A practitioner's argument that meeting summarizers fail in the same way regressions fail when you skip the part where you ask what the data can support.