Earlier quoted context omitted.
I don't think the fact that the process is automated is a problem (and I don't see anyone saying so). The problem is that existing algorithms are bad (like in the article where an algorithm decided two people are the same based on name and year of birth alone, when all other factors speak against it) and that complaints are not handled properly,
[from the article] > the defendant, unsurprisingly, had the same legal first name, last name and middle initial. We were also both born in 1987. But our similarities end there: He has half a dozen convictions. He’s black. He’s currently incarcerated. And an arrest warrant indicated that he was homeless > when all other factors speak against it Given that the agency thinks they're the same person, the number of convic…
I would assume they can get this information, considering that some reporter seemed to be able to get this information easily.
Making hard decisions on wether two people are the same is hard. But that's exactly why "name and year of birth match" seems like a naive algorithm. I would expect a match probability based on the treasure trove of statistical data that these agencies possess. A person that is currently incarcerated is less likely to be the same person that is currently spending and earning money, and for which landlords request reports. Skin color is strongly correlated to place of residence: if you know (automatically or manually detected from a mugshot) that one person is black, you can get the statistical probablility that the other person is black from the neighbourhood he lives in. etc.
>I think the root of the problem is getting a criminal background check and a credit report from the same place.
I agree that this is also not optimal. In Europe this type of information is obtained by requiring a criminal record certificate, a certificate anyone can get from a police station to show their (lack of) criminal record.