As many of you probably know, in an Italian court early last weekend, two young students, Amanda Knox and Raffaele Sollecito, were convicted of killing another young student, Meredith Kercher, in a horrific way in November of 2007. (A third person, Rudy Guede, was convicted earlier.)

If you aren't familiar with the case, don't go reading about it just yet. Hang on for just a moment.

If you are familiar, that's fine too. This post is addressed to readers of all levels of acquaintance with the story.

What everyone should know right away is that the verdict has been extremely controversial. Strong feelings have emerged, even involving national tensions (Knox is American, Sollecito Italian, and Kercher British, and the crime and trial took place in Italy). The circumstances of the crime involve sex. In short, the potential for serious rationality failures in coming to an opinion on a case like this is enormous.  

Now, as it happens, I myself have an opinion. A rather strong one, in fact. Strong enough that I caught myself thinking that this case -- given all the controversy surrounding it -- might serve as a decent litmus test in judging the rationality skills of other people. Like religion, or evolution -- except less clichéd (and cached) and more down-and-dirty.

Of course, thoughts like that can be dangerous, as I quickly recognized. The danger of in-group affective spirals looms large. So before writing up that Less Wrong post adding my-opinion-on-the-guilt-or-innocence-of-Amanda-Knox-and-Raffaele-Sollecito to the List of Things Every Rational Person Must Believe, I decided it might be useful to find out what conclusion(s) other aspiring rationalists would (or have) come to (without knowing my opinion).

So that's what this post is: a survey/experiment, with fairly specific yet flexible instructions (which differ slightly depending on how much you know about the case already).

For those whose familiarity with the case is low:

I'm going to give you two websites advocating a position, one strongly in favor of the verdict, the other strongly opposed. Your job will be to browse around these sites to learn info about the case, as much as you need to in order to arrive at a judgment. The order, manner, and quantity of browsing will be left up to you -- though I would of course like to know how much you read in your response.

1. Site arguing defendants are guilty. 

2. Site arguing defendants are innocent.

I've chosen these particular sites because they seemed to contain the best combination of fierceness of advocacy and quantity of information on their respective sides that I could find. 

If you find better summaries, or think that these choices reflect a bias or betray my own opinion, by all means let me know. I'm specifically avoiding referring you to media reports, however, for a couple of reasons. First, I've noticed that reports often contain factual inaccuracies (necessarily, because they contradict each other). Secondly, journalists don't usually have much of a stake, and I'd like to see how folks respond to passionate advocacy by people who care about the outcome, as in an actual trial, rather than attempts at neutral summarizing. Of course, it's fine if you want to read media reports linked to by the above sites.

(One potential problem is that the first site is organized like a blog or forum, and thus it is hard to find a quick summary of the case there. [EDIT: Be sure to look at the category links on the right side of the page to find the arguments.] If you think it necessary, refer to the ever-changing Wikipedia article, which at the moment of writing seems a bit more favorable to the prosecution. [EDIT: I'm no longer sure that's true.] [EDIT: Now I think it's true again, the article having apparently changed some more. So there's really no telling. Be warned.])

After you do this reading, I'd like to know:

1. Your probability estimate that Amanda Knox is guilty.
2. Your probability estimate that Raffaele Sollecito is guilty.
3. Your probability estimate that Rudy Guede is guilty.
4. How much you think your opinion will turn out to coincide with mine.

Feel free to elaborate on your reasoning to whatever degree you like.

One request: don't look at others' comments until you've done the experiment yourself!

For those whose familiarity with the case is moderate or high:

I'd like to know, as of right now:

1. Your probability estimate that Amanda Knox is guilty.
2. Your probability estimate that Raffaele Sollecito is guilty.
3. Your probability estimate that Rudy Guede is guilty.
4. How much you think your opinion will turn out to coincide with mine.
5. From what sources you've gotten the info you've used to arrive at these estimates.

Then, if possible, do the experiment described above for those with little familiarity, and report any shifts in your estimates.

Again, everyone should avoid looking at others' responses before giving their own feedback. Also, don't forget to identify your prior level of familiarity!

If the level of participation warrants it, I'll post my own thoughts (and reaction to the feedback here) in a later post. (Edit: That post can be found here.)


New Comment
266 comments, sorted by Click to highlight new comments since: Today at 3:01 AM
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

On a cursory reading of Wikipedia the obvious interpretation is that Knox and Sollecito are innocent and Guede is guilty. I didn't go through all the sites so I don't know if this would qualify as a litmus test, and assigning probabilities in this state of knowledge would be extra work.

EDIT: Read comments and am surprised at how many estimates of "Knox and Sollecito probably didn't do it" have probabilities in the range of 40% attached that they did. If it were a binary judgment or a confidence interval, then yes we should avoid extreme probabilities and widen intervals to compensate for known overconfidence biases. But in this case the hypothesis space of equally plausible possibilities is large, and some low-probability symbols were used to write the message (multi-person rape-murder vs. single-person rape-murder, female rape-murder vs. male rape-murder). It may not always be easy to unravel crime scenes (though this one sounds pretty straightforward) but to focus on Knox or Sollecito seems like privileging the hypothesis.

