Why aren't trials based on Bayesian reasoning?
Because it is illegal
http://understandinguncertainty.org/court-appeal-bans-bayesian-probability-and-sherlock-holmes
>>7586563
>only 100 total DNA-based suspects in the city
Great job, you just increased the odds he is guilty from 1-in-10,000 to 1-in-100
Criminal cases are based on reasonable doubt, no the balance of probabilities. Probability estimates are not evidence. They're supposition, circumstantial at best. Given the ability for so-called experts to get the calculations wrong and for jurors to misinterpret results, it should be used very carefully if at all (that case of the Dutch nurse prosecuted for murdering patients is another example – see Bad Science).
This is not anti-science or anti-math. It’s recognizing what probability theory is actually telling you. After all, banks use very sophisticated probability/statistical analysis to evaluate financial risk. The use of these approaches contributed to the banks thinking their financial instruments were low risk on the downside, and yet the downside happened anyway. You want your guilt evaluated in the same way (even partially)?
>>7586563
Prior data tells me P( guilty | on trial) is very high.
Thus, unless I am presented with substantial evidence to the contrary, I'm voting guilty.
>>7586815
What prior data? We don't even know if courts are better at proving guilt than a coin flip.
Criminal law is based on doubt, not probability. Saying how likely it is someone committed a crime isn't really all that useful.
>>7586563
Since when did law courts ever use reasoning?
I once successfully defended myself in court using similar reasoning. (Fortunately, it wasn't the sort of trial that required DNA testing. And no, I didn't do it.)
The key point is that the faulty stats derive from an assumption of guilt. If you are assumed to be guilty, then a positive DNA test reinforces that. But if instead you are assumed to be innocent until proven otherwise, then the test doesn't provide much evidence.
>>7586588
No, you have it backward. Using faulty stats, it was likely that he was guilty by a factor of 10000. Using proper logic, it was unlikely that he was guilty by a factor of 100.
>>7586815
I hope you end up on trial some day, with someone like you on the jury.
>>7586638
The law and legal system does not determine truth. It determines who wins the case/trial.
>>7587597
exactly, and it is the same in math/logic
>>7587516
>No, you have it backward. Using faulty stats, it was likely that he was guilty by a factor of 10000. Using proper logic, it was unlikely that he was guilty by a factor of 100.
You are saying exactly the same thing but wording it differently.
What do courts decide, if you are innocent or not innocent? No, they decide if you are guilty, or not guilty.
A 1-in-100 chance of being guilty means you are 99% likely to be innocent, but the OP still decreased the odds from 1-in-10,000 (99.99% innocent).
It's sad that truth is not what is determined in a courtroom. No justice. Just us.
>>7586570
>Anyway, I teach the Bayesian approach to post-graduate students attending my 'Applied Bayesian Statistics' course at Cambridge, and so I must now tell them that the entire philosophy behind their course has been declared illegal in the Court of Appeal. I hope they don't mind.
Hahaha! Also, settings -> Quotes and Replies -> Legacy Captcha
>>7587516
Shit, I ( >>7586588 and >>7588431 ) just reread OP.
I kinda defaulted to seeing it as a 1:10,000 chance of being guilty because the logic is so flawed in the first place.
A sequence occurring once in 10k doesn't mean anything, what if there is an abundance of people with that sequence who all live in one area?
Or you know, relatives of the accused.
Jesus fuck has anyone actually been prosecuted under this "reasoning"?
