Can AI Predict Our Next Social Moves? How Language Models Are Reshaping Social Science

LLMs are transforming social science by automating critical research processes. They can generate and test hypotheses, design experiments, simulate human subjects, and analyze data, all without direct human intervention. This contrasts with traditional methods that are labor-intensive, requiring extensive data collection and analysis. By leveraging LLMs, researchers can accelerate the pace of discovery, test theories at scale, and gain deeper insights into social dynamics. The use of LLMs in social science facilitates in silico experimentation, allowing researchers to simulate social interactions and test hypotheses in a controlled, scalable environment.

Structural Causal Models (SCMs) provide a framework for organizing and conducting AI-driven social science research. They offer a formal language to express causal relationships, allowing researchers to define variables, their potential causes, and how they interact. In the context described, SCMs are integrated with LLMs to create computational systems. These systems use SCMs to generate hypotheses, design and run experiments on LLM-powered agents, and analyze the results. This approach enables researchers to explore a wide range of social scenarios and test social science theories in a controlled and scalable manner.

The systems using LLMs and SCMs can simulate various social scenarios, such as: bargaining over a mug to understand factors influencing deal outcomes; simulating bail hearings for tax fraud to identify biases; modeling lawyer interviews to examine hiring decisions and uncover biases; and analyzing open ascending price auctions to test auction theory. These simulations are designed to test specific hypotheses, identify causal relationships, and understand the underlying dynamics of social interactions. Each simulation provides a virtual laboratory where experiments can be conducted rapidly and at a fraction of the cost of traditional methods.

The integration of LLMs and SCMs offers several benefits. It automates the process of hypothesis generation and testing, accelerating research and allowing for larger-scale studies. It enables in silico experimentation, which is faster, cheaper, and more controlled than traditional methods. This allows researchers to explore a wider range of social scenarios, test theories more rigorously, and uncover new insights that might have been hidden. AI can identify causal relationships and provide a deeper understanding of the forces that shape human interactions, leading to new theories, better policies, and a more informed society.

Automated social science transforms the role of social scientists by providing them with powerful new tools. Rather than being solely focused on data collection and manual analysis, researchers can now leverage AI to generate and test hypotheses, design experiments, and analyze data. This allows for a shift toward higher-level thinking, focusing on the interpretation of results and the formulation of new theories. The collaboration between humans and AI promises to accelerate the pace of discovery, uncover new insights, and improve our understanding of the social world, leading to a more efficient and impactful approach to social science research.