It doesn't need to be effective, it needs to be a computerized excuse to go after more people. Whenever computers are used to target people, the output is always given far too much weight. Recently we had guns drawn on a child because a computer vision algorithm classified his doritos as a gun with a low confidence score, with explicit advice to only investigate further and not assume correctness. But that child still had multiple guns trained on them.

Take that, apply it to here, and it's clear that effectiveness would actually be counterproductive.