Unless there's major prosecutorial evidence not in Wikipedia, then this seems like a case of paying too much attention to other people's opinions (the jury, in a case where we have further information that the verdict gained media attention as possibly inaccurate), and I would suggest that anyone who gave a probability higher than .15 be more arrogant in the future.

If, of course, I just haven't done enough reading, then ignore the above.

I would suggest that anyone who gave a probability higher than .15 be more arrogant in the future.

This does not mean my assigned probability was 15%. It means, "Even after accounting for fudge factors and people using different numbers to express similar emotional degrees of certainty, if you gave a number higher than FIFTEEN PERCENT it means you've got a MAJOR problem with paying WAY too much attention to really lousy evidence, what other people think, and the authority of idiots."

The most important lesson here has nothing to do with innocence and guilt, but with people's confidently paraphrasing the opinions of others. How much the more so when those opinions aren't saved as text, both to give the interpreters another chance to parse properly and to introduce the possibility that they could be embarrassed if later shown wrong.
The main reason my estimate is so high is because * I know my information came from two heavily biased sites, and * I found the "innocent" site a lot easier to follow and therefore paid more attention to it, so I know my information is particularly biased in that direction. That said, I did consider a more-arrogant probability of 0.25 or so. My caution in this case isn't on general principle, but because I have something of an old history of embracing cause celebre cases like this only to decide on further reading that the person I'm defending is guilty as hell.
I agree with Eliezer but like Maxwell's point about assigning extra probability to Knox and Sollecito because the guilty argument was so poorly formated. "She was convicted but I don't get why, perhaps I don't understand this." That said I think 15% or less more then accounts for this uncertainty. I gave Knox a 6% probability. Side note, I am surprised that more people are not assigning probability to the chance that none of them did it.

I began with zero familiarity with the case.

  1. Knox: 8%
  2. Sollecito: 10%
  3. Guede: 95%
  4. Agree with komponisto: 80%

Rationale for considering Sollecito more likely than Knox: They're linked quite closely here, but there's enough confusion surrounding the case that I can't take that completely for granted. That being the case, it's unlikely-but-possible that one of Knox or Sollecito was directly involved while the other wasn't, and my prior for a male committing a violent rape/murder is a lot higher than for a female.

Guede is clearly guilty. He fled town immediately after the murder. His DNA was found in the victim's body, by far the most difficult-to-contaminate piece of DNA evidence in the case, making it extremely likely he's the rapist. Very low prior on a rape/murder being committed by separate parties.

The inconsistencies in Knox and Sollecito's stories are definitely worth paying attention to. But there are several factors diluting their importance:

  • I already had a reasonably high prior on the prevalence of brutality and corruption in Italian police forces. This doesn't leave me with much faith in their competence, especially when it comes to interrogations.
  • It's known that Knox and
... (read more)
I've read the other comments, and for the most part, my estimates haven't moved. However, this remark [] by Psychohistorian jumped out at me: Wow, I definitely missed this when I read about the case. The prosecution's theory was already setting off alarm bells. If this is true, that shaves another percentage point or two off of my estimate of guilt for K+S. In general, I'm a bit surprised how much faith some commenters have in their causal understanding of human behavior and psychology, when that understanding seems to be derived from a process of imagining what they would do in those circumstances. I know that I would certainly try like hell to maintain a coherent account of the night's events, but taking this kind of interpersonal analogy for granted when assessing such a delicate situation strikes me as willingly throwing oneself into the arms of the typical mind fallacy []. Some commenters seem particularly focused on some of the more arcane details of the case, e.g., the toilet. How much can you really infer from this sort of thing? I get the distinct impression that people are falling back on intuitions gleaned from detective shows and mystery novels [], which by construction tend to involve cases that hinge on the little things.
The estimates I came up with were a lot less confident than yours, but on reading yours they seem more like my intuition. I think I got burned on the calibration test and now I'm avoiding extremes.
The facts as presented are not accurate so that is throwing off your calculation. For example, when discounting Knox's statement to the police most people consider that it was after a length interrogation but the truth is the questioning lasted one hour. They accept that Knox was mistreated but all the evidence points to them treating her toughly but as expected for a murder suspect. What is never mentioned in the interrogation story is that Sollecito told the police Knox was not with him the night of the murder and that he lied at Knox's request. This information was in fact what led to Knox placing herself at the crime and accusing an innocent man. It should also be noted that this is the second innocent black man Knox had tried to implicate. Also missing from consideration is the fact that Luminol revealed footprints matching both Amanda Knox and Raffaele Sollecito’s feet in the corridor between Knox’s room and the victim’s room and nowhere in the house. The prosecution advanced the theory that these were made by the accused in the victim’s blood while the defense presented the argument that these could have been made at a different time. When one of the footprints contained both Knox and victim’s DNA the expert argued independent deposit. These are not equivalent explanations. The defense position requires that there be a reason for Knox and Sollecito to have blood on their feet. No such reason was ever given but since they had dated for less than two weeks if such an event had occurred it certainly would have been fresh in everyone’s memory. So not only does it require that there be an occurrence unrelated to the murder where these footprints were made but further both Meredith and Amanda had to deposit DNA presumably by spitting of some kind of nasal discharge in the exact same spot. We should also add a reason for why these prints only appear between the victim’s room and Knox’s room and not in the rest of the corridor. That or we accept the obvious that t

