I don’t mean to single you out particularly but your probabilities show a pattern common to several other commenters that I find surprising. What is your prior probability that a rape-murder would be committed by three people, a man and a woman who had been dating for a couple of weeks and a third man who was a stranger to both of them, rather than by a single man acting alone?
My prior probability for three people acting together like this would be extremely low. The fact that Knox and Sollecito had only been dating for a couple of weeks doesn’t seem to be in dispute, nor does the fact that they did not previously associate with Guede. I’m surprised that people don’t revise down their probability of Knox and Sollecito being guilty significantly if they believe Guede is guilty.
The first site below suggests US wrongful conviction rates range from .5 percent to 10 percent. It cites for the lower rate a source who I think is the author of the next URL:
I would be very surprised if Italy’s legal system turned out to have a significantly worse rate.
“Convicts eventually proven innocent” is, sadly, bound to be a lower fraction than wrongful conviction rate—i.e. you get wrongful conviction rates by extrapolating one way or another from attested cases.
Thanks for the link. I get that “Convicts eventually proven innocent” is most likely a lower bound (probably not too many guilty ones later exonerated?), but I figured I’d have to work from there to get a crude guess.
On one hand, even 10% isn’t all that bad in an absolute sense- most of the deterrent is being had with not too much additional waste. On the other, that means that our trial and jury system is probably worse than I thought, if it’s true that “most” people charged with a crime plead guilty.
I’ll look up the statistics and report back in a couple days.
I’m inclined to believe that in general it’s very strong—that if all you know about X is that they were convicted of rape and murder, then the likelihood that they raped and murdered someone is vastly greater.
In this particular case, adding it to the other things I’ve skimmed about it, I’m coming to something like a .20/.20/.70 estimate, but that’s after reading the comments here.
I’m inclined to believe that in general it’s very strong—that if all you know about X is that they were convicted of rape and murder, then the likelihood that they raped and murdered someone is vastly greater.
Yes...but that’s never all you know, unless you visit a prison. Otherwise, something has drawn your attention to the particular person, which is a lot of information. In particular, my prior for the innocence of people whose conviction is a cause celebre is 75%.
It seems even harder to estimate the rate of “cause celebre wronful convictions” than to estimate wrongful convictions in general.
Moreover, I’d be concerned that this particular reference class leaves you vulnerable to availability bias. You’re more likely to remember cases where a convicted person was eventually proven innocent; we never see newspaper headlines proclaiming “Conviction of X still not overturned”.
The one big example that comes to my mind of a cause celebre conviction that turned out to be proper is Alger Hiss, though I note from Wikipedia that many still dispute his guilt.
Also, in general, the Communist Party USA really was trying to get secret allies in high government positions and really was under the control of Soviet intelligence.
I actually can’t remember how much non-leftist resistance the red scares got while they were going on. But certainly after the fact they have been portrayed as merely irrational, politically motivated witch hunts.
Edit: And there were definitely members and allies of CPUSA who had no idea they were part of or working with a Soviet front and defended the Party because of that belief. CPUSA did whole lot of solid civil rights organizing—work that no one else (at least no other groups with significant white participation) was doing at the time. When I say the CPUSA was a Soviet front that doesn’t mean that it was exclusively a Soviet front.
The fact that my first two probabilites were as high as they were has more to do with the minimal time spent and tired mental state I was in (eg I had trouble even keeping the relevant evidence in my short term memory) and fully expected later estimates based on further processing the same information to be much more extreme.
I did significantly revise my probability due to low priors on that sort of 3 person crime once I decided Guede was guilty, but I didn’t trust myself enough at the time to put a prior on it that was that low nor really keep track of evidence.
Still though, I probably could have guessed which side the answer would fall on and should have put something somewhat lower.
I had a quick look for some statistics to verify my priors that both female on female murder and multiple offender murder are unusual. These data seem relevant: gender; multiple offenders. These appear to back up my intuition that finding a very likely male suspect (Guede) should have greatly reduced the prosecution’s odds for the guilt of a female suspect (Knox) in a multiple offender attack. A plausible case for a male offender committing the crime alone should greatly lower the probability of guilt of a female suspect acting either alone or as an accomplice.
Agreed. The only reason that this might not drive down the probability of Knoxs guilt is that there’s some chance that the system is behaving somewhat rationally, and the fact that they were convicted is evidence that there’s information that you don’t have.
Of course, once you hear all the prosecutors arguments, that goes away.
That brings up an interesting question though. If the base rate is low enough and base rate neglect is common enough, maybe you really can confidently claim that Knox is innocent even though she was convicted given only the fact that Guede is guilty and plausibly could have done it himself.
If the base rate is low enough and base rate neglect is common enough, maybe you really can confidently claim that Knox is innocent even though she was convicted given only the fact that Guede is guilty and plausibly could have done it himself.
I’d love to see an article doing the research to make this claim. At the very least, I’d be entertained.
Sure this type of murder is rare, otherwise it wouldn’t have made the news. That’s not evidence the other two aren’t guilty. Not when we have her hand on the knife and his foot in her blood.
The Wikipedia page seems to be changing quite a lot at the moment. The impression I got from the various sources was that Guede was acquainted with the people who lived on the 1st floor of the building (not house mates but neighbours of Knox and the victim) and Knox said she recognized him but that Sollecito claimed never to have met him. I haven’t seen any claims that they were friends but there was an eye witness who claimed to have seen them together. There seems to be some dispute over whether they had ever previously met or talked with each other but no claim from the prosecution that there was any kind of longer term association between them prior to the murder.
I don’t mean to single you out particularly but your probabilities show a pattern common to several other commenters that I find surprising. What is your prior probability that a rape-murder would be committed by three people, a man and a woman who had been dating for a couple of weeks and a third man who was a stranger to both of them, rather than by a single man acting alone?
