Fighting murder convictions that rest on shoddy stats
science.org
science.org
While the suspicion had lingered for a decade or two that sumo matches were rigged, the analysis of the paper precipitated a scandal years later that led to the cancellation of a sumo tournament in 2011, and the expelling of 23 wrestlers.[2]
[1] https://web.stanford.edu/~mgduggan/Research/MD_Levitt_2002.p...
[2] https://en.wikipedia.org/wiki/Match-fixing_in_professional_s...
If anyone has any know-how to interpret the following diagram, I’m open to some help here. Sorry if it annoys you to hear a devil’s advocate defending a potential murderer, but what if said nurse was just incompetent and people took advantage of her? (Or she was the primary person willing/forced to work shifts known to be ripe with gang activity.)
I would guess that reality is closer to "the carb heavy food pyramid turned out to be total hogwash and had negligible impact because almost nobody cares what the government wants them to eat".
It’s not about what happens when you have a choice on what to eat but what information official sources use when you are reliant on another authority for meals.
Edited to add that: it’s important to know why the food pyramid was created and the context of them. Government provided nutrition guidelines were used for meal provision during the Great Depression, food rationing during war, etc. and are still largely created by the USDA which answers to agribusiness and not medicine. (There are several news articles of the USDA actively withdrawing and editing nutritional guidelines due to agribusiness outcry!)
They definitely fried food in oil and lard rather than delicately avoiding fats. No sense wasting good bacon grease when you can toss some eggs in there. And some bread too. Nothing like toast all sopped up in bacon grease.
For anyone saying citizens didn't listen to these experts, they've apparently never seen the magazine rack in the grocery store checkout. Or any of the millions of lifestyle articles and weight loss blogs informing desperate folks to run from anything with butter in it. I don't think my parents had a real stick of butter in the house until after I graduated high school. Low fat everything in that fridge until they saw me lose weight and gain so much more energy by eating low carb and high fat.
When you've got teachers in the classroom teaching the food pyramid to elementary kids, you're going to have some kind of impact on the way people eat.
In 1980, I was twenty-five, and had never lived in a household with a deep fat fryer. Occasionally I guess we had fried chicken from my grandmother's kitchen.
Spoiler: https://cms-static.wehaacdn.com/hoards-com/images/Obesity-19...
(Also, your graph doesn't show obesity, and the pre-pyramid 45% rate would already be alarmingly high if it was obesity, which it isn't.)
For the sake of balance (in these comments), sadly some similar incidents have been very real and the wheels of justice very slow: https://www.texasmonthly.com/true-crime/how-san-antonio-baby...
In a report peer reviewed and distributed by the Royal Statistical Society (RSS) in September 2022, Gill and colleagues detailed the statistical missteps in past medical murder trials and made recommendations for how legal systems can do better. Gill hopes the report will help with the case of another British nurse, Lucy Letby, who is now on trial for the alleged murder of seven babies and attempted murder of 10 more in a neonatal unit at the Countess of Chester Hospital.
It seems this was in fact the case.
I agree with your broader point however.
0. https://www.independent.co.uk/news/uk/crime/lucy-letby-nurse...
The evidence in her case does seem too circumstantial to convict, a real injustice.
Unfortunately it is impossible to explain to people that a confession is meaningless.
When a Dutch nurse named Lucia de Berk...
Important evidence at the appeal was to be the statement of a detainee in the Pieter Baan Center, a criminal psychological observation unit, where de Berk had said during outdoor exercise, "I released these 13 people from their suffering". However, during the appeal, the man withdrew his statement and stated that he had made it up.Haven't you ever read The Green Mile?
Let’s say someone who’s mentally unwell wrote notes to confess to crimes, is it enough to convict them based on that evidence alone? And if so, should it be?
So if they find a body with multiple stab wounds and I go in and say I stabbed the person, I can be convicted. If I say I murdered Audrey Farber, and the police can't even find a record of an Audrey Farber gone missing, then I can't be convicted because there's no evidence than an Audrey Farber was murdered at all.
You mean you don't know for sure? Do you want the investigators to stop investigating if a suspect "sounds" mentally unwell? Should prosecutors refuse to prosecute if the suspect "sounds" mentally unwell?
What I'm saying is ... why, in this particular case, should that make any difference to the prosecution or investigation? We already find mentally troubled people guilty, we just sentence them differently.
So, yeah, the sanity/insanity of an individual makes no difference to whether they are guilty or not, it only changes the specific nature of the charge (i.e. premeditated vs culpable) and the sentencing (asylum or prison).
Someone is is genuinely mentally unwell in a way that results in them harming other people must still be kept away from society!
IOW, people who are dangerous to others need to be locked up. Whether they are insane or not is not relevant to keeping them locked up, it is only relevant to where they are locked up, and how they may be rehabilitated.
