AI That Rejects You May Soon Tell You How to Fix It

Every day, computer programs decide who gets a loan, a job interview, or insurance — and people who get turned down are usually told nothing. Two researchers, Matthew Olckers and Toby Walsh, studied a simple question: why would a bank or an employer bother explaining how to fix a rejection? Their answer is that it can be good business, not just good manners. Helping someone improve their score makes that person more valuable to the company later. So explaining rejections isn't charity — it's self-interest.

The system also works as a filter. Offering a fix-it path only helps the company if the people who take it are genuinely better applicants. Because following through costs time and effort, only the motivated ones bother. That means the offer itself sorts people: those willing to do the work prove something about themselves just by trying.

The researchers found the smartest approach is three buckets. Low scores get rejected outright. High scores get accepted immediately. Middle scores get a "here's what you'd need to change" list. But there's a twist: that middle group overlaps with people who would have sailed through under the old system. So some applicants who used to get in for free must now clear an extra, costly hurdle to earn the same result.

The practical takeaway is that explaining AI rejections is valuable but not automatically fair. Whether it helps you depends on where you land. Companies will likely offer these fix-it lists where it benefits them, not necessarily where it's needed most. And if you're a borderline case, expect homework.