How does more data hurt an investigation, if you are statistically literate? What is the conditional probability on that DNA test when it turns up a match with someone who also knew and was in recent contact with the murdered person and had a motive?
Same with watch lists, doesnt all the ineffectiveness go away if they have tiering or triage system?
> Same with watch lists, doesnt all the ineffectiveness go away if they have tiering or triage system?
If properly used, perhaps. However, the standards for government forensics have been... questionable.
Generally speaking, government "tests" seem more focused on confirming the prosecutor or investigator's theory than on actually resolving truth about facts.
You seem to be assuming a great deal of statistical literacy on the part of people utilizing these database tools, which I'm not sure is a solid assumption.
But a lot of arguments about surveillance tech seem to be really arguments about people not doing their jobs. If they get overwhelmed with the data, it's not the fault of technology. So preventing them from using new technology doesn't strike me as a right solution.
I don't think you can dismiss that argument until people actually do there jobs. Case in point there has been a big push to stop collecting SSN on so many forms. Not because you can't collect that information saftly in theory, but because people and systems have repeatedly failed to protect it in the past.
You’re now basing your case on circumstantial evidence which is considered near worthless for a reason.
If you have 5 suspects and then do DNA testing that's solid evidence. Ex: Paternity test. But, if you find someone with DNA testing you need to completely ignore that evidence after that point as 100% of its predictive power has already been used.
PS: To show just how easy it is to mess this stuff up. Suppose a murder occurred in NY city and the DB returned 20 people so you look at the closest 3 suspects and aha someone lives just 10 miles from the crime! However, note you already picked someone that was close by so you need to consider the odds that someone who lives close by happens to lives in the area. EX: A high percentage of people living in say NY state also live within 10 miles a crime in NY City.
That seems to be a textbook case of people failing to apply Bayes' theorem properly, not a fault of tests or technology. They, just like doctors analyzing cancer tests, need to remember to multiply one more number...
Same with watch lists, doesnt all the ineffectiveness go away if they have tiering or triage system?