> Let’s say someone who’s mentally unwell wrote notes to confess to crimes, is it enough to convict them based on that evidence alone?
No one is currently convicted on the basis of a confession alone[1]. Normally a confession just means that the investigation into the confessor is more thorough than it would ordinarily be.
Is it perfect? No, but it is a lot better than you appear to believe.
I think if you thought about it for more than a few seconds, you'd realise the questions you are asking have obvious answers.
[1] Have you any idea how many people claimed to be famous serial killers? it happens more often than you think. It's also why specifics of a crime may not make it to the news, because the police use those specific details to identify those confessors who are not the perpetrator.
> Someone is is genuinely mentally unwell in a way that results in them harming other people must still be kept away from society!
That is extremely poor and dangerous argument for putting them in jail in bad evidence. But yeah, when innocent people get into jail, it is fairly often on arguments like this.
> No one is currently convicted on the basis of a confession alone[1]. Normally a confession just means that the investigation into the confessor is more thorough than it would ordinarily be
Bases in innocence project, people who were provably not guilty were sentenced to death on confessions. And there is literally zero reason to think it is stopped happening.
> That is extremely poor and dangerous argument for putting them in jail in bad evidence.
Who made that argument?
Here's what I said:
>> the sanity/insanity of an individual makes no difference to whether they are guilty or not, it only changes the specific nature of the charge (i.e. premeditated vs culpable) and the sentencing (asylum or prison). Someone is is genuinely mentally unwell in a way that results in them harming other people must still be kept away from society!
Why on earth would you snip away my text about ASYLUM, and instead warble on about putting those people in JAIL?
I'm genuinely curious about the motivation to pretend that my argument is different to what I said, and then argue against your pretend version of my argument.
Genuinely curious.
The submission on HN is coincidence. I don't know about the publication of the article.
These forensic labs have their own brand of "Never the twain shall meet" while they do blind analysis especially when the lead crime investigator has to clue them in ("fentanyl", in one case).
When doing network-based forensic, just because a "ping" happened, doesn't mean that "causation equal correlation".
Worse is the misuse of end-users' search history; a lone query made 7 years ago being used as a lead-in (but lone) evidence to a suspect's alleged behavior. Really?
Or an innocuous Google Scholar watch doing a trapping of some suspected but heavily watched URL?
One simple lab exercise of such forensic analysis is a reconstruction on the chaining of events given a set of evidence; it gets hilarious at times to see such "conspiratorial-like" theories.
Now the world witness the continuing saga of misusage of statistics from the US FDA VARS database being flooding with death by COVID inoculants.
There could very well be similar bias and shoddy statistics at play there
I'm not saying the guy is a saint, but low-effort cheating in low-stakes online games by looking at an engine in the background seems substantially different than what would be required to cheat completely undetectably in major tournaments with many people watching.
On the one hand, I'm inclined to agree with the assertion that dishonesty damages one's credibility.
On the other hand, there's degrees of dishonesty - should someone who (intentionally lied, and) denied going over the speed limit be automatically considered lying when he also denies premeditated murder?
This person cheated in an exam in middle school so shouldn't be trusted with exams in med school.
Because, yeah, when he cheated he was a kid. Kids frequently do stupid things.
In any case, do these legal systems not presume innocence until actually proven guilty beyond a reasonable doubt? If not, then these cases seem like perfect case studies as to why said presumption of innocence is mandatory for a society to be considered free and just - and if they do presume innocence, then I'm baffled by how these cases not only resulted in convictions but upheld convictions given the abundance of reasonable doubt.
1. Isn't 1 in 49 still hugely statistically significant and therefore suspicious? The probability of this happening by chance is still only 2%
2. What were De Berk's the "great secrets" and "secret compulsion"? Was there any other evidence presented against her besides the raw statistics?
3. In the case of Folbigg I just wonder: Why would she have so many kids if all she does in her diary is complain about their existence ... ??
Imagine someone goes to a casino, and play a game where you flip a coin 20 times in a row, and win if they're all heads. There's about a 1 in 1 million chance of flipping 20 heads - so people are suspicious that the person cheated.
However, imagine that during that same time period, 10 million people played that game. Looking at that, you'd expect there to be 10 winners of the game. Suddenly it's not suspicious at all.
So I think the argument is that this is the chance a nurse experienced this many deaths is 1 in 45 - so with 45 nurses, you'd expect one to experience it.
2% means a false positive for every 50 nurses.
People aren’t willing to acknowledge the high probability that someone just had a string of bad partners
Nonsense. Choosing a partner is not a random event, like drawing a card from a deck, or spinning a roulette wheel.
