Economics Research Seminar

“Structuring Search Into the Unknown"

We study a principal-agent problem where an agent searches for answers in a space of questions, modelled in the spirit of Carnehl and Schneider (2025) and Callander (2011). The principal can verify correct answers and knows how they compare to the “canonical” or expected answer, but lacks the expertise to put questions into context; the principal cannot tell if a question is close or far from existing knowledge. We show that, as a result of this lack of expertise, the principal distorts the degree of novelty of research downward, and pays for overly-surprising results. If the search technology differs for specialist and generalist researchers, the principal may allocate them to different tasks in order to economize on information rents, with generalists working on more novel questions.