Social scientists frequently make use of datasets where one or more variables are generated through human coding decisions (e.g. democracy scores, political party manifestos).
But human-coded variables typically carry some amount of measurement error.