My prior probability for three people acting together like this would be extremely low. The fact that Knox and Sollecito had only been dating for a couple of weeks doesn’t seem to be in dispute, nor does the fact that they did not previously associate with Guede. I’m surprised that people don’t revise down their probability of Knox and Sollecito being guilty significantly if they believe Guede is guilty.
The fact that they were convicted is also evidence, of course.
How strong a piece of evidence do you think it is?
Before or after conditioning on the rest of the available information?
If before, there’s gotta be statistics out there. What fraction of people charged plead guilty? What fraction that plead not guilty are convicted?
What fraction of convicts are eventually proven innocent?
The first site below suggests US wrongful conviction rates range from .5 percent to 10 percent. It cites for the lower rate a source who I think is the author of the next URL:
http://www.caught.net/innoc.htm
http://www.abanet.org/crimjust/spring2003/conviction.html
I would be very surprised if Italy’s legal system turned out to have a significantly worse rate.
“Convicts eventually proven innocent” is, sadly, bound to be a lower fraction than wrongful conviction rate—i.e. you get wrongful conviction rates by extrapolating one way or another from attested cases.
Thanks for the link. I get that “Convicts eventually proven innocent” is most likely a lower bound (probably not too many guilty ones later exonerated?), but I figured I’d have to work from there to get a crude guess.
On one hand, even 10% isn’t all that bad in an absolute sense- most of the deterrent is being had with not too much additional waste. On the other, that means that our trial and jury system is probably worse than I thought, if it’s true that “most” people charged with a crime plead guilty.
I’ll look up the statistics and report back in a couple days.
I’m inclined to believe that in general it’s very strong—that if all you know about X is that they were convicted of rape and murder, then the likelihood that they raped and murdered someone is vastly greater.
In this particular case, adding it to the other things I’ve skimmed about it, I’m coming to something like a .20/.20/.70 estimate, but that’s after reading the comments here.
Yes...but that’s never all you know, unless you visit a prison. Otherwise, something has drawn your attention to the particular person, which is a lot of information. In particular, my prior for the innocence of people whose conviction is a cause celebre is 75%.
How do you get that figure ?
It seems even harder to estimate the rate of “cause celebre wronful convictions” than to estimate wrongful convictions in general.
Moreover, I’d be concerned that this particular reference class leaves you vulnerable to availability bias. You’re more likely to remember cases where a convicted person was eventually proven innocent; we never see newspaper headlines proclaiming “Conviction of X still not overturned”.
The one big example that comes to my mind of a cause celebre conviction that turned out to be proper is Alger Hiss, though I note from Wikipedia that many still dispute his guilt.
Also, in general, the Communist Party USA really was trying to get secret allies in high government positions and really was under the control of Soviet intelligence.
Genuine question: were there a lot of people who were not supporters of the CPUSA but who argued that one or both of these were not true?
I actually can’t remember how much non-leftist resistance the red scares got while they were going on. But certainly after the fact they have been portrayed as merely irrational, politically motivated witch hunts.
Edit: And there were definitely members and allies of CPUSA who had no idea they were part of or working with a Soviet front and defended the Party because of that belief. CPUSA did whole lot of solid civil rights organizing—work that no one else (at least no other groups with significant white participation) was doing at the time. When I say the CPUSA was a Soviet front that doesn’t mean that it was exclusively a Soviet front.
I’d be interested to read more about this if you have any good pointers—thanks!
The fact that my first two probabilites were as high as they were has more to do with the minimal time spent and tired mental state I was in (eg I had trouble even keeping the relevant evidence in my short term memory) and fully expected later estimates based on further processing the same information to be much more extreme.
I did significantly revise my probability due to low priors on that sort of 3 person crime once I decided Guede was guilty, but I didn’t trust myself enough at the time to put a prior on it that was that low nor really keep track of evidence.
Still though, I probably could have guessed which side the answer would fall on and should have put something somewhat lower.
I had a quick look for some statistics to verify my priors that both female on female murder and multiple offender murder are unusual. These data seem relevant: gender; multiple offenders. These appear to back up my intuition that finding a very likely male suspect (Guede) should have greatly reduced the prosecution’s odds for the guilt of a female suspect (Knox) in a multiple offender attack. A plausible case for a male offender committing the crime alone should greatly lower the probability of guilt of a female suspect acting either alone or as an accomplice.
Agreed. The only reason that this might not drive down the probability of Knoxs guilt is that there’s some chance that the system is behaving somewhat rationally, and the fact that they were convicted is evidence that there’s information that you don’t have.
Of course, once you hear all the prosecutors arguments, that goes away.
That brings up an interesting question though. If the base rate is low enough and base rate neglect is common enough, maybe you really can confidently claim that Knox is innocent even though she was convicted given only the fact that Guede is guilty and plausibly could have done it himself.
I’d love to see an article doing the research to make this claim. At the very least, I’d be entertained.
Sure this type of murder is rare, otherwise it wouldn’t have made the news. That’s not evidence the other two aren’t guilty. Not when we have her hand on the knife and his foot in her blood.
The wikipedia article states that Guede was known to the couple and to the others in their house.
The Wikipedia page seems to be changing quite a lot at the moment. The impression I got from the various sources was that Guede was acquainted with the people who lived on the 1st floor of the building (not house mates but neighbours of Knox and the victim) and Knox said she recognized him but that Sollecito claimed never to have met him. I haven’t seen any claims that they were friends but there was an eye witness who claimed to have seen them together. There seems to be some dispute over whether they had ever previously met or talked with each other but no claim from the prosecution that there was any kind of longer term association between them prior to the murder.