accordingly, the problem representations are interpreted as concerning issues of control and the lack thereof.
This commentary introduces the concept of platformed migrant to theorize how digital platforms fundamentally reshape migration experiences through processes of datafication and platformization. Drawing from an intersectional feminist perspective, it examines the paradox whereby platforms simultaneously offer migrants unprecedented opportunities for self-representation and expose them to intensified surveillance and control. The platformed migrant emerges at the intersection of platform capitalism, state surveillance, and algorithmic governance, where social media presence becomes both necessity and liability. The commentary reveals how platforms function not merely as communication tools but as border infrastructure itself, actively (re)producing migration conditions by determining who can move and under what conditions. Understanding the platformed migrant becomes crucial as platforms increasingly align with authoritarian politics, transforming digital presence into a contested site where contemporary migration is lived, narrated, governed, and resisted.
