The classical principal-agent problem assumes a single, shared model of the world: the principal delegates a task, the agent takes an action, and a payment is made based on the observed outcome. The standard solution aligns incentives through a contract that accounts for the agent's effort and the stochastic link between effort and outcome.
A new arXiv preprint, "Contracting with an Agent under Model Mismatch," relaxes a key assumption. It considers what happens when the principal and the agent do not share the same model of the environment — for example, they may disagree on the probability of success given the agent's effort, or on how much the outcome benefits the principal.
The abstract is brief, but the direction is clear: model mismatch can alter the optimal contract, potentially requiring different incentive intensities or payment structures than the classical solution. The paper appears to provide a general framework for studying how such disagreements affect contract design, rather than focusing on a single application.
As a preprint, the work has not yet been peer-reviewed, and the abstract leaves many details unspecified. Still, it points to a realistic and underexplored dimension of contracting — one where parties bring their own beliefs to the table, and the contract must work despite, or because of, that divergence.