The persuasive effects of political microtargeting in the age of generative artificial intelligence

The increasing availability of microtargeted advertising and the accessibility of generative artificial intelligence (AI) tools, such as ChatGPT, have raised concerns about the potential misuse of large language models in scaling microtargeting efforts for political purposes. Recent technological advancements, involving generative AI and personality inference from consumed text, can potentially create a highly scalable “manipulation machine” that targets individuals based on their unique vulnerabilities without requiring human input. This paper presents four studies examining the effectiveness of this putative “manipulation machine.” The results demonstrate that personalized political ads tailored to individuals’ personalities are more effective than nonpersonalized ads (studies 1a and 1b). Additionally, we showcase the feasibility of automatically generating and validating these personalized ads on a large scale (studies 2a and 2b). These findings highlight the potential risks of utilizing AI and microtargeting to craft political messages that resonate with individuals based on their personality traits. This should be an area of concern to ethicists and policy makers.

Introduction. In the summer of 2016, the world was struck with the Brexit refer­ endum results, in which the United Kingdom had voted to leave the European Union. On the other side of the Atlantic, Donald Trump was only months away from getting elected to be the 45th president of the United States. At the same time, Alexander Nix, a CEO of a relatively obscure company called Cambridge Analytica, promoted the company’s ostensible success at swaying voters from one political candidate to another based on exploit­ ation of voters’ particular psychological vulnerabilities. This tac­ tic involves deducing psychological attributes that are not readily observable, such as personality traits, from individuals’ online behavior and personal data. Subsequently, these inferred psychological features are leveraged to craft highly personalized messages tailored to each individual.

Discussion / Conclusion. Our findings indicate that political microtargeting is an effective technique and can be automated using off-the-self generative AI. While we show consistent efficacy across four studies, it is im­ portant to recognize that the demonstrated effect sizes are rather small. Nonetheless, small effect sizes can turn substantial at scale, and the automation of political microtargeting is pivotal to achieving such scale. To put this effect size in context, based on the change between the median mismatch and no mismatch between individuals and ads, we have simulated 100,000 re­ sponses. Given a cutoff of 3 (midpoint of the 5-point perceived per­ suasion scale), we find that out of every 100,000 individuals exposed to political messages tailored to their personality, 2,490 individuals would now be expected to be persuaded due to the style. By extrapolation, the change from the worst matching to the best matching would mean an increase to 11,405 individuals. Considering that elections are often decided by fractions of a per­ centage point, shifting a few thousand individuals out of 100,000 can substantially impact the results.