What is completely sad (besides this horrible murder case), is the inability of either website linked to present a coherent, rational argument. In fact, I haven't been able to find one website that reveals all facts and then explains them with their point of view in a rational (or even semi-rational) way.

I find this situation almost as depressing as the murder. I couldn't come to any conclusion based on the poor quality of reasoning used on most websites. Wikipedia, as usual, presents a decent collection of facts.

From the Wikipedia article I could only ascertain that Rudy Guede is very likely guilty. My probabilities for the other two being guilty are low (but have a lot of uncertainty), certainly not enough for me to feel that the verdict is correct.

Perhaps this is after all a litmus test for rationality, in a different sense than (I suspect) komponisto intended.

I mean, here I am looking at a 100+ comments thread, discussing a highly charged issue, and everyone is thoughtful, respectful, and willing to take others' points into account.

That is... extremely unusual.

Few of us have had much investment or interest in the case before, I'm thinking. Presented as an abstract almost-philosophical problem, with a common framework of epistemology, it's much easier to discuss well.

Having lived for 14 years in Italy, my impression is that several commenters severely overestimate the rationality and fairness of the italian police force.

5% for the couple, 99% for the first convict. 90% that my probability estimate is close to yours in the sense that you think the two are innocent and the one is guilty.

I'd read a bit about this in the news, and I checked out those sites and wikipedia.

Given the fact that there is no evidence of prior acquaintance of the couple and the man, combined with the fact that the man did not attempt to implicate the couple despite the overwhelming evidence against him, make it very unlikely that they were involved. That, and one person being crazy/desperate/disturbed enough to commit a brutal rape-homicide is much, much more likely than one person and a completely unrelated couple he's never met before being disturbed enough to commit a rape homicide. The defense's response to forensic evidence appeared pretty strong, and the pro-guilt group did not seem like they tried to seriously rebut this (they mentioned that one defendan't DNA was on the bra strap, but failed to mention three other unidentified people's DNA on it as well).

The fact that the prosecutor is under investigation for previously using crazy psychotic hypersexual homicidal maniac theories without basis does a lot to explain how... (read more)

Thanks, your comment has changed my mind about Knox and Sollecito's guilt.
While it may indeed be legitimate to wonder if my having posted this implies anything about my opinion, I'll note that I didn't in this post cite the jury's decision itself as an example of rationality failure, but merely indicated -- in the context of an erupting international controversy and ongoing Internet flamewars --that the whole topic is by nature fraught with obstacles to rationality.

I was unfamiliar with the case. I came up with: 1 - 20% 2 - 20% 3 - 96% 4 - probably in the same direction, but no idea how confident you were.

From reading other comments, it seems like I put a different interpretation on the numbers than most people. Mine were based on times in the past that I've formed an opinion from secondhand sources (blogs etc.) on a controversial issue like this, and then later reversed that opinion after learning many more facts.

Thus, about 1 time in 5 when I'm convinced by a similar story of how some innocent person was falsely convicted, then later get more facts, I change my mind about their innocence. Hence the 20%.

I don't think it's correct to put any evidential weight on the jury's ruling. Conditioning on the simple fact that thier ruling is controversial screens off most of its value.

This is a very interesting analysis -- I like your choice of reference set and your Outside View approach.
I disagree. Do we have specific data about the correlation between the controversy of jury rulings, and their accuracy (or some half-decent proxy, like the likelihood of the rulings being sustained in appeal)? Most of the controversy in this specific case appears to originate from people who have significantly worse access to the factual evidence than the jury; and it's likely to be in the interest of some entities reporting about this case to play up the controversy to attract readers. I don't think there's any strong evidence to be gained from this, and consider the original ruling to still be significant evidence even after taking the controversy into account.


Amanda Knox is guilty: 75%, Raffaele Sollecito: 75%, Rudy Guede: 25%. I shall abbreviate future percentage triples as 75/75/25.

No knowledge of the case before reading this post. My prior is due to my assumption that trial people know what they are doing, and on the fact that I imagined that the trial was trying to show that the guilty were K+S instead of G.

acquiring information

Reading about G's DNA, which should be rather good evidence: switching to 50/50/75. I contemplated switching all the way to 25/25/75, but I figured there had to be some reason for the new trial.

