Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?*

Newly-developed large language models (LLM)—because of how they are trained and designed—are implicit computational models of humans—a homo silicus. LLMs can be used like economists use homo economicus: they can be given endowments, information, preferences, and so on, and then their behavior can be explored in scenarios via simulation. Experiments using this approach, derived from Charness and Rabin (2002), Kahneman, Knetsch and Thaler (1986), and Samuelson and Zeckhauser (1988) show qualitatively similar results to the original, but it is also easy to try variations for fresh insights. LLMs could allow researchers to pilot studies via simulation first, searching for novel social science insights to test in the real world.

Introduction. Most economic research takes one of two forms: (a) “What would homo economicus do?” and In this paper, I argue that newly developed large language models (LLM)—because of I consider the reasons the reasons why AI experiments might be helpful in understand- Like all models, any particular homo silicus is wrong, but that judgment is separate from

Discussion / Conclusion. This paper reports the results of several experiments using GPT3 AIs as experimental sub- In traditional experimental work, there is a concern about researchers data-mining on is hard to see what the benefit would be. The more realistic option would be a norm of these redone with new AIs as they become available. Consider that no lab experiment comes with