What AI Agents Carry into the Loop: Network Positions of Production and the Reconfiguration of Epistemic Cultures

Society for Social Studies of Science (4S), Toronto · accepted · October 2026

Panel: Imaginaries of AI in Collaboration: Expertise and Authority in AI-Mediated Scientific Practice — a 20-minute paper talk on Wednesday, October 7, 2026.

In plain terms: an AI agent is built somewhere in particular, carries the marks of that place into whatever work it joins, and arrives without the community of practice that made it — so which AI enters a collaboration helps decide what the collaboration becomes.

Nonhuman Attentional Control: Organizational Import of External Institutional Logics

Academy of Management Annual Meeting 2026, OMT Division · Academy of Management Proceedings 2026(1) · July–August 2026 · doi.org/10.5465/amproc.2026.18749abstract

Accepted March 2026. Presented as a pre-recorded talk in Session 22935 — the session number is the conference program's, not a citation; the Proceedings DOI above is the citable record.

In plain terms: given the same organizational decisions, five AI models from different institutional origins recommended systematically different courses of action, while each stayed consistent with itself across repeated runs.

Watch the pre-recorded talk.

The Institutional Origins of AI: Evidence from Large Language Models

SSRN working paper 6749643 · May 2026

Two studies: five frontier-lab models across 25 organizational decision scenarios (3,125 observations), and nine models across six open-ended knowledge-production tasks (162 outputs) — institutional signatures reach into what models treat as evidence and explanation. The paper does not claim that production origin rather than co-varying capability causes the differences; that is what the registered, capability-matched study tests.

This work has evolved into a pre-registered study now underway.

For preprints, replication materials, or collaboration inquiries, write to [email protected].