Reading about the police's claim that the murder was linked to a group sex game; thinking that this would be a ridiculous motive. This made me think that maybe the trial people didn't knew what they were doing after all. Switching to 25/25/80.

Finally realized that the trial was in fact trying to show that the guilty were K+S+G instead of just G, not K+S instead of G. Stopped keeping track of percentages for some reason.

Reading about the police switching from K+S+L to K+S+G, which lowered my esteem of the police even more.

Reading about the DNA of K+S, figured it was natural for a woman and her boyfriend to have... (read more)

I like the rapist theory. It's not like Amanda was the only promiscuous American collegian around - birds of a feather... And who would rob an exchange student? No, a consensual meeting gone bad sounds like a far more common scenario to me.
Why would you murder someone just because she didn't wanted to have group sex? And even if you did, why the hell would you call the police yourself? Furthermore, this case was decided by a jury. A group of average people with no real juristic knowledge at all. Just some random dumbasses who couldn't care less and just want to get home (no, I'm not narcist. Just consider: 50% of all people are by definition dumber than the average, and a jury is randomly selected. ie the average jury ranges from IQ 80 to 120, while a professional judge should start at about 125.) Do you really believe they would come to the correct decision? And just think of the charges against the prosecutor (about inventing crazy conspiracy theories) . . I guess for me its 5/5/90 for K/S/G.
Not sure what you are thinking, but to clarify: I meant an encounter between just Meredith and Guede, nothing to do with Knox or Sollecito. I could see a single guy (Guede) expecting sex and then resorting to rape and then murder to cover it. (And I suspect the jury is on average better than the populace, since just about every country gives into the temptation to lard on extra conditions like 'if you're a felon you forfeit forever rights such as being on a jury or voting', which would disparately affect the lower percentiles.)
Ah, I thought you referred to an encounter between all of them. In this case, I agree with you - that's also what I think happened. Regarding the juries: I've read to many reports about bad juries than to believe in them, and the fact that they are almost certainly less educated than a judge still remains.
The Italian system is different from ours. This particular jury included two judges.
That's an interesting variant. There may well be advantages to such a system to counterbalance the disadvantages. I know, for example, that I just felt my 'confidence of innocence' adjust itself downwards. (I would expect a judge to be more likely to be corrupt than a random citizen but also to have less naive vulnerability to obvious manipulations. The latter is relevant here.)
[comment deleted]

Familiarity pretty good - I've read the Wiki page, revisited several articles from when the murder was first discovered and I watched Sky news the day of the verdict and saw/heard Prof of Criminology, feminist journalist, UK barrister and two Italian barristers. I frequently search the web, hence I found this site.

(I don't understand the up/down system.)

I find the logic of the murder disturbing - if the murder was a game gone wrong, then it was not premeditated, so unlikely gloves were worn. If bleach was used to clean, then why was Guede's DNA all over the body and room? The only DNA evidence for Raffaelle is highly suspect and a physical impossibility to leave DNA only in that one place I would think. That really is the beginning and the end of it.

The Prof of Criminology's view I have to take seriously. But, in saying that Amanda's diary reads like a gap-year Rosemary West he failed to draw contrasts as well as comparisons. Rosemary West was severely abused from early chidlhood, she did not study languages or develop an interest in creative writing.

I have an alternative scenario - a young woman from a comfortable secure All-American upbringing, who has no idea of the level of cor... (read more)

I've often seen that pattern. When someone is murdered and someone is convicted for it, the bereaved insist, no matter how controversial the trial, that justice has been done and that any querying of the verdict is an insult to the memory of their loved one. It's completely barking mad, but then, people are crazy []. (Edit:) And welcome, erica.
In fact, the bereaved normally do this before and throughout the trial, and sometimes continue afterwards even if the accused are not convicted. The media loves displaying them during the trial.
Thanks for the Welcome. I laughed at the illustration. * I'm in the UK, are most other people in the US?
Thanks for your replies. Orientation increased to 37% :)
Most but not all. There is a small contingent of UK people, myself included. There were meet ups for a bit in London, but those have fallen by the wayside recently.
Somebody posted a breakdown by country a while back. You can probably find it in the search. I think the US was the largest group by raw count. I'm from Australia by the way. Also, welcome.
Erica Are you being ironic?, "But...she did not study languages or develop an interest in creative writing."
I don't think I'm being ironic. First, I don't think character analysis is necessary in this case but as the prosecution and support for the verdict both rely on character analysis, I have attempted to put forward an alternative analysis that depicts Amanda as cosseted, well-educated, literary and imaginative. This is the opposite of Rosemary West, so I raise the question as to whether the two styles of writing are directly comparable.

I averaged up the answers given so far in this thread for Knox and got 35% mean, 20% median.

