Calibrating Behavioral Parameters with Large Language Models
Brandon Yee, Krishna Sharma
TL;DR
This paper presents a framework to treat large language models as calibratable measurement instruments for eight canonical behavioral biases relevant to asset pricing. By using profile-driven prompts, the authors induce and validate parameter shifts (e.g., $\lambda$, $\theta$, $\rho$, $w$, $\gamma$, $\tau_N/\tau_F$) against human benchmarks, establishing a calibration framework with monotonicity, range coverage, stability, and coherence. Four biases—loss aversion, herding, extrapolation, and anchoring—achieve strong validation and reach human-relevant magnitudes; others show weaker performance or instability, highlighting domain limits (affect, social pressure, real stakes). Embedding calibrated parameters into a simple ABM reproduces momentum and longer-horizon reversals consistent with empirical findings, linking calibrated cognitive biases to market dynamics. Overall, the work extends behavioral finance measurement by providing scalable, validated instruments for cognitive biases while delineating boundaries where LLM-based calibration remains informative rather than substituting human subjects.
Abstract
Behavioral parameters such as loss aversion, herding, and extrapolation are central to asset pricing models but remain difficult to measure reliably. We develop a framework that treats large language models (LLMs) as calibrated measurement instruments for behavioral parameters. Using four models and 24{,}000 agent--scenario pairs, we document systematic rationality bias in baseline LLM behavior, including attenuated loss aversion, weak herding, and near-zero disposition effects relative to human benchmarks. Profile-based calibration induces large, stable, and theoretically coherent shifts in several parameters, with calibrated loss aversion, herding, extrapolation, and anchoring reaching or exceeding benchmark magnitudes. To assess external validity, we embed calibrated parameters in an agent-based asset pricing model, where calibrated extrapolation generates short-horizon momentum and long-horizon reversal patterns consistent with empirical evidence. Our results establish measurement ranges, calibration functions, and explicit boundaries for eight canonical behavioral biases.