Accepted & Forthcoming
Nonhuman Attentional Control: Organizational Import of External Institutional Logics
Academy of Management Annual Meeting 2026, Organization and Management Theory (OMT) Division. Philadelphia, PA. July 31 – August 4, 2026.
AI models carry institutional logics from their originating social systems. Given identical organizational scenarios, different AI models produce systematically divergent outputs because they have been socialized differently through their training. A 5-model × 25-scenario × 25-iteration experimental design shows 68% divergence in recommendations across models, while within-model consistency remains 91.4% when the same scenario is run repeatedly. AI attention is systematic, not random, and reflects the institutional logics of the model's origin.
Full paper and supplementary materials available on request: [email protected]
In Preparation
The Social Signatures of Artificial Intelligence
In preparation. Target venue: Humanities and Social Sciences Communications (Nature Portfolio) or PNAS Nexus.
Co-author invitations under discussion.
A four-study architecture testing whether 20+ AI models exhibit stable, distinguishable institutional signatures across behavioral, perceptual, deliberative, and ecosystem-level tasks. Extends the AOM 2026 result to a broader model population and a wider range of organizational contexts.
Design and protocol documentation in active preparation.
Earlier Work
Nonhuman Knowledge in Collaborative Knowledge Production
AOM TIM Paper Development Workshop, 2023. Technology & Innovation Management Division, Academy of Management.
Argues that large language models act as knowledge brokers in expert collaboration — overlapping with many specialists' domains simultaneously and bridging otherwise disconnected areas of expertise. Draws on the tertius gaudens framework from network theory to warn that nonhuman brokers can become political actors, turning knowledge brokerage into a form of power. An early articulation of themes — nonhuman agency, logics embedded through training — that the current program has since developed.
Future publications will be added as they are accepted or released. For preprints, replication materials, or collaboration inquiries, write to [email protected].