Algorithm Effects on Performance and Human Capital Development

Abstract

Organizations increasingly deploy algorithms and AI to guide human decision-makers to improve performance. While such technologies can improve productivity, their overall impact depends both on human-AI interactions that directly shape the integrated performance of human-AI systems and indirectly through their effects on human skill acquisition and retention. We study design choices regarding algorithmic advice through a principal-agent lens, where an algorithm designer (principal) determines the signals a user (agent) receives. In this talk, we will examine the impact of two key dimensions of these decisions: (i) the level of detail of information conveyed in AI signals, and (ii) the degree of user control over their timing and frequency. Using experimental evidence, we show how these decisions shape performance and human learning.