The Rise of the Homo-Silicus: Artificial Personalities and Macroeconomic Dynamics

Large Language Models (LLMs) hold the potential to revolutionize science and the creation of knowledge. In the field of Economics, which has long been tied to the concept of rational agent, these models offer the possibility of disciplining behavioral analysis, given their immense capacity to process data quickly and efficiently, thus producing coherent, well-founded, and statistically robust behavioral rules. In this context, one may expect economic theory to undergo a major transformation in the near future, as the concept of Homo-Economicus begins to be challenged by a new idea, namely that of Homo-Silicus (i.e., the generative economic agent forged with the support of LLM algorithms). The latter will not be a single or homogeneous being like the former; on the contrary, it will be a much more complex, multifaceted, and diverse entity. To illustrate the potential of the new paradigm, this article uses an LLM to assist in assembling a classification of economic agents based on the definition of different personality traits. The exercise allows for the characterization of thirty-two distinct agent types, with their economic behavior diverging according to three dimensions: economic conservatism, social orientation, and propensity for risk and innovation. A simple macroeconomic model capable of accommodating the different behavioral profiles is outlined, revealing that long-term aggregate outcomes depend on which personality types are dominant in the economy. The interaction among heterogeneous agents gives rise to complex macro patterns of behavior, with aggregate variables potentially evolving along irregular and unpredictable pathways.