The COMPAS (Cоrrectiоnаl Offender Mаnаgement Prоfiling for Alternative Sanctions) algorithm is currently used in US court systems to predict the likelihood that a defendant would become a recidivist (i.e. to commit another crime after being released). After a careful analysis, ProPublica found that there was noticeable discrepancy in the predictions based on race and occupation. Northpointe, the company that created COMPAS, countered that any discrepancies could not be taken as evidence of bias since race and occupation were never used as variables in their model. What is the BEST explanation of this reasoning?