Large Language Models Are Democracy Coders with Attitudes

Abstract

Current political developments worldwide illustrate that research on democratic backsliding is as important as ever. A recent exchange in Political Science & Politics (February 2024) highlighted again that the measurement of democracy remains a challenge. With many democracy indicators consisting of subjective assessments rather than factual observations, trends in democracy over time could be due to human biases in the coding of these indicators rather than empirical facts. This article leverages two cutting-edge Large Language Models (LLMs) for the coding of democracy indicators from the V-Dem project. With access to huge amounts of information, these models may be able to rate the many “soft” characteristics of regimes at substantially lower costs. Whereas LLM-generated codings largely align with expert coders for many countries, we show that when these models deviate from human assessments, they do so in different but consistent ways. Some LLMs are too pessimistic and others consistently overestimate the democratic quality of these countries. Although the combination of the two LLM codings can alleviate this concern, we conclude that it is difficult to replace human coders with LLMs because the extent and direction of these attitudes is not known a priori.