Feeling Thermometer Reveals How Quickly Social Media Freezes Political Warmth

Social scientists have long used “feeling thermometers” to measure political attitudes, asking people to rate how warm or cold they feel toward various groups. New research employed this classic methodology to quantify algorithmic influence with unprecedented precision, revealing that one week of altered social media feeds can shift political temperatures by degrees that previously required years to change.
The experiment involved over 1,000 X users during the 2024 presidential election. Researchers manipulated feed content to either increase or decrease exposure to anti-democratic attitudes and partisan hostility. After one week, participants rated their feelings toward political opponents on a zero-to-100 degree scale, with higher numbers indicating warmer, more positive feelings.
Results showed changes exceeding two degrees on this scale—a shift matching the polarization increase typically seen across four decades between 1978 and 2020. To appreciate the magnitude, consider that these shifts represent substantial changes in political animosity compressed from years into days through algorithmic curation that most users never consciously perceived.
The feeling thermometer methodology captures something crucial beyond simple policy disagreement. Political opponents can disagree profoundly on issues while still respecting each other’s humanity and good faith. But when feelings toward opponents turn cold—when people view those across the political divide as not just wrong but as threats or enemies—democratic compromise becomes nearly impossible.
This research demonstrates that social media algorithms significantly influence political temperature. Feeds can be optimized to warm or cool partisan animosity depending on what content is amplified or suppressed. Current optimization for engagement tends to promote the most emotionally provocative content, which typically cools feelings between political camps. But platforms could choose different optimization objectives that foster mutual respect even amid disagreement.