Is this an arithmetic mean or a geometric mean? Which is the correct mean to use for averaging probabilities, anyway?
The arithmetic mean of the log odds [] is pretty natural. It is 27%, but the median looks like 30% to me.
Neither really, but there's no easy way to do it. The mean has the problem that if a lot of people claim near ignorance (like I did), then that counts against Knox, when really, it doesn't mean anything. The problem with the geometric mean is that it is biased towards the low end of the spectrum, so it depends on if the statement is negated or not. (GM(.001,.999)<4%) The median is probably better than both, but it's still not the right way to do it. Ideally you'd try to count up how much evidence each person saw and add those, but it is no easy task to estimate how correlated the evidence is (though it's probably a worthwhile subject to put thought into, since thats how you determine how much an additional persons belief is worth []) Even with a large degree of overlap, this is probably one of the cases where sharing beliefs should make everyones beliefs more extreme. I'd sorta like to see what it'd look like on round two.
[comment deleted]

I arrived at the site from the HP fanfiction after reading the author's notes concerning the case.

Probability estimate that Amanda Knox is guilty -- 99%

Knox and Sollecito's alibis are contingent on the other. However, I recognize that there is a probability that one may lie to cover the other, so it is not implausible that one may be guilty without the other being so. While they certainly had the opportunity to murder, there appears to be neither motive nor weapon. The only credible evidence against Sollecito is his DNA being on the bra hook material,... (read more)

I'm here similarly after reading the aforementioned author note. There's no legitimate controversy regarding the bra clasp. The clasp was cataloged six weeks after the murder and after being handled by multiple investigators. Sollecito had, of course, visited the apartment multiple times in the two weeks prior to the murder investigation beginning, so his DNA was present in the apartment. The DNA found on the bra clasp was entirely consistent with contamination, and the circumstances under which the clasp was cataloged make contamination inevitable. No reputable judge would ever allow the jury to consider the bra clasp as evidence. Given the lack of any other evidence against Sollecito and the compelling evidence against Guede, no rational person would attribute any weight to the bra clasp. Anyway, I'd heard about the case before, but hadn't followed it. There's no physical evidence against either Knox or Sollecito, the inconsistencies in their statements are classical examples of the inconsistencies brought about by sleep deprived intimidation and interrogation by the police, and all the "theories" about how they participated in the murder and covered up the physical evidence of their participation are tin-foil hat material. And there's an absolutely compelling suspect who isn't them. Probability that AK is guilty: indistinguishable from 0%. Probability that Sollecito is guilty: indistinguishable from 0%. Given that Guede was in the apartment at the time of the murder (admitted by him), his DNA was found inside the victim, he initially offered a ridiculously implausible story about how the murder happened ("Guede claimed he suddenly needed to use the bathroom, and while he was sitting on the toilet listening to his iPod, a stranger entered the cottage and attacked Meredith. Guede said he emerged from the bathroom and grappled with the stranger, who ran off into the night after shouting "a black man found is a black man condemned.") and changed his story to na
I had much the same experience as Sinai. Personally, though, I didn't bother to try this test after having already been biased by Eliezer's opinion on the matter. Also, I feel that the uselessness of the linked pro-guilt site hurts the overall experiment too much. I know it wasn't translated at the time, but using the Massei report as Sinai did makes the test much more effective.
It's not just that it hadn't been translated; it hadn't even been written! It came out in March, around the same time as Bruce Fisher's excellent Injustice in Perugia [] site went up. If I were proposing this experiment today, those would be the sources I would use (i.e. the Massei-Cristiani report for pro-guilt, and IIP, including the appeal summaries [], for pro-innocence). Believe it or not, True Justice for Meredith Kercher represented about the highest quality pro-guilt advocacy available at the time of this post.

Page one of the site arguing defendants are guilty has nothing that would count as evidence for guilt. When I got to the bottom of the page and saw that there were 24 more pages, I lost patience for the exercise because the low quality of the argumentation on page one (most saliently, the picture of the vicitim when she was five, which if course is not evidence at all, but which will tend to evoke biased thinking in the reader) was a sign that the other 24 pages would be very sparse in actual evidence.

Aren't there enough opportunities for us to practice r... (read more)

