An enraging story by the Associated Press spells it out: computers used by background checkers mix up two people with the same name. Blameless woman gets tagged with a criminal record that isn’t hers, can’t get work and ends up homeless.

GettyImages_rba1_Robots.jpg

We’ve written about the danger in relying on the “intelligence” of computers here, and the limitations of Google here. The message in these as well as the AP story about the ruined lives of background checks gone wrong are the same: there is no substitute for having an intelligent person conduct your research.

We saw such a problem earlier this year: a client called and said he wanted to hire (let’s call him) Robert M. Johnson for a high-paying job. Robert M. Johnson is an accomplished professional with a great resume and wonderful references. How then to explain the criminal record for fraud turned up by one of the databases? Robert M. Johnson the candidate said, “I’ve never been convicted of any crime. You’ve got the wrong person.”

Turned out he was right and the database was wrong. How could we tell?

1. We looked up the criminal case in question. While the AP story is correct that more and more criminal records are on line, the cases themselves – the critical documents – often are not and in any case need careful reading. We looked at the sentencing report in this case and found that the criminal Robert M. Johnson was sentenced to serve his time in a particular southern state to be close to “home and family.”

2. We then found that the Robert M. Johnson up for the job had no ties whatsoever to this southern state. His parents didn’t live there, he had never lived there, his Social Security number was not issued there, and he was working in Connecticut while the criminal Robert Johnson was committing his crimes down south.

This seems like simple work, but the databases were incapable of making the mundane connections that distinguished two very different men with the same common name. Why would this be?

To start with, not all databases come up with the same information. One may know where a person has worked for the last ten years, while another will have no information about employment. Still, the employment-light database will do a great job associating a person with companies he may own or on whose boards he serves. Both kinds of information are vital, but no machine can yet put them together to draw a complete picture.

Complete pictures come when the human mind gets involved, and even then, databases can only take you so far. We’ve written here about that too, and how interviewing is sometimes the only way to get the information you’re looking for.

But interviews or not, automated searching is like letting a robot build a car and then having another robot inspect the thing. The robots are certainly useful and lower the cost of the vehicle, but would you drive that car before a person had looked it over?