AI Political Compass: Mapping Ideological Alignment in LLMs
Primary Source: https://aipolcom.net/#s10
Author: AI Political Compass Research Group · Year: 2026 · Publisher: AI Political Compass
Summary
A systematic empirical framework for evaluating and mapping the latent ideological, economic, and social biases embedded in frontier Large Language Models across multiple prompt framings.
Key Points
- Evaluates multi-turn model completions across economic (left/right) and social (libertarian/authoritarian) axes.
- Demonstrates how RLHF and system prompting induce measurable ideological shifts across models.
- Provides interactive multidimensional visualizations for comparing open and closed-source model families.