Sorry about that; it's scattered accross the site, which as I said is basically a blog. Try the category links on the right side of the page; such as this []. I believe I know "the answer" with high probability, and want to see what others think. Obviously, anyone else is welcome to post whatever rationality exercise they wish.
Do you have a reason to believe that your opinion is more likely to be correct than other commenters on this site? Do you believe them to be guilty and linked to an impassioned site full of logical fallacies over a more informative one? (I don't mean to impune your post, just guessing that this is the solution to your rationalist puzzle) I think this experiment is going to be of limited success at best due to the fact that people on the road to rationality are far less likely to acquire new beliefs with both strong emotional component and poor grounding in facts. That's kind of the point of being a rationalist, true beliefs.
impugn []?
The experiment will be a success if there is significant participation. I certainly don't expect people here to "acquire new beliefs with both strong emotional component and poor grounding in facts." It's because LWers make an effort to avoid this that I'm interested in hearing what they have to say. Which site are you talking about? This isn't a "puzzle". To some extent, it's a sanity check I'm performing on myself.
Well, I wish your original post had been a little clearer about your reason for posting. I had somehow gotten the idea that my doing the exercise would be a good way for me to improve my own rationality or the rationality of a bunch of other readers, not just helping you assess your own rationality. Asking everyone here to do your exercise is an expensive way to do the latter. Reading about a current trial and its aftermath is a very inefficient way for the reader to improve his or her own rationality; would you not agree? Do you think that it is an efficient way for the reader to assess his or her own rationality? (I do not unless there is unusually strong evidence, e.g., from DNA, that is held back till everyone has given their probabilities.) If I did not regard you as a positive contributor to LW, komponisto, I would not bother making these comments on your post. But, OK, now that you have my attention, I will respond to your human need: yeah, it is shocking how badly judicial systems can malfunction and how bad most people (even judges) are at rationality. Probably the thing that has most helped me tolerate these shocks is caring personal relationships with people whose rationality is above average. There is something about really being heard and understood by another human being about some point that will trip up many people that makes this sort of thing more tolerable.
Anyone participating can make their own judgements before seeing the probabilities given by other commenters. I found that quite interesting and not solely for komponisto's personal benefit.
On reflection, this statement of mine was misleading. Now that I've thought more about why I believe what I believe -- the fundamental question of rationality -- and am preparing to write up my answer in the form of my next post, I now realize that this post could in fact be considered a sort of puzzle, or test, the answer to which will be given in the next one.
I agree. A lot of the stuff on that site was just silly. It was mostly appeals to emotion and random stuff like (I'm paraphrasing), "A witness heard people running on some metal stairs, therefore the defendants are guilty!" I think this is a test to see if we all come to the same conclusion in a case that stirs up a lot of emotions. If komponisto is Italian, he might have vastly different probability estimates than some of us Americans.

Never heard of the case before, after reading the wikipedia page on the crime and its associated discussion page I think it's very unlikely that knox and sollecito are guilty. Certainly the evidence does not seem at all sufficient to convict them, and interrogating someone for 14 to 30 hours without recording the interrogation is downright idiotic.

I expect you agree.

I'd like to see more posts of this nature. This site has too much theory and not enough practice.

I guess we were all guilty, in a way. We all shot him, we all skinned him, and we all got a complimentary bumper sticker that said "I helped skin Bob".

my somewhat admittedly sketchy reasoning:

I go to the University of Washington where there is considerable interest in the case. Of the people who have only been marginally involved in the case, most believe that Amanda Knox is innocent. Of the people who are interested in the case, many believe she is guilty. There's an obvious hometown effect here which biases towards innocence so I'm assuming those who look into the case are taking that into account when and still reach a guilty verdict.

Therefore, I assign a 70% probability to Amanda Knox being guilty (+ or - 30%).

Given the information you're going on, that's not a bad estimate. Actually reading on the case may change your opinion dramatically, though; why not try it?
  • P(AK=guilty) = .01
  • P(RS=guilty) = .01
  • P(RG=guilty) = .995

Do I think my assessment will coincide with yours? Of course I do, we're supposed to be rationalists!

I had zero familiarity with the case before reading the links provided, and did not read any of the comments in reaching my estimate.

I admit to having non-trivially updated based on my perception of the lack of seriousness of the pro-conviction site's domain name (what is this, Marvel Comics?)

3Paul Crowley13y
Those are very high confidences, could you say a little about that?

The prior probability of a three-way conspiracy to commit rape and murder is way, way lower than the prior probability of it being a one man job.

I didn't see any evidence that would move much probability mass away from the prior probabilities, but this could just be due to the slice of the evidence I saw given my 30 minutes or so of reviewing it.

How much of a shift do you think this accounted for? What would your estimates have been if the same site had been called something different?
The name is just one piece of evidence. Overall the pro-conviction site did not impress me at all. It seemed to be full of irrelevant statements, and things like site usage statistics.


  1. False
  2. True
  3. True

Why am I giving (most of) these in boolean terms rather than probabilities? Bayesian probabilities aren't useful in cases where the most probable scenario for (AK guilty) is something like "Two of the perpetrators were secretly ninjas". There really is no rational way to convict someone for leaving no forensic evidence in a room whatsoever.

I have to admit here though that I peeked at your article before posting this. And incidentally, predicted what it would say pretty damn well. (AK not guilty with a probability that r... (read more)

Jury decisions that prompt public scrutiny certainly seem to be!

Would anyone actually be up for discussing the specifics of the case? (I don't know why but I find myself oddly interested in this case.)

