That is why there are metrics like inter rater reliability and significance tests and confidence intervals, which seem to be all missing in paper...