A polarization score is supposed to describe a community, not its surroundings. But a new preprint on arXiv argues that this assumption breaks down when a minority community is embedded in a larger network. The authors show that the measured polarization of such a community can change—rising or falling—depending on the size and structure of the majority around it.

This means that polarization scores are not neutral or comparable across different contexts. A minority group's score may reflect the surrounding majority more than the group's own internal divisions or consensus. The paper frames this as a fairness problem, asking who is actually being described by the score.

The work is an audit of existing network polarization measures, highlighting a systematic bias rather than proposing a single fix. As a preprint, it has not yet been peer-reviewed, but it raises a concrete concern for researchers and policymakers who use polarization metrics to compare communities of different sizes and social contexts.