The field of sustainable food systems and AI-driven social simulations is witnessing significant developments, with a growing focus on reducing meat consumption and leveraging large language models (LLMs) to simulate social behaviors. Research is exploring the effectiveness of interventions such as Meat-Free Days in reducing greenhouse gas emissions and promoting healthier dietary habits. However, challenges such as customer retention and adherence to dietary guidelines remain. In the realm of AI-driven social simulations, LLMs are being utilized to model complex social systems and behaviors, including the simulation of persuasive dialogues on meat reduction. Nevertheless, limitations and biases in these models are being identified, highlighting the need for rigorous validation and calibration. The development of novel methods, such as Mixture-of-Personas language models, is addressing the issue of capturing behavioral diversity in target populations. Noteworthy papers include:

• A study on simulating persuasive dialogues on meat reduction with generative agents, which found promising results in producing outcomes consistent with theoretical expectations.
• A proposal for Mixture-of-Personas language models, which demonstrated improved alignment and diversity metrics in simulation tasks.
• A critical review of generative social simulations, which highlighted the limitations and challenges of using LLMs in this context, including the lack of awareness of historical debates and poorly addressed validation.