The study, published in the peer-reviewed Journal of Information Science by researchers at New York University and the University of Edinburgh, is methodologically careful in ways that make it difficult to dismiss. Using a dataset of over 2.3 million social media posts across three platforms, it applies a computational model that identifies political orientation independent of self-reported labels, then tracks content moderation outcomes — removal, demotion, and account suspension — against that classification.
The findings: content classified as conservative or right-leaning faces moderation actions at a rate 2.8 times higher than equivalent content classified as liberal or left-leaning, controlling for engagement, virality, community reporting rates, and platform-stated policy violations.
Platform Responses
Representatives for two of the three platforms that cooperated with the data request disputed the study’s methodology, arguing that the political classification model introduced bias. The lead researcher, Dr. Samantha Chen, published a 40-page methodology appendix in advance of peer review precisely to address such objections. Three of the paper’s four peer reviewers found the methodology sound; one raised concerns about sample selection.
What makes the study particularly significant is what it does not claim. The researchers are explicit that they cannot determine whether the observed disparity reflects deliberate policy choices, algorithmic emergent behavior, or differential reporting patterns by user communities. “We can measure the outcome,” Dr. Chen told Conservative.to. “We cannot, from this data alone, attribute intent.”
Leave a Reply