Risk-averse algorithmic support in resource allocation problems

Abstract

We live in the age of algorithms. In this dissertation, I study how temporary exposure to risk-averse algorithms can influence decision-making in classic resource allocation problems such as portfolio management in finance and inventory management in operations. Using the anchoring and adjustment heuristic, I derive my predictions regarding algorithmic influence and test them using laboratory experiments. In chapters 1 and 2, I focus on project portfolio management and the multi-item newsvendor problem, respectively. In both these domains, I find that highly risk-averse algorithmic recommendations have the strongest influence on resource allocation decisions, despite individuals hedging away from the advice. Importantly, the changes in resource allocation decisions tend to persist even after the algorithm is no longer available. Chapter 3 reveals that these effects are similar regardless of factors such as decision autonomy (i.e., whether the algorithm is externally assigned or chosen by the subjects themselves) and source of advice (i.e., human or algorithm). Additionally, I find risk-averse algorithms can be used to counteract the “pull-to-center” bias in the low-profit newsvendor regime. Overall, I demonstrate the mutability of human behavior when temporarily exposed to risk-averse algorithmic aids. The findings are of notable consequence to firms strategically looking to utilize risk-averse algorithmic tools to improve resource allocation decisions without curtailing managerial autonomy.