Personalized Persuasion by Shaping Beliefs About the Multidimensional Features of Objects

Kazunori Terada, Yasuo Noma, Masanori Hattori

Abstract: Personalization consistently improves the impact of persuasive agents, recommender systems, and nudge interventions. However, most implementations still rely on coarse-grained demographic or personality attributes such as age, sex, or broad personality traits. Ultimate personalization instead tailors messages to an individual’s nuanced beliefs and utilities. Multiattribute utility theory formalizes personal utility as the sum of utilities over an object’s features, highlighting that negative beliefs about even a few features can suppress overall acceptance. Leveraging multiattribute utility theory, we test whether countering the features of each individual value can increase the acceptance of a complex technology—fully autonomous vehicles. For each participant (N = 197), a belief-manipulation algorithm identified the five fully autonomous vehicle features with the lowest subjective utilities and generated counterpropositions during a semistructured dialog with a virtual agent. Pre- and postdialog measures captured (i) the monetary valuation of fully autonomous vehicles, (ii) the desire to ride, and (iii) the perceived social obligation to accept fully autonomous vehicles. Two comparison conditions served as controls: a nonpersonalized counterargument condition and a baseline small-talk condition in which the agent discussed everyday topics unrelated to autonomous vehicles. The results showed that the proposed method improved the social obligation to accept fully autonomous vehicles more than the nonpersonalized method and the baseline method did but had no effect on the monetary value or the desire to ride them. This suggests that personalized belief manipulation may not be effective in enhancing the “want to” desire or utility of an object but may only improve the thought of “ought to do”. These findings have broad implications for understanding belief manipulation across diverse domains where acceptance involves complex cognitive and motivational processes, including healthcare decision-making, educational engagement, and sustainable behavior adoption.

External IDs:doi:10.1109/access.2025.3575369