When Rejection Stops Being Local

An applicant can send the same résumé to ten employers and encounter the same disqualifying rule at all ten. The rejection emails arrive from separate addresses.

An institution’s accuracy score describes how often its decisions are right. It does not tell you whether the people it misjudges are also misjudged everywhere else. Where employers make different errors, rejection at one may be corrected by recognition at another. A more accurate shared screen can overlook fewer capable applicants at every firm while overlooking exactly the same people across the market.

Suppose that screen wrongly treats a career break as disqualifying. Ten employers applying the rule have not made ten independent assessments. The tenth refusal adds no independent evidence that the applicant cannot do the work.

The mechanism predates artificial intelligence: a default wrongly attributed to a borrower can block applications to several banks because they share faulty evidence, even if their decision systems differ.

A shared system may remove prejudices or clerical errors that independent assessors perpetuated. Those gains can outweigh the loss of alternatives, leaving fewer people excluded overall. Yet the remaining errors can become harder to escape, concentrating exclusion on a smaller group of applicants.

Nor does common software necessarily produce common outcomes. A reviewer empowered to disregard the mistaken rule can preserve a second chance within an institution when switching institutions cannot. The danger is dependence on the same uncorrected mistake, not the mere use of the same tool.

The question belongs to the market: can someone wrongly refused here receive an independent assessment elsewhere? An individual accuracy score cannot show whether competition supplies correction or merely repeats a verdict.

For those people, the number of competitors is not the number of chances. Ten doors do not offer ten ways through if the same error keeps them locked.