As far as I can tell, the biggest pro-defendant evidence is that there is no major DNA evidence of Sollecito and Knox in the room where murder took place. We are told that there is a bra clasp with Sollecito's DNA and a knife that has both Amanda's and Kercher's DNA - both of these DNA traces are 'weak' in the sense that they are not that obvious, require a hefty search and are hard to see in lab. On the other hand, ther... (read more)

I, too, find myself oddly fascinated by the case. I assumed Sollecito and Knox were guilty until just before the verdict came in, when the story was gaining more traction here in the U.S. I can't recall what it was that I read that made me question their guilt, but it set me off on a quest to learn as much as I could about it. I've basically taken details reported in the media, blogs, etc., that disturbed me and looked for the defense's OR prosecution's take on that detail. Here are the main points, and what I understand to be the truth behind the "evidence" - listed in no particular order: I wish I could provide sources, but I haven't kept track.

  • One of the main things that I keep in mind as I read about the case is that the prosecution leaked many details to the public which have since been proven false and public opinion was turned against Knox & Sollecito very early on based on a lot of incorrect information. The same incorrect information still abounds on the internet and in many minds.

  • Bleach. I had read a number of times that Knox and/or Sollecito had purchased bleach around 7 AM of the morning following the murder. I'd even read that there were receipts conf

... (read more)
Thanks for your summary. The only place I differ from you is the cartwheel part. This behavior strikes me as genuinely insensitive and disrespectful but being disrespectful and insensitive doesn't make one a murderer. I'd like to believe that the prosecution has a case but for the life of me, I can't see one. One thing that struck me as weird is that Kercher's family was 'pleased' with the verdict - do they really think that Knox and Sollecito took part in the murder? Why do they think that way? I'd like to know. Surely, the Kercher family must be reasonable people - so why are they pleased with the verdict? The horrifying prospect is: do they know something I don't? If so, I must search for it and learn it... :-(
Why look for rationality in the desires of a typical bereaved family? Surely if they had their way, anyone associated with the event at all would be punished, so great is their loss.
In response to the cartwheel part - here's a possible explanation. It's from a pretty clearly biased source, but it does sound reasonable. [] At the very least I doubt she was leaping around exuberantly and spontaenously.
From the link you give: Thanks for this - one more mystery solved.
Oh, definitely. This is a known bias []; fight it.
0michellesings13y has some current stuff
Thanks for the link! I will try to fight it :-).
People can behave strangly under pressure when they don't want to except reality. If you were never in a situation which such pressure it's hard to estimate your own reactions to it.
Actually she was doing Yoga to calm herself. No lie.
  1. Knox guilt: 80%
  2. Sollecito guilt: 80%
  3. Guede guilt: 99%
  4. prob of agreement: 0.8

  5. I was unfamiliar with the case before hearing about the verdicts on the news. I spent about 10 minutes on the Wikipedia article, then20 minutes on each of the sites you linked, and made these judgments before reading others' comments.

Guede's guilt seems almost beyond doubt, given the DNA evidence, his implausible story, and his flight without calling police. Regarding Knox & Sollecito, I'm convinced of the prosecution's version of events mainly by the cell-phone record e... (read more)

