Measuring Political Preferences in AI Systems: An Integrative Approach