Better and Faster Decisions with Recommendation Algorithms
84 Pages Posted:
Date Written: December 09, 2024
Abstract
While recommendation algorithms are increasingly powerful and prevalent, their influences on individual decision-making remain largely unexplored. To address this question, we conduct a randomized controlled experiment where subjects of a US representative sample make risky decisions. Subjects receive no recommendations in one baseline condition and random recommendations in another baseline condition. In three treatment conditions, subjects receive recommendations based on decisions of the majority, their own past decisions, or decisions of similar subjects. Compared with baseline conditions, subjects tend to follow recommendations and they exhibit less stochastic choices, behave more consistently with expected utility, and make faster decisions. Moreover, subjects are willing to pay to receive recommendations for subsequent decisions. This study helps understand behavioral mechanisms underlying recommendation algorithms and sheds light on the design of choice architecture with the assistance of artificial intelligence.
Keywords: preference, noise, risk, recommendation algorithm, experiment
Suggested Citation: Suggested Citation