PVP: An Image Dataset for Personalized Visual Persuasion with Persuasion Strategies, Viewer Characteristics, and Persuasiveness Ratings

TL;DR

The paper introduces the Personalized Visual Persuasion (PVP) dataset, a large multimodal resource linking image persuasiveness to viewer psychology by pairing 28,454 images across 596 messages and 9 strategies with 2,521 annotators' demographics and personality traits. It presents two core tasks—a personalized persuasive image generator and an automated persuasiveness evaluator—and demonstrates that incorporating psychological characteristics improves both generation and evaluation, with baseline models showing promising performance. The authors provide extensive analyses of topics, strategies, and viewer traits, compare image sources (DALLE vs Google), and offer substantial baselines and insights to guide future research in personalized visual persuasion. The work highlights ethical considerations and outlines limitations, while establishing a foundation for developing responsible, psychology-informed persuasive AI systems with broad practical implications in advertising, public health, and political communication.

Abstract

Visual persuasion, which uses visual elements to influence cognition and behaviors, is crucial in fields such as advertising and political communication. With recent advancements in artificial intelligence, there is growing potential to develop persuasive systems that automatically generate persuasive images tailored to individuals. However, a significant bottleneck in this area is the lack of comprehensive datasets that connect the persuasiveness of images with the personal information about those who evaluated the images. To address this gap and facilitate technological advancements in personalized visual persuasion, we release the Personalized Visual Persuasion (PVP) dataset, comprising 28,454 persuasive images across 596 messages and 9 persuasion strategies. Importantly, the PVP dataset provides persuasiveness scores of images evaluated by 2,521 human annotators, along with their demographic and psychological characteristics (personality traits and values). We demonstrate the utility of our dataset by developing a persuasive image generator and an automated evaluator, and establish benchmark baselines. Our experiments reveal that incorporating psychological characteristics enhances the generation and evaluation of persuasive images, providing valuable insights for personalized visual persuasion.

Figures (15)

• Figure 1: Examples in the PVP dataset. Asterisks (*) indicate important elements.
• Figure 2: Data construction process. See the main text for details.
• Figure 3: Persuasiveness scores by persuasion strategy.
• Figure 4: Spearman correlations between evaluator scores and human judgments across topics.
• Figure 5: Correlation between image persuasiveness and psychological characteristics across different strategies: These plots illustrate the correlation coefficients between image persuasiveness and various characteristics scores across multiple strategic approaches.