Hacker News Reader

Comment by rootusrootus | Hacker News Reader

I have a problem with the recent practice of using innocuous places like ancestry.com as a tool to find murderers through their familial relationships. Definitely I think we need to see some more regulation there.

But, as a separate issue, when there is enough evidence to convince a judge to sign a warrant, I do not have an objection to exhuming a body. I have to trust the judicial system to handle this sanely.

View on HN

What are the downsides of using databases like ancestry.com to solve crimes?

Forensic testing of DNA looks at 13 positions. 9 out of 13 is considered a match.

When you match against a very large database using such a small number of polymorphisms, you get many matches.

Detectives then take that list and look for people who lived sort of in the crime area. The matches who don't happen to have an alibi they can prove for a crime sometimes decades ago are then deemed "guilty".

It's complete pseudoscience.

If they did a full DNA sequence and matched that, sure. But they never do.

Also there is the problem that many DNA samples contain DNA from multiple individuals, in which the match can be assembled piecemeal. This completely destroys the statistical validity.

DNA forensics are absolute pseudoscience. Any court case relying on DNA should be discarded by the jury.

These are not conventional forensic DNA tests.

It's not "pseudoscience" but literally how everything is investigated, DNA or not. There is always a process of elimination, strong vs weak evidence, probable cause, and many more factors. Audio/video surveillance, alibis, even eyewitness reports and just about everything else can be unreliable, incorrect or faked. That's why police gather all kinds of evidence and present it to a jury.

> That's why police gather all kinds of evidence and present it to a jury.

There was no jury evaluating evidence and deciding guilt or innocence in this case. There was also no case. Only a proclamation of guilt against a man who died long ago and had no ability to defend himself, no defense attorney assigned, no arguments made before a jury whatsoever. The only "forensic" evidence made by some dodgy startup pushing this story to sell their "services".

That's not justice at all. It's not even a caricature of justice. It's complete and absolute nonsense. Holding this up as an example of a done guilty man found is an atrocity and everyone doing so should be deeply ashamed.

Did you not read that they dug him up to do a full dna analysis to compare to seen found in the victims underwear? And that there was other evidence to link him which is what led to the permission to exhume him?

The only reason that this wasn't taken further is as he was dead.

The issue remains, though, that if he were alive all this evidence would result in a trial -- at which the prosecution would have had to prove their case beyond reasonable doubt. The accused would have the opportunity to defend himself, to have legal representation, to present evidence that might call into question or contradict aspects of the prosecution's case, etc.

As it stands, it seems that due process has been bypassed.

I'm troubled by the idea that in the case of a dead defendant, it's OK to bring in a guilty verdict on the basis of an entirely one-sided "trial".

Legally, this is not a guilty verdict, precisely because of the due process issue you cite.

The due process here was getting a warrant. No rights were violated.

The poster you're replying to is referring to the misuse of statistics, essentially. Think about it this way: it's not amazing that someone wins the lottery. It'd be pretty amazing if a specific person won the lottery (statistically speaking).

Here's where DNA evidence shines: through other means you have a suspect or a short list of suspects and you use DNA evidence from the crime scene to eliminate suspects or concur with existing evidence.

Here's where DNA evidence is abused: if you take essentially a "hash" of DNA and look for matches to that same hash. There are enough people that you're likely to find unrelated hits (collisions, essentially) and then law enforcement puts the onus of proving their innocence on those people. This works particularly badly for the poor who are less likely to have credit card receipts and other evidence that might otherwise constitute an alibi.

Don't believe me? Consider [1] for issues in fingerprint matching. DNA databases are getting sufficiently large that adventitious matches are much more common than you might think [2]. I believe there's been at least one murder case in the US where searching the fingerprint database brought up a completely unrelated person who had the exact same fingerprint (based on the markers) but I can't find a reference to it now.

tl;dr Fingerprints and DNA are good for eliminating or indicting someone you otherwise have legal cause to suspect and shouldn't be used as a blanket scan.

[1] https://theconversation.com/fingerprinting-to-solve-crimes-n...

[2] https://www.theguardian.com/science/2004/sep/09/sciencenews....

It's not even really a "hash" in the sense that there's no predictable relationship of the outputs of similar inputs. The "hash" of my DNA is going to be far more similar to people in Seattle than to people in Tokyo, so it becomes far less meaningful to say my DNA matched and I was Seattle at the time.

Hashes don't need to have large changes in output for small changes in input. A hash just transforms input data into a fixed space.

For example, perceptual hash functions like Locality Sensitive Hashing (LSH) for text and pHash for images are designed so that small changes in input produce small to no change in output.

Also, your DNA is an order of magnitude (or more) more similar to someone in Tokyo than it is different to someone in Seattle.

There was an article on HN a while back which demonstrated how presenting this kind of evidence works in practice. Police arrested a suspect on a DNA match, put on a dog and pony show for the jury about how absurdly unlikely it was for the DNA to match by chance, he was convicted and put away for a long time. What the jury didn't know - because the prosecutors managed to block the defence from telling them - is that the police had found half-a-dozen locals who were DNA matches and chosen the most unsavoury-sounding one to arrest. The tale they'd told the jury about the absurd odds of a DNA match if he wasn't the criminal was a bunch of nonsense, and the defence weren't allowed to clue them in about this. He's probably still in jail right now.

Regulation is only to stop harm. Who does this hurt ? I don't think capturing criminals counts as harm. It's not trust that's needed, what's not to trust ?

Reply on news.ycombinator.com