The odds are very much against "poor guy/gal had 10 boyfriends/girlfriends in the last 10 years, all of whom turned out to be meth-heads" being a random event.
In relationships, what is true most of the time (the odds on) is that a string of bad relationships for a man/woman has one common factor - that man/woman.
Over 5 years?
Two is not "a string of bad partners".
Three meth-heads in a row is almost certainly NOT "very plausible" (probability approaching 1). The subject is almost certainly selecting for meth-heads.
Three alcoholics in a row over 5 years is also not "very plausible", but it is at least somewhat plausible - alcohol addiction can be hidden very well from someone you don't live with.
Three cheaters in a row (over 5 years) is maybe what I would consider to be approaching "very plausible". Even then, at around 20 months per relationship (the most optimistic way to have each relationship be long), you gotta wonder why the subject is choosing only players to date. Still looks like a selection bias on the subjects side.
I guess what I'm trying to say is that your original assertion could not be any more wrong than it already is.
Someone with a string of bad relationships most probably has a selection bias for certain traits in a partner which make them unsuitable as a partner.
So at the least the person with a string of bad partners is very, very bad and choosing romantic partners.
[1]: https://podcasts.apple.com/us/podcast/wrongful-conviction/id...
(This is distinct from the problem of qualified immunity for the police, who (1) regularly kill people under dubious at best circumstances, and (2) regularly launder falsified or coerced evidence into the legal system. It makes sense, in my opinion, to introduce personal liability there.)
I’m curious what your “Blackstone number” is. The Blackstone principle is that it is better for ten guilty men to go free than one innocent one to be imprisoned. It seems you might go higher than ten—but how high? Should we be so concerned about wrongful convictions that we allow 1000 guilty men to go free for each one? 10,000?
Then, turn that around. Given how often prosecutors have good motives and adequate grounds for moving forward with a conviction, how many innocent prosecutors would you put through the wringer just to catch one guilty one?
It seems like you have an expectation that human systems could be perfect, and I’m curious where that comes from or what sustains it.
If you want 0% chance for innocent people to get convicted, you have to let many guilty people go.
If you can drive down the rate of innocent people punished below, say, 1 in a trillion, that's probably close enough to zero in practice?
The problem is that we are not achieving the blackstone number more than 1/10th of people in prison are innocent.
That means the criminals responsible for those crimes are still out there
> That means the criminals responsible for those crimes are still out there
That's more than a little concerning. Link?
This is incoherent; it is obvious that you'd want to set this differently for different crimes.
We can also ask for certainty that people who are guilty in fact get convicted. ("It's bad when guilty men go free.") This is the opposite of the other kind of certainty. Here, we'd want a lot of that type of certainty for crimes with severe punishments (we assume that such crimes are very serious), and we don't really care about this for crimes with no particular impact on victims. So we want more certainty-of-conviction in case 1 and less certainty-of-conviction in case 2.
Certainty-of-conviction and certainty-of-guilt are, in the general case, directly in conflict with each other. We don't want either (in a relative sense) for trivial crimes, so it doesn't really matter what happens. We want lots of both for serious crimes, so what happens matters, but we can't actually achieve both goals.
In conclusion, when setting your policy, you need to look at other aspects of the crime than the one(s) you mention here, which are not informative as to this question. (Though they are informative as to the question you literally asked; we want more certainty of every type on serious crimes.) Perhaps the punishment for a particular crime is out of proportion to how serious that crime is. If the punishment is too heavy, we'd want more certainty-of-guilt; if it's too light, we'd want more certainty-of-conviction. Perhaps it's much easier than usual to reliably determine guilt-in-fact for a particular crime. Then we'd want more certainty-of-conviction.
There is nonsuch thing. Even short conviction destroys people's ability to hold jobs, find housing and to find partners. People inyour life stop trusting you and that is super big one too. And it may cause divorce with law abiding lartner and factor into custody dispute.
Any Criminal conviction has huge ramifications.
The relevance of a Blackstone number is it basically sets all this aside and looks at an ideal circumstance, and you can adjust from there, if you want to take peculiarites into account.
Of course they should have tried to correct the mistake once pointed out, but the justice seems doesn't seem to like doing that.
Putting prosecutors on trial for something less than clear prosecutorial over-reach is, in my view, a major mistake. Perhaps the bar should be lower than it is, but there's no way it should be as low as "didn't get a conviction" or "overturned on appeal".
(And I see the strain between this comment and my previous one. If the prosecutor has discretion, then they have responsibility for how they use it. But the bar for coming after the prosecutor needs to be considerably higher than just a "not guilty" verdict. The prosecutor probably should be more responsible for abusing their discretion than they are. But when the pendulum moves, it often moves too far.)
A job which also includes prosecutorial discretion.
IOW, it's also literally their job to decide which cases to pursue and which to drop.