Website CV Paper
Fields
Behavioral & Experimental, Applied Micro
Job market paper
"Recommended For You": Decision-Making under Algorithmic Recommendations
Concerns have been growing over the effect of recommender systems on consumer choices. I study how recommendation presentation, preference-based targeting, and disclosure shape choices through two incentivized experiments using theory-based and data-driven personalization. Recommendations exert substantial influence even without targeting: recommending a randomly selected alternative increases its selection by approximately 20 percentage points. Personalized recommendations increase agreement with model predictions, but the welfare implications of theory-based and data-driven personalization diverge. Theory-based personalization reduces average welfare shortfalls by approximately 43% relative to unaided choice. Data-driven personalization increases prediction but is not accompanied by statistically detectable improvements over unaided choice in expected utility or expected monetary payoff. Disclosure increases recognition of personalization, but evidence of an additional choice effect is limited. These findings highlight a tension underlying concerns about recommender systems: recommendations can effectively direct choices even when random, while personalization can attract choices without delivering gains in welfare.