Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Bad science and bad statistics in the courtroom convict innocent people

scientificamerican.com

Post reply on HN
Print view View original

31–40 of 205 posts

Re: Bad science and bad statistics in the courtroom convict innocent people

#31

Post by 10000truths »

WalterBright wrote: post #29

I read about a case once where a man was convicted of rape. The evidence was that the rape was done face-to-face with the victim, and the victim positively ID'd him. How could she not, when she was face-to-face with him? It was a slam dunk. After some years in prison, the convicted person managed to get a DNA test, which proved he was not the rapist. How could this be? It turns out the DNA matched a convicted sex off…

Do you have a link to the case details? At the very least, there has to have been some additional evidence to convict, even if circumstantial. I find it hard to believe that someone can be ruled guilty of a crime as serious as rape based solely on the plaintiff's testimony.

Re: Bad science and bad statistics in the courtroom convict innocent people

#32

Post by bjornsing »

A few years back there were some stories in a Swedish newspaper about a method for medical age estimation of asylum seekers developed by a government agency. I could hardly believe my eyes when the person responsible for the accuracy of the method was quoted as saying that it was very reliable, because it used two unreliable methods to estimate if the person was above 18 years old, and then concluded that they were 18+ if either method so indicated. You don’t need to be a genius to realize your false 18+ rate will be higher when combining two methods in this way than it would be if you used a single method, but he argued the opposite.

This caught my attention and I spent an inordinate amount of time, together with a group of other concerned people, trying to convince this government authority they were wrong, and later to expose them for it / stop them from continuing. But it sadly turned out to be next to impossible. They persisted with inaccurate/false affidavits and court testimony for years, until finally the minister of justice ordered a formal probe into the method. They then patched it up as best they could without exposing their previous errors, and pretended like nothing had happened. The formal probe issued an interim report with pretty harsh commentary, and was then cancelled by the minister of justice before it could give its final report.

Forever changed my view of the Swedish justice system.

Re: Bad science and bad statistics in the courtroom convict innocent people

#33

Post by graphe »

s1artibartfast wrote: post #22
graphe wrote: post #7

Earlier quoted context omitted.

Innocent meaning they didn’t do the specific crime, or innocent as in someone with a clean record that was railroaded? If you’re looking for the criminal that killed a bunch of people but not that specific case it’s not that uncommon. If you’re looking for a squeaky clean guy that was there at the wrong place at the wrong time it’s very rare. Most criminals are repeat offenders.

I believe, and I think the data supports, that the US prison population is due to the number of laws we have, and the harshness of our penalties. A false conviction rate of a few percent does not explain an order of magnitude level differences in incarceration between countries. I think an interesting question is what portion of the US population would be incarcerated if the totality of our laws were applied to every…

US law is applied unequally. It’s very harsh, sometimes random and I’ve heard all Americans break the law daily. We would all be in jail and sometimes prison.

US crime is like driving across the only road that is covered in black ice with toll trolls on the way, you have to cross it, you know it’s there, and you can do it carefully but you never know what will happen or where you’ll slip up.

Re: Bad science and bad statistics in the courtroom convict innocent people

#35

Post by gruez »

adamsb6 wrote: post #17

Even photo and video evidence suffers from courtrooms that don’t understand how nearly all cameras will record and encode images. JPEG encoding (and thus MPEG encoding) takes 8x8 blocks of pixels and represents them as a discrete cosine transform. In a very real way these pixels don’t actually exist. If your video evidence relies on such a small portion of MPEG encoded video, you can easily mislead a jury to convicti…

>I’m reminded of the Rittenhouse trial where the prosecutors had a technician massage a still from a video for hours until some very tiny collection of pixels, enlarged from the original by quite a bit, were arranged in such a way that they said demonstrated Rittenhouse initiated aggression against the first man he shot.

Is there an article with more on this?

Re: Bad science and bad statistics in the courtroom convict innocent people

#36

Post by graphe »

WalterBright wrote: post #29

I read about a case once where a man was convicted of rape. The evidence was that the rape was done face-to-face with the victim, and the victim positively ID'd him. How could she not, when she was face-to-face with him? It was a slam dunk. After some years in prison, the convicted person managed to get a DNA test, which proved he was not the rapist. How could this be? It turns out the DNA matched a convicted sex off…

This indicates that the jury/judge also failed the innocent man. What would have saved him? Are these cases relics before dna? Would it happen again today?

