Researchers routinely compare machine-labelled sentiment scores across social media platforms, interpreting differences as shifts in public opinion. A new arXiv preprint argues this practice rests on a fragile assumption: that the platform itself does not influence the measurement.

The paper, "Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment," proposes a protocol and simulation study to examine whether classifiers produce equivalent labels across platforms. The title itself suggests the platform may be part of the measurement, meaning observed cross-platform differences could be artefacts of platform-specific language, user demographics, or moderation.

The authors call for testing cross-platform equivalence before drawing substantive conclusions. Because the abstract is truncated, the full details of the protocol and simulation results are not available here, but the framing highlights a methodological risk in comparative social media analysis.