The Emergence of Economic Rationality of GPT

TL;DR

The paper investigates whether GPTs act as economically rational agents by testing their choices against revealed preference theory across risk, time, social, and food domains. Using 10,000 GPT tasks and a parallel human study, the authors compute GARP-based rationality scores (CCEI) and perform structural estimation of domain-specific preferences. GPT consistently achieves near-perfect rationality, outperforms humans, and displays more homogeneous preferences, with robustness to randomness and demographics but sensitivity to price framing and discrete-choice formats. The findings suggest LLMs can emulate rational decision-making in controlled settings, while also underscoring contexts that can induce deviations and the need for deeper mechanism-level understanding.

Abstract

As large language models (LLMs) like GPT become increasingly prevalent, it is essential that we assess their capabilities beyond language processing. This paper examines the economic rationality of GPT by instructing it to make budgetary decisions in four domains: risk, time, social, and food preferences. We measure economic rationality by assessing the consistency of GPT's decisions with utility maximization in classic revealed preference theory. We find that GPT's decisions are largely rational in each domain and demonstrate higher rationality score than those of human subjects in a parallel experiment and in the literature. Moreover, the estimated preference parameters of GPT are slightly different from human subjects and exhibit a lower degree of heterogeneity. We also find that the rationality scores are robust to the degree of randomness and demographic settings such as age and gender, but are sensitive to contexts based on the language frames of the choice situations. These results suggest the potential of LLMs to make good decisions and the need to further understand their capabilities, limitations, and underlying mechanisms.

Figures (28)

• Figure 1: Cumulative Distributions of the CCEI Values. This figure consists of four subplots for four preferences domains. Each subplot depicts a cumulative distribution function (CDF) plot, which shows the proportion of CCEI values less than or equal to a specific threshold. The light dotted lines represent simulated subjects, the dark dashed lines represent human subjects, and the solid lines represent GPT observations.
• Figure 2: Cumulative Distributions of the Spearman's Correlation Coefficient of $\ln(x_A/x_B)$ and $\ln(p_A/p_B)$. This figure contains four subplots for four preferences domains. The dashed (solid) lines represent human subjects (GPT observations).
• Figure 3: Scatter Plots of Estimated Parameters. This figure contains four subplots for four preferences domains. Each hollow circle (solid square) points represent a human subject (a GPT observation).
• Figure 4: Mean CCEI Values of GPT Observations across Different Variations. This figure displays the average CCEI values and 95% confidence intervals for GPT observations under different conditions: baseline, temperature of 0.5, temperature of 1, price framing, and discrete choices, and various demographic settings.
• Figure D1: Rationality Score in Prior Studies with Human Subjects. This figure presents the average CCEI values of human subjects in revealed preference studies ahn2014estimatingandreoni2002givingbanks2019educationcappelen2021developmentcarvalho2016povertycettolin2020cortisolchen2023consistencychoi2007consistencychoi2014moredean2016measuringdrichoutis2020economicechenique2011moneyfisman2007individualfisman2023distributionalhalevy2018parametriccrawford2010habitskim2018roleli2021povertymuller2019anatomy