Both the selection (survivorship bias) and weighting (mentions rather than votes/discussion) are problematic.
Discussion itself can be a very poor metric, as when there's a great deal of general agreement there's frequently little (or at least less) as when compared with highly contentious topics.
Research methodology can often be restricted by data availability. Finding novel ways to torture^W explore data can be insightful, though raises its own risks (p-hacking, as an example).
Discussion itself can be a very poor metric, as when there's a great deal of general agreement there's frequently little (or at least less) as when compared with highly contentious topics.
Research methodology can often be restricted by data availability. Finding novel ways to torture^W explore data can be insightful, though raises its own risks (p-hacking, as an example).