Friday, May 25, 2007

MySpace and sex offenders -- can this get any more ridiculous?

I created a MySpace page once, just to see what the fuss was about. Hint, it's about a fundamental biological activity that's not eating, sleeping, or breathing. Most of the profiles are of people from ages 15 to 25.

Astonishingly some of the clientele tend to be ... on the prowl. I thought that was kind of the point, but some are nastier prowlers than others. Recently MySpace decided it was going to try to screen out members who've been convicted of "sex crimes". Now, how are they going to do that?

This is known in my day job as a problem in "patient matching". In the absence of a unique identifier, how can you tell two people are the same? Well, it's not easy -- even if your population is cooperating. Matching is statistical. If the gender, us-legal-names, address history, birth dates, etc more or less match, then you assume the two people are the same. A social security number is often (mis)used in healthcare matching, but a phone number can do almost as well. Sooner or later you do get a false match, but, if people aren't trying to hide their identity and you can ask for a SSN or phone number, you can do pretty well. All of these obvious matching algorithms, by the way, are patented.

Homeland security tries to do the same thing with less data. They probably severely inconvenience over a thousand innocent people for every bad actor they may deter. It's completely pointless.

So, what did MySpace do? They hired a company that tries to track felons based on a (probably patent violating) matching algorithm. MySpace doesn't have a SSN though, so the matches are highly problematic. Not to mention that any "predator" who ever entered correct data has now changed it. Oh, and do you think the birth dates on MySpace are accurate?

The company had to find some matches, so they relaxed their algorithm a tad ...
ABC News: MySpace Error: Woman No Sex Offender:

... 'The Jessica Davis in question is absolutely not a sex offender,' Cardillo told ABCNEWS.com, explaining that beyond sharing a similar and common name, Jessica Davis the non-sex offender and Jessica Dawn Davis the sex offender also had birthdays two days off as well as two years off and had lived in Florida at roughly the same time.

Cardillo, who called the initial match an'unfortunate circumstance,' said that the database worked exactly as intended.

'It was so close,' Cardillo told ABCNEWS.com. 'It was one of those rare instances where there was nothing else we could have done but flag her. If we get an offender and I'm looking at a date of birth that's two days off, we're going to assume were dealing with the offender.'"
In other words, they deliberately err on the side of false positives. There's no appeal process, MySpace simply deletes the user's profile page and all their content. I'll bet, even considering that the MySpace user base almost certainly has a higher than average concentration of bad actors, that the majority of their matches are false positives. Wait until MySpace then turns over the names and addresses of their "matches" to the police.

Sigh.

No comments: