LLMs are not weird: Comparing AI and human financial decision-making

In this paper, we explore how large language models (LLMs) make financial decisions by systematically comparing their responses with those of human participants across the world. We presented a set of commonly used financial decision-making questions to several leading LLMs, GPT-4, GPT-4o, GPT-5, Gemini 2.0 Flash, and DeepSeek R1, each evaluated across multiple temperatures, yielding a total of 21 model-temperature combinations. We then compared their outputs to human responses drawn from a dataset covering 53 nations. Our analysis reveals three main results. First, in cross-national comparisons, the aggregate responses of LLMs cluster together, forming a distinct group separate from all nations, clearly not WEIRD (Western, Educated, Industrialized, Rich, Democratic), contrary to what has been suggested in previous studies conducted in other contexts. Second, LLMs generally exhibit a risk-neutral decision-making pattern, favoring choices aligned with expected value calculations in lottery-type questions. Third, when evaluating intertemporal trade-offs between present and future rewards, LLMs often generate internally consistent and economically rational responses. These findings contribute to the understanding of how LLMs emulate human-like decision behaviors and highlight potential cultural and training influences embedded within their outputs.