Strategy

Speaker: Carolyn Fu: Harvard Business School

Time: 12:00 PM - 1:30 PM

Location:

Speaker: Nicolaj Siggelkow: Wharton

Time: 12:00 PM - 1:30 PM

Location:

This paper investigates the performance consequences of how managers update their contingency beliefs in response to feedback, emphasizing the interaction between first-order learning (updating performance expectations) and second-order learning (updating contingency beliefs). Managers often begin tasks with inaccurate contingency assumptions—either overly simple or overly complex—which shape their initial decisions. Our research explores whether starting with simple or complex contingency beliefs is more advantageous when learning occurs in environments characterized by varying degrees of true contingencies. Employing a contextual multi-arm bandit simulation, we reveal nuanced learning dynamics: without second-order learning, complex initial beliefs consistently outperform simpler ones. However, when second-order learning is enabled, initially simple beliefs significantly enhance managerial decision-making over time, even potentially surpassing the performance of managers with initially accurate beliefs. Conversely, those with overly complex initial beliefs demonstrate limited improvement through second-order learning alone. Importantly, this pattern changes substantially when managers have considerable prior information about action outcomes across contexts; under these conditions, complex initial beliefs become advantageous. Thus, the benefits of simplicity versus complexity in contingency beliefs critically depend on the interplay between first- and second-order learning and the availability of prior information, offering actionable insights for managerial decision-making in uncertain environments.