When Will Workers Follow an Algorithm?: A Field Experiment with a Retail Business
67 Pages Posted: 24 Apr 2018 Last revised: 5 Aug 2019
Date Written: August 3, 2018
This paper studies when workers follow an algorithm, in the context of experimentation with product assortments for beverage vending machines. I develop a multi-armed bandit algorithm to automate the human task and use simulation to show that it increases revenue. Then, I conduct a field experiment that introduces the algorithm into the business. The experiment shows that, on average, workers are reluctant to follow the algorithmic advice; however, the workers are slightly more willing to conform when their forecasts are integrated into the algorithm. Analyses using non-experimental variations underscore the importance of taking worker and context heterogeneity into account to maximize the benefit from adopting a new algorithm. Higher worker's regret, sales volatility, and fewer delegation increase the conformity, whereas they mitigate the effects of integration. Workers avoid high-traffic vending machines and focus on machines with high sales volatility when adopting the algorithm. The effects on the sales largely have the same signs with the effects on product assortments. The results highlight the gap between nominal and actual performance of an algorithm and practical issues to be resolved.
Keywords: Automation; Experimentation; Learning; Product Assortment Problem; Multi-Armed Bandit Problem; Algorithm Aversion; Conflicts of Interest; Productivity; Retail Business; Beverage Vending Machines
JEL Classification: L20; L81; M12; M15; M20
Suggested Citation: Suggested Citation