Search on Rugged Landscapes: An Experimental Study

Search is fundamental for understanding human decision-making. Recent theoretical and empirical contributions in management have drawn on the NK model of rugged performance landscapes to examine search in complex settings. This paper reports findings from an experiment on search behavior on rugged performance landscapes. The results from our experiment confirm basic predictions of the NK model about search performance. However, our results point to a different set of behavioral rules governing search behavior. Human subjects adapt their search behavior to problem complexity. Our results suggest that the properties of performance landscapes systematically influence feedback conditions and thereby actual search behavior. What is more, the cost of exploring new alternatives systematically leads to the stopping of search before reaching a local peak. In sum, while we find broad support for the predictions of the model, our findings also point to an alternative causal mechanism to connect the properties of complex problems to search behavior and performance.