Scholarship on science has long contained a tension between accounts emphasizing convergent forces and those highlighting divergent practices. This article addresses that tension through an ethnographic study of data science. It advances the concept of epistemic idiom —a recognizable and portable pattern of practices that evolve alongside knowledge-making tools, analogous to musical idioms arising from the interaction of material affordances and learned techniques. Through analysis of a project attempting to sequence chromosomal centromeres, I show participants enacting the nomad idiom of data science: a data-led approach involving tools for iteration and exploration, cross-disciplinary borrowing, bricolage construction of bespoke assemblages, and the motivation to scale. Though the project failed to sequence centromeres, it generated unexpected discoveries, challenged established theories, and yielded lasting changes to scientific practice. The case reveals how epistemic trends can spread without being totalizing—the nomad idiom enabled both convergence (shared technical toolkit) and divergence (local adaptations, surprise outcomes) simultaneously. The epistemic idiom concept complements existing STS frameworks by explaining how practices can migrate across epistemic communities, while those communities retain distinct cultures. It speaks to concerns over computational universalism and epistemic monocultures, demonstrating that alternatives exist—ways for scientific trends to be borrowed rather than imposed, becoming remixes instead of replacements.
The use of AI-based technologies is reshaping mathematics in ways that extend but also strain the classic Big Science analogy. As in Big Science, the new AI-based mathematical work relies on concentrated computation, specialized and hierarchical teams, sociopolitical justification, and large-scale collaboration. But AI also changes how epistemic labor is distributed and attributed: Systems now generate constructions, explore proof spaces, and formalize arguments, prompting mathematicians to renegotiate how credit, authority, and accountability are assigned across hybrid human–machine assemblages. This reconfiguration redistributes epistemic authority toward corporate labs and elite hubs that control models, data, compute, and narrative, thereby gating which problems appear tractable and legitimate. Analyzing four dimensions (resource concentration, division of labor, political embedding, and collaboration), and examining recent cases, we argue that AI-based mathematics constitutes a novel sociotechnical regime we call ‘Big AI Mathematics’, in which truth claims are increasingly mediated by opaque pipelines and institutional power, making the future of mathematical knowledge as much a governance problem as an epistemic one.
This paper draws on two affectively charged fieldwork encounters in a Swedish daily activity centre for persons with profound intellectual and multiple disabilities (PIMD) to explore how affect intervenes in fieldwork and reconfigures what becomes knowable. Situated within Science and Technology Studies (STS), the analysis reflects on how moments of discomfort, intimacy, and apprehension unsettle a research orientation initially shaped by material semiotics, prompting renewed attention to questions of complicity and the epistemic terms of practice-oriented analysis. Rather than treating affect as noise or private response, it is approached as a register through which broader histories of segregation and exclusion become palpable in the encounter. Tracing how affect shapes what is registered as relevant or knowable, the paper develops epistemic adhesion as a conceptual companion to disconcertment—understood here as moments of felt epistemic disruption—to name the ease or alignment through which dominant orientations are affirmed rather than troubled. In attending to how disability is rendered affectively and epistemically available, the analysis offers a situated reflection on the stakes of knowing—and not knowing—with PIMD.
Natural Language Processing (NLP) technologies—ranging from topic models to today’s large language models like GPT—have rapidly entered the social sciences, reshaping methodological practice. Yet researchers often overlook the stark political-economic contrasts between academia and the AI research-industry symbiosis. Identical algorithms, once embedded in different institutional settings, acquire different meanings and standards of evaluation. This paper shows the divergence by examining topic modeling, a classical NLP technique in computational social science. Social scientists grapple with the instability of applying topic models to the same corpus, whereas in the AI industry such variability matters little, given different evaluative priorities. Through a comparative analysis of topic modeling’s trajectory across AI and social science, I show how organizational contexts and goals shape the development of the same algorithms, and why framing instability as a purely technical issue is problematic in the social sciences. The findings reveal that algorithms are not simply technical tools but products of material political-economic regimes. Recognizing this, I argue that STS scholars have a vital role to play in computational social science: not only by critically examining and developing methods, but also by interrogating the material-political-economic regimes in which algorithms are enacted, and by working toward more just alternatives.
Ghostwritten medical papers are a cornerstone of pharmaceutical marketing. We trace how one manuscript written under the instructions and supervision of opioid producer Mallinckrodt was produced and eventually published in a medical journal to present a marketing message. We systematically analyse an archive of internal communications made public after litigation through the lens of Ricœur’s work on imaginaries. Along his first analytic axis, we show how the ghostwriting process shapes and reshapes identities, ascribing agency and portraying social order. Along the second, legitimacy was constructed by the expert voice of a peer-reviewed article ‘authored’ by researchers, but written by backstage medical writers. Third, the result was a twisted phantasmagory of both academic writing and pain management. Professional medical writers generated an imaginary of pain management to expand the prescription of opioids by redefining pain categories and reassigning agency and responsibility of actors in pain treatment. Meanwhile, backstage writers also reproduced an imaginary of scientific writing, with efficient writing as an editorial service that misrepresents authorial contribution and responsibility. We suggest that Ricœur’s analytic scheme could also be used to disentangle technoscientific imaginaries in other settings.
