Two completely different forms of personalization have led to significant outcomes, with one taking place in a controlled AI-driven lab experiment and the other in the realm of British electoral politics. A recent study conducted in a lab setting demonstrates the persuasive power of AI-driven large language models (LLMs) when it comes to personalizing arguments. Salvi, Ribeiro, Gallotti, and West (2024) reveal in a working paper that a personalized LLM was able to persuade individuals significantly more than humans in an online setting, with an increase of over 80% (p<0. 01). In other words, when faced with an LLM equipped with demographic information to personalize its arguments, humans were 81. 7% more likely to agree compared to when debating with another human. The study refers to previous experiments from the past year which have shown that LLMs are just as effective as human participants and even professional propagandists when it comes to crafting persuasive text (Bai et al. , 2023; Palmer & Spirling, 2023; Goldstein et al. , 2023; Karinshak et al. , 2023). While the discussion topics in the study were deliberately chosen to be accessible, and participants were assigned to either a pro or against condition, this slightly compromises the study's real-world applicability. However, topics such as animal research ethics and race as a factor in college admissions are undoubtedly relevant to ongoing debates. In anonymous debates, both humans and the LLM had access to demographic data about the individuals they were debating. The Power of Personalization in Election Campaigns Moving from Swiss labs to the real world, the influence of personalization was recently demonstrated in Rochdale, England during a by-election. Former Labour MP George Galloway sent letters to specific areas of his new constituency, focusing on certain foreign policy stances in some letters and social values in others. This sparked controversy as some viewed it as targeted messaging for the Muslim community in one set of letters, and for the White community in the other. Despite the rudimentary nature of this "technology, " Galloway emerged victorious.

While the impact of his dual-messaging strategy on his victory is up for debate, this is not an isolated example of the effectiveness of tailored and varied messaging. Other instances include Cambridge Analytica's notorious involvement in the 2016 U. S. presidential election (though it is worth mentioning that Barack Obama apparently used personalization in his successful 2012 reelection campaign). These cases provide real-world applications that may go beyond what is reflected in research papers alone. Uncharted Territories Salvi, Ribeiro, Gallotti, and West (2024) express concerns regarding the potential malicious use of these tools. With increasingly advanced LLMs having access to vast amounts of personal data, coupled with more human interactions occurring online and an escalation of polarization on crucial political and societal issues, a storm is brewing. It is likely that these developments will continue to shape various aspects of human behavior, including voting patterns and consumer choices. However, the full extent of their impact and transformative potential remains unknown. In fact, it may be challenging to gauge, as AI technology advances and becomes further detached from human cognition in the future.