LLM services in the management of social communications

This paper proposes enhancing social communication management with a behavioral economics approach through artificial intelligence instruments. The research aims to explore the influence of social communication on citizens' behavior using large language model services and assess its effectiveness. The paper builds on Daniel Kahneman's dual-process theory, highlighting the intuitive system (System 1) and the rational system (System 2) in decision-making. The author introduces a third system, System 3, representing rooted in identity socially conditioned behavior influenced by societal norms and self-awareness. On this theoretical basis, the paper emphasizes automating communication management through large language model services, freeing up citizens' potential for self-determination and self-organization. By leveraging these services, messages can be crafted to support social transformation while respecting historical, cultural, and political contexts. Based on the preconditions and restrictions described above, we use GPT-4 model to generate messages based on these narratives. The experiment will use an observational study design with virtual persons. To compare the impact of original and modified messages according to the addressee's mentality, we used the Claude 3.5 Sonnet model. We can see that the potential activity of respondents after perceiving the changed message does not change much, and the original message is perceived. Modifying messages by LLM services crafted to support social transformation while respecting historical, cultural, and political contexts cause attitudes to become substantially more negative (2.5 units downward shift in median); the intentions showed a slight positive increase (0.2 units upward change in median).