AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › People & Organizations

Henry Farrell

medium confidence · updated 2026-06-06

SAIS / Johns Hopkins political scientist working at the intersection of political economy, information technology, and democracy. Co-author of AI as Social Technology (Knight Columbia, May 2026) with entities/cosma-shalizi, the framework treating LLMs as social technologies in the Long Industrial Revolution lineage. Co-PI on the broader AI as Cultural Technology research program with Alison Gopnik, James Evans, and Shalizi.

Contents

Henry Farrell is a political scientist working on the political economy of information, technology, and democracy. He is co-author, with Cosma Rohilla Shalizi, of AI as Social Technology (Knight Columbia, May 11, 2026), a framework that treats large language models as social technologies in the lineage of markets, bureaucracies, and the price mechanism.

Background

Farrell is SNF Agora Professor of International Affairs at the Johns Hopkins School of Advanced International Studies (Sais) and an SNF Agora Institute affiliate. He took his Ph.D. in Government at Georgetown in 2000 and was a longtime political-science professor at George Washington University before moving to Johns Hopkins in 2021 to take the SNF Agora chair.

He co-founded the blog Crooked Timber (2003–) and was an editor of The Monkey Cage at The Washington Post (2010–2018). He contributes to Foreign Affairs, Aeon, The Guardian, and his Substack, Programmable Mutter.

AI as Social Technology

Farrell co-authored AI as Social Technology with Cosma Rohilla Shalizi (Knight Columbia, May 11, 2026; see AI as Social Technology and AI as Social Technology (Farrell + Shalizi, Knight Columbia, May 11 2026)). The framework treats LLMs as social technologies — systematic means of reorganizing social relationships — continuous with markets, bureaucracies, democracy, and the price mechanism.

Central to the essay is the Long Industrial Revolution framing: the centuries-long transformation of human capacities to organize economic, social, and political life via lossy coarse-grainings such as bureaucracies, markets, opinion surveys, print capitalism, and now AI. The framing re-anchors AI debate on the question of what kind of social technology AI is, rather than on capability benchmarks or alignment frontiers.

The work sits alongside three related framework programs. AI as Cultural Technology (Farrell-Gopnik-Evans-Shalizi 2025) is Farrell's own earlier and broader research program with Alison Gopnik (UC Berkeley developmental psychologist), James Evans (Chicago sociologist), and Shalizi. AI as Normal Technology (Sayash Kapoor and Arvind Narayanan, Knight Columbia 2025) is a companion framework emphasizing diffusion lag and deployment friction rather than reorganization of social technologies. Sociotechnical Risk Governance (Deirdre K. Mulligan, Nik Marda, and Victor Zhenyi Wang, Knight Columbia 2026) is a companion framework on how risk-governance institutions reach AI.

Intellectual lineage

The AI as Social Technology essay locates its argument inside a specific lineage of thinkers, which Farrell and Shalizi treat as a reading list for the framework:

  • Herbert A. Simon — bounded rationality; the Logic Theorist; the analytic move that treats organizations and markets as information-processing devices.
  • James C Scott (page planned) — Seeing Like a State (1998); legibility, simplification, and the coarse-grainings that state-like institutions impose (Seeing Like a State (applied to AI) planned).
  • Elinor Ostrom (page planned) — governing the commons; institutional design as a substitute for both market and hierarchy.
  • Joseph Schumpeter — creative destruction.
  • John Dewey — democracy as experimental inquiry (the Simon-Dewey synthesis Farrell-Han 2025 elaborates).

These names identify who sits on the political-economy side of the AI-as-social-technology debate, whether or not each is separately page-worthy.

Major works

  • Farrell, Gopnik, Evans, Shalizi (2025) — Large AI models are cultural and social technologies, Science 380 (commentary). The foundational essay for the framework cluster.
  • Farrell and Shalizi (2026-05) — AI as Social Technology, Knight Columbia, ingested as AI as Social Technology (Farrell + Shalizi, Knight Columbia, May 11 2026).
  • Farrell and Fourcade (2023) — "The Moral Economy of High-Tech Modernism," Daedalus. The high-tech-modernism thesis holds that platforms impose simplified representations on social life, with consequences Farrell frames as analogous to Scott's earlier critique of state legibility.
  • Farrell (2025) — Democracy as a Social Technology — argues democracy is itself an information-processing institution and that AI's reorganization of information has direct democratic-stability implications.
  • Farrell and Han (2025) — Democracy and the Simon-Dewey Synthesis, joint with political scientist Hahrie Han; the political-experimentation reading of democratic institutions.
  • Farrell (2024) — Programmable Mutter Substack — the public-facing channel, cited for shorter-form positions not in the academic record ((Source: programmablemutter.substack.com/...)).

Recurring positions

Across the published record, four positions recur in Farrell's work:

  1. AI is new in scale, not in kind. Farrell argues that LLMs are doing what bureaucracies, markets, and the price mechanism have always done — coarse-graining human knowledge into operationalizable signals — but at a new scale and via a new substrate.
  2. The political-economy reading should precede the alignment reading. Where the alignment community asks how to keep models aligned with human values, Farrell asks what social institutions a technology reorganizes, and at whose expense.
  3. The democracy-AI relationship is not primarily about deepfakes. Farrell argues it is about how AI reshapes the information-processing infrastructure on which democratic deliberation depends, with Farrell-Han 2025 as the clearest statement of this view.
  4. Industrial-revolution analogies are useful but lossy. Farrell adopts the Long Industrial Revolution framing while flagging that AI's reorganization of cognition, not just labor, makes the analogy partial.

Relationships

Sources

Categories: People & Organizations | academic | framework | ethics | frontier-models | political-economy | democracy
What links here (15)