[SCALE]: Software, Community, Territory

Public discourse, much of it expressed on social media, diagnoses our era as both the smartest and dumbest in history. AI boosters celebrate artificial intelligence as a neutral solution to our crisis of truth. At the same time, many journalists, scientists, and liberal professionals double down on simplified calls for “just the facts,” overlooking the difficulty of reaching scientific consensus, even among credentialed experts. Both responses intersect today to revive, at scale, naïve notions of liberal epistemology: the belief that neutral technology or appeal to authoritative facts and experts alone can restore truth. Parts of media and communication studies have become equally rigid, casting technologies as inherently dystopian. All these views avoid the harder (and known) work of situated analysis. Drawing on feminist science studies, the history of science, and anthropology, I propose a framework of sturdy knowledge built on transparency, humility, and accountability, in which AI is neither a savior nor an inherent threat, but a pragmatic tool that can help users navigate and participate in knowledge production. I also argue that addressing this crisis calls for treating humanistic and social-scientific ways of knowing as indispensable and equal partners with the natural sciences.

Open source software communities build critical global infrastructure without the coercive power structures that define corporations and universities. No one can be fired, and yet these communities produce some of the most consequential technology in the world. Drawing on 15 years of experience in the Ruby and Rust programming language communities, this talk examines how culture scales in the absence of formal authority. Some communities thrive under strong centralized leadership until that leadership’s culture calcifies, driving contributors away and blinding the project to outside innovation. Others deliberately build open, inclusive cultures that hold for years, only for implicit power structures to overtake the explicit governance as the community grows. Today, new tools and new anxieties are fracturing these communities along surprising lines, stress-testing the social systems that open source depends on. The story of open source at scale is the story of what happens when you try to build community without coercion.

Enterprise AI service contracts — big tech contracts with large companies — make up 80% of Anthropic’s revenue. Though enterprise is only 20% of OpenAI’s revenue, the company’s current stock valuation has already priced in expectations of growth in corporate, educational, and government contracts, including military contracts. Nevertheless, most research and reporting frames AI’s users as private individuals related to their own personal chatbots by possessive pronouns: my AI, my chatbot, my agent. This possessive framing is encouraged and facilitated by platforms marketing customizable interfaces (Rocha 2025) and secondary markets that mediate between the user and the Big tech companies that own the underlying models (Whitaker et al 2023). This talk argues that framing “users” as “owners” erases the corporate contractual relations in which the prompt and the output meet, and that the confidentiality of the service contracts between corporate entities that underlie most enterprise AI use contributes to the erasure of these relations.

In studying the emergence of a field of nature study in early modern Japan, historians face four key challenges: first, accurately reconstructing the meanings and uses of scholars’ conceptual frameworks that mediated the entire enterprise of natural knowledge in the Tokugawa period; second, faithfully reconstructing how these scholars justified the authority of their cognitive claims on plants, animals, and ecosystems at the institutional, sociological, and epistemological levels; third, justifying these reconstructions historiographically despite the scarcity of documentary sources explaining and philosophically justifying Tokugawa naturalists’ cognitive practices; and fourth, reflecting on how historians’ own modern conceptual frameworks influence their reconstructions. These four challenges relate, respectively, to conceptual history, historical epistemology, historiography, and epistemology.

“Scale” plays a role in knowledge formation both in the context of the historical practices that historians aim to reconstruct and at the self-reflective level of their own cognitive claims. This paper explores how the idea of “scale” functions semiotically to create distance from the objects of study, making their manipulation possible.

This presentation considers the implications of using generative AI for high-stakes decisions (e.g., hiring, lending, admissions). While such decisions have been handled traditionally by predictive AI, there is growing interest in using generative AI for these purposes because it is easier to access, costs less to use, and requires little or no training data. In this presentation, I show, however, that decision-making based on generative AI lacks many of the qualities commonly valued about decision-making based on predictive AI: such decision-making can be more ad hoc, less consistent, and less insulated from irrelevant or even illegal factors. These developments challenge the commonly held view that algorithmic decision making is not only a way to automate — and thereby scale — decision making, but to ensure its regularity. As a result, many of the policies and tools for governing the use of AI for decision making will need to be reconsidered in the transition from predictive to generative AI.

Drawing on fieldwork with venture capital (VC) investors, startups, and their government partners in China, this talk examines how venture capitalism intersects with Chinese governance in the business-to-government (B2G) market — specifically, how startup companies partner with local governments to fulfill central state mandates. By reflecting on how participants see themselves as nodes within a larger algorithmic assemblage to generate policy, this talk considers the Chinese state’s latest iteration of adaptive and experimental governance. It asks how the impact and influence of AI technologies — such as machine learning — might extend beyond domain-specific applications, lending scientific authority to the Chinese state’s approaches for leveraging and managing the nation’s scale.

Both the Chinese and American state position AI as a “return” to a past of national strength. While the detailed articulations vary, each frames the other country’s approach to AI as a threat to its national rebuilding. Their political leaders rely on the production and circulation of feelings from fear of a threatening enemy to happiness attached to a strong nation, wholesome family values, and a “simpler” past. In this talk, I offer a theory of governance and power amidst a global embrace of AI that examines the enduring role of affect in the data-driven management of land, populations, and industries. I draw from in-depth ethnographic research on the uptake of AI to transform the countryside, agricultural industries, and various bureaucratic organizations in both China and America. I challenge the idea that the authoritarian state is capable only of coercive control enabled by an all-seeing machinery of vision technology. And I show that America’s and China’s approaches to AI have much in common, with AI functioning as a useful “emotional technology” for political regime formation.