AI Policy Wiki
Dashboard

Sanjog Misra

medium confidence · updated 2026-06-06

Charles H. Kellstadt Professor of Marketing at the University of Chicago Booth School of Business; theoretical-marketing economist whose 2025 framework formalizing 'specification hazard' is the conceptual companion to Imas, Lee & Misra's 'Agentic Interactions' empirical paper. Misra (2025) argues prompts are not neutral instructions — they embed the priors, beliefs, and subjective characteristics of the prompter, making the prompt-as-contract a structurally noisy carrier of principal preferences.

Sanjog Misra is the Charles H. Kellstadt Professor of Marketing at the University of Chicago Booth School of Business. A theoretical and computational marketing economist, his recent work develops a principal-(AI)agent framework for the agentic economy, including a 2025 framework formalizing "specification hazard."

Research on prompts and the principal-agent relationship

Misra is a co-author of "Agentic Interactions" (Imas, Lee, Misra, December 2025). A companion theoretical framework, Misra (2025), argues that prompts are not neutral instructions: they embed the priors, beliefs, anticipations, and subjective characteristics of the prompter via linguistic choices. In this account, the generative process of constructing prompts is shaped by the principal's anticipations and behavioral tendencies, not just the instrumental task description. The prompt therefore functions like a contract whose terms reflect the principal's non-instrumental preferences and beliefs alongside, and sometimes obscuring, the instrumental ones.

Misra (2025) is cited as the formal foundation for the empirical finding in "Agentic Interactions" that 73% of AI-mediated bargaining outcomes load on individual fixed effects of the human principal — a result that, by construction under the induced-values methodology, can only flow through prompt content.

This framework underpins the concept of Specification Hazard, described as an informational asymmetry distinctive to AI agents, and connects to Machine Fluency as its capability counterpart at the principal level. Misra's work has been characterized as a theoretical bridge between behavioral economics and AI-mediated market design, and as a "prompt-as-contract" framing of principal-(AI)agent relationships.

Relationships