Date of Degree
Management Information Systems
Digital nudge, algorithmic nudging, e-commerce, cognitive theory, trust, transparency, recommendation dissonance, user decision making, recommendation badge, randomized experiment.
Most recently, academic research has started to address the relationship between nudges and Artificial Intelligence/Machine Learning (AI/ML) and found that personalized targeting algorithms can influence individual and collective behaviors in ways that may lead to undesired consequences for both end-users and firms. This dissertation investigates the effects of digital nudging, particularly algorithmic nudging, on user’s decision quality in online badge recommendations in the e-commerce context. Nudge is a choice architecture that alters individual’s behavior in a predictable way while preserving all available options and keeping the same economic incentives. In this dissertation, I conducted three independent and relevant studies and unveiled the impacts of algorithmic nudging on user decision making in the online shopping environment by employing a series of between-subject randomized experiments.
The first study explored the relationship between recommendation dissonance and decision quality where recommendation badges are used as nudges. We conducted a randomized experiment focusing on recommendation dissonance, and tested user decision quality across two treatment groups.
The second study examined the influence of nudging transparency on user decision quality. Drawing on nudging literature and cognitive theory, we proposed two types of recommendation badges based on transparency: ambiguous badge (e.g., Amazon’s Choice) and specific badge (e.g., Best Seller). The data analysis, based on the randomized experiment and survey instrument, revealed the significant difference of choice confidence between different groups of badge types (ambiguous vs. specific). Additionally, the regression analysis of perceptual constructs confirmed the significantly positive impact of perceived badge transparency on users’ decision confidence.
The third study examined users’ decision quality on a smaller platform (as a proxy of trust in platform) in the algorithmic nudging setting. In the third randomized experiment, I used the platform as the treatment and compared user decision outcomes with those from the second study. The findings indicated that on a smaller platform (e.g., Wish.com), there was no significant difference in choice confidence between recommendations featuring different badge types (ambiguous vs. specific). A post-hoc analysis was conducted with the merged data, and the regression results showed that trust in platform has a significantly positive impact on choice confidence.
These three studies explored algorithmic nudging on the e-commerce platform and will contribute to the nudging literature and biased recommendation research in IS. Furthermore, the results will enhance the overall understanding of the cognitive processes associated with recommendation badges and bring ethical implications to use of AI/ML algorithms in practice.
Luo, Yuxiao, "Investigation of Algorithmic Nudging on Decision Quality: Evidence from Randomized Experiments in Online Recommendation Settings" (2024). CUNY Academic Works.