Working Paper Abstract
Unbundling Formation from Performance: An Institutional Theory of Artificial Intelligence and Human Capability
Abstract
This paper develops an institutional theory of artificial intelligence as a force that can unbundle visible performance from human formation. Many organizations historically joined output and formation because the two were difficult to separate: people learned judgment, responsibility, attachment, and tacit skill by performing the work itself. Generative AI lowers the cost of separation. It can deliver plausible performance while bypassing the human process through which capability was reproduced.
The paper locates the mechanism in an asymmetry between private and collective incentives. Unbundling is often cheap, immediate, and privately beneficial, while rebuilding formation is costly, delayed, and institutionally collective. A simple model shows why the danger is uneven. Unbundling becomes more likely when immediate gains are high, maintenance costs are high, and decision-makers have short internalized horizons. Conditional on unbundling, social formation loss rises with the strength of the original bundle. The highest-risk cases lie near the preservation threshold, where formative value is large but insufficiently internalized to stop separation. The contribution is a positive theory of where AI hollows institutions, where it does not, and when formation must be rebuilt deliberately.
Keywords: artificial intelligence, institutions, formation, judgment, apprenticeship, collective action