Left Leaning Models: AI Assumptions on Economic Policy

2025-07-21

AI Summary

This study investigates implicit value orientations and structural biases in large language models (LLMs) regarding economic policy evaluation. Using a conjoint experimental design, we systematically quantify the response sensitivity of 12 prominent LLMs across seven policy dimensions: unemployment, inequality, financial stability, environmental risk, economic growth, inflation, and public debt. Results reveal a consistent, statistically significant left-leaning preference—strongly prioritizing social equity and environmental sustainability while comparatively downweighting conventional macroeconomic objectives such as growth, inflation control, and debt management. This bias persists across models and contextual variations. To our knowledge, this is the first experimental study empirically identifying and quantifying the latent economic value preferences embedded in LLMs. By establishing a rigorous, experimentally grounded methodology, the work advances understanding of implicit biases in AI-driven policy analysis and provides foundational evidence for enhancing model interpretability, transparency, and reliability in socioeconomic applications.

Technology Category

Humans and AI: Learning Human Values and PreferencesMachine Learning: Large Multimodal Models (LMMs)Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

Economics, Online Markets and Human Computation: Cost models of using LLMs in production systemsUser Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendationSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

How does AI think about economic policy? While the use of large language models (LLMs) in economics is growing exponentially, their assumptions on economic issues remain a black box. This paper uses a conjoint experiment to tease out the main factors influencing LLMs' evaluation of economic policy. It finds that LLMs are most sensitive to unemployment, inequality, financial stability, and environmental harm and less sensitive to traditional macroeconomic concerns such as economic growth, inflation, and government debt. The results are remarkably consistent across scenarios and across models.

Problem

• Uncover AI assumptions on economic policy factors
• Analyze LLM sensitivity to unemployment and inequality
• Assess consistency of economic evaluations across AI models

Innovation

• Uses conjoint experiment to analyze LLMs
• Focuses on unemployment, inequality, stability, environment
• Ignores growth, inflation, debt in evaluations