What Directs Search? Experimental Evidence on Some Triggers of Exploration

Abstract

Search is fundamental to managers and organizations, a cornerstone of the behavioral theory of the firm. But although theory and firm-level evidence rely on presumed individual mechanisms, these remain untested or are contested. Here we join the effort to uncover the micro-processes of search by identifying some conditions that encourage exploration. To do that, we design and conduct incentivized, randomized behavioral experiments, and replicate them across populations. We theorize that search can be reactive, a response to immediate negative experiences — but also proactive, guided by deliberation and anticipation. We begin by testing the prevalent assumption that exploration is a reaction to trailing aspirations. But then we demonstrate that exploration can be also prompted by stable performance (low variability). Moreover, performance stability (or variability) also interacts with the response to a negative aspiration–performance gap, seemingly changing how it is interpreted. We proceed to compare the observed human behavior to that of artificial intelligence (AI) agents. We find the populations behave similarly, starting with a broad exploratory search, which may not be immediately lucrative. The evidence, we propose, demonstrates that decision-makers can search proactively, not just in response to a failure or haphazardly. Rather, they act as if they use feedback to refine a mental model, a model that directs subsequent search.

Keywords: Exploration, exploitation, performance, feedback, individual, experiment