Not yet human-reviewed

This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.

Knowing Enough to Be Dangerous: The Problem of “Artificial Certainty” for Expert Authority When Using AI for Decision Making and Planning

Citation. Leonardi, P. M. & Leavell, V. (2026). Knowing Enough to Be Dangerous: The Problem of “Artificial Certainty” for Expert Authority When Using AI for Decision Making and Planning. Organization Science, 37(2), 516–543. https://doi.org/10.1287/orsc.2023.18224

Summary. This comparative ethnographic study examines how two urban planning organizations used the same AI-powered simulation system while developing long-range regional plans. It theorizes artificial certainty: AI representations can make uncertain futures look definitively knowable, while process experts can preserve useful uncertainty by modulating how outputs are scoped, presented, and interpreted.

Key claims

• Artificial certainty is not an inherent property of an AI tool; it emerges through organizational practices that make AI-generated representations appear more detailed, definitive, and technologically authoritative than their underlying uncertainty warrants.
• At the site that enhanced the model’s detail, stakeholder immersion, and technological prominence, stakeholders increasingly treated model outputs as predictions and the planners’ interpretive authority weakened.
• At the site that modulated the model’s scope, stakeholder engagement, and meaning, planners retained the model as one exploratory input and preserved their authority to keep uncertainty visible in decision making.
• The study distinguishes representations of the future from representations for the future: treating outputs as conditional aids to deliberation rather than forecasts to implement changes the role of both expertise and AI in planning.

Evidence & limitations

• Comparative ethnographic study of two U.S. metropolitan planning organizations using UrbanSim, drawing on 110 semistructured interviews, observations, and internal and public documents.
• The mechanism is developed from a shared AI simulation tool in regional planning; its transfer to other expert settings, model types, and organizational decision processes requires further evidence.

Feeds

• expert-authority-in-ai-decisions
• Proposed new topic: artificial-certainty — how AI representations create an illusion of predictability and how organizations preserve uncertainty for accountable deliberation.