Re: Bad science and bad statistics in the courtroom convict innocent people

#37

Post by swells34 »

causal wrote: post #28
tiahura wrote: post #21

Earlier quoted context omitted.

Being primarily a civil attorney, I'm well aware of the disproportionate number of former prosecutors on the bench. A couple of points: 1. Most prosecutors are liberals. It's not like right wing law students go looking for jobs in prosecutor's offices. I can't think of any from my class way back when. 2. The problem is more that judges and prosecutors are old colleagues. The slant is that they all know the defendants…

Thanks for that insight. My discomfort is with the notion that they could know that the defendant is guilty- sure there may be some obvious cases, but I do not trust most humans to differentiate their intuition from hard evidence.

And unfortunately, judges are the only legal entity empowered to use their intuition for legal decision making. Sentencing is a good example; judges can kinda just decide sentences based on feel. Just throw some number out there.

Re: Bad science and bad statistics in the courtroom convict innocent people

#38

Post by andrewla »

albert_e wrote: post #3

I was thinking about this just today randomly. Every bit in computer memory or data store is just a bit flip away from being completely wrong. In other words all data corruption -- unintentional or deliberate-- can be statistically shown as a bunch of off-by-one errors. Of course the chance of multiple random off-by-one errors lining up exactly to help one party is astronomically low. But would a typical court of law…

I don't think this categorization is technically correct, and even in the places where it is partially correct I don't see that the "off-by-one" phrasing adds any value.

Off-by-one errors are a very specific type of error in algorithm implementation, where a closed boundary and an open boundary are confused. A general bit flip is not really an example of this unless it is connected to this specific kind of mistake.

And data corruption is much broader, and can involve adding or removing data from a stream in addition to mutating existing data. Saying that all data mutations are a "bunch of off-by-one errors" is just asserting that there exists a hamming distance between any two binary representations of the same length, and I don't think this adds a lot of insight.

On the other hand, using the term "off-by-one error" carries with it a lot of baggage and connotations, most of which do not apply to scenarios you have described or general data corruption.

Re: Bad science and bad statistics in the courtroom convict innocent people

#39

Post by twisteriffic »

Lord-Jobo wrote: post #12

Wow, a hackernews topic thats actually directly related to my specialty. Anyway, this article is spot on and this is a really rampant specific issue. “expert witnesses have often overstated the probative value of their evidence, going far beyond what the relevant science can justify.” This is the absolute center of the issue. Experts are used by the prosecution AND defense, but the experts used by the prosecution are…

Wondering how you got in to that field?

A few years back I was asked to informally provide a technical review the prosecution's evidence in a case involving a former colleague. The police expert's sole credentials were a 10 year old A+ certification and a few employer-sponsored day courses, but was confidently representing themselves as an expert in networking/computing/technology. Their theory of the case was brutal - filled with nonsensical jargon, fabrications and misrepresentations. Things like "a computer behind a NAT is unreachable from the Internet under any circumstances". Instances where timestamps were implied to correlate but were actually 6+ hours apart once time zones were factored in. Pages and pages of things in that vein.

None of that mattered - they were convicted anyway. It broke my faith in the justice system. The prosecution behaved in exactly the same way as today's legions of right-wing conspiracists, grifters and politicians do, because the truthfulness of their words didn't matter to them. Because the defense cared about the accuracy of what they were stating, they could be easily drowned out by confident, rapid bullshit and innuendo.

I desperately want to find a way to help make the system better, but at every avenue the response has been "it's your funeral".

Re: Bad science and bad statistics in the courtroom convict innocent people

#40

Post by sneak »

Zigurd wrote: post #13
causal wrote: post #8

Earlier quoted context omitted.

I've always been perplexed by the incentive to get convictions. This is guaranteed to result in prosecutors preferring any method with a lot of false positives.

Separating evidence processing from law enforcement would be a good step forward.

I think the system of American policing was originally cooked up to enforce the existing social hierarchy via selective enforcement, and it appears to remain mostly unchanged because it seems to continue to serve that same purpose.

Why would people with power to change it do so, when it is working as intended? Prosecutors don’t bring charges against police and judges and senators and large landowners/rightsholders (or the DNI).

Post reply on HN