After reading more comments, I've updated my probabilities significantly. Here's what influences me: 1. Different commenters focus on different subsets of evidence. This makes me suspect that my own focal subset was incidental and probably depended on what order I encountered various claims. 2. Many items of evidence presented as fact (and which I relied on), such as cleaning supply shopping and the contents of the washing machine, are said by others to be rumors and were never presented at trial. This undermined a lot of what I based my judgment on. 3. The judgment of people who have followed the case very closely (e.g. jenmarle) is that Knox and Sollecito are innocent. Together, these things make me throw my hands up in the air. I don't think I can clear things up without spending much more time on it, definitely not without seeking out new sources, and I don't know where to look. I now believe: Knox: 50%, Sollecito: 50%, Guede: 99%.
The cell phone evidence and the changing stories are what gives me reason to a probability of guilty to S and K above .05. (See my comment for my probabilities). The effect of this evidence was mitigated for me by: 1. Their likely intoxication during this period. 2. Police coercion leading to confusion about the events. 3. It looks like a lot of what the pro-guilty site claims ended up being unproven rumors not produced in trial, this makes it very unclear to me what story S and K actually ended up giving in trial. Since the site exaggerates other claims their claims about the cell phones and the stories can't be fully trusted. You weren't troubled by the lack of motive and the fact that after covering up her involvement in a crime but intentionally leaving evidence of Guede's presence Knox went on to point the finger at someone who couldn't have been involved after being coerced by the police? Shoot, I can't figure out why Knox, if she was guilty, wouldn't have just stayed at Sollecito's until someone else discovered the body.
What I can't understand is, if all three were in it together, given the evidence against Guede, why he didn't rat out the other two in exchange for a reduced sentence. I'd be amazed if the Italian legal system doesn't cut such deals, and I'd be amazed if the prosecutors didn't try to get him to rat out the other two. If they were actually involved, the odds that he'd turn on them in that situation seem well over .9.
Hmm, thinking back, I didn't consider motive too closely. I rather got caught up in the evidence of Amanda's evasion. In retrospect, this would lower my probabilities. I didn't judge that evidence against Guede was left intentionally; I could easily have missed details indicating this. Knox's fingering of Lumumba seemed to me to be a natural part of a collection of inconsistent attempts to establish an alibi and deflect suspicion.
FYI, I've updated in your direction since my first response. Those phone call lengths are driving me nuts. Part of the suspicious behavior was that Knox didn't flush the toilet (where Guede had used it) the morning she returned. Also, Knox and Sollectio's 'bloody' foot prints had to be luminaled when right next to them were Guede's visible foot prints. And over all, the shear amount of physical evidence against Guede compared to the near total lack of physical evidence against the other two suggests that if there was a clean up they really cared not at all about Guede getting caught. Which is believable, but they did too good of a job. They're intoxicated the night of, come back a few hours later and can distinguish all of their prints from all of Guede's? And they didn't think he had left enough physical evidence (on the victim) that they thought they should leave the toilet unflushed? They appear to have done a really good job cleaning up and a really bad job getting their story straight- which seems inconsistent to me. Though come to think of it I'm not really sure why the toilet wasn't flushed at all. If Guede used the bathroom before joining Knox and Sallecito killing Kercher, why wouldn't he flush it? Hearing the screaming is actually a plausible explanation for this. Not flushing would make sense if Guede was trying to provide evidence for the story he was planning on telling but that seems way to smart for him. I also don't know why that would be his planned explanation if he was working with Knox and Sallecito. Part of the problem is that the details of what happened in the room were never released in English (and that might be a good thing) so there may be good reason to think the evidence indicated three people were involved, etc.
This seems to be a major focus of the pro-guilt site, though it isn't really backed up at any level of detail. Were it true, it would increase the probability that K and S were involved, but still far short of beyond a reasonable doubt, I would think.

No prior familiarity; thus I started with no information and no particular beliefs about their guilt or innocence either way.

The first thing I saw was that Ann Coulter is convinced that Amanda and Raffaele are guilty. I immediately moved my belief in their guilt way down. When Ann Coulter takes a strong position on a controversial issue, she is almost always wrong.

From there it was mostly downhill for the prosecution, as far as I could tell.

"Later, when a airtight alibi forced the authorities to release Lumumba, they substituted Guede as the third pa... (read more)

That's what the police always want. Their role in the game of law is being conviction-maximizers. Given leave they'd tile the universe with convicts. That's why courts have judges who are supposed to be independent of the police. (At least, in the better class of country.)
Okay, I'll bite. Could you name a few examples of this, especially the ones that cemented this belief?
For a claim that she is almost always wrong 'a few' examples won't be substantial evidence - assuming she has more than a few opinions. We'd need either an almost exhaustive catalog of her opinions, or a way to sample them in a sufficiently random fashion.
Agreed. All that I have is a highly unscientific impression based on my own personal experiences with her. So far she's batting pretty close to 1.000 though. The fact that the consensus of this community is contrary to Coulter's conclusion I'm counting as one more data point.
Oh jeez you're asking a lot. Too many to count. Google her if you feel up to it. And honestly, I don't really believe this is a serious guide to truth and falsehood. Every time I test it, it comes out right. But I can't run enough tests to know for certain.
Actually, I want to thank you (and Dan, below) for making me think a little more carefully about this. I now think that the constant wrongness of Ann Coulter isn't an accident. She is an almost perfect example of a pure anti-rationalist: someone who will always and only believe things that accord with her ideology. You can predict what she will say about many issues via a simple process. For instance, take the sentence "Muslims are bad," and apply it simplemindedly to any issue involving Muslims, and you will be able to predict her beliefs. She insists that Muslims had nothing to do with the advance of knowledge; that Islam has never been a religion of peace or tolerance; that Sirhan Sirhan was a Muslim (he wasn't). She writes: "Muslims ought to start claiming the Quran also prohibits indoor plumbing, to explain their lack of it." And on, and on. Similarly with "liberals are bad." She believes that liberals are always wrong. Among the conclusions she draws: liberals believe in evolution, therefore evolution is false. I don't know, it's a pretty impressive record she's got going. She will, no doubt, be right about things on occasion, by accident. But I'm starting to feel better about the reliability of my "shortcut to truth." :)
Most real-life issues admit of more than two answers. It's just the political and media approach to paint things as black and white. In other words: reversed falsehood is not truth. You can't get at the truth by taking the opposite position from Ann Coulter. Because the vast majority of statements don't have a single opposite position. So while you can judge her always wrong, it will only help you to be right by dismissing her opinions, not by suggesting the right ones.
Of course. It's an exceedingly limited heuristic and valuable only in rare circumstances.
I'm very opposed to this kind of statement, because one of my core beliefs is that reasonable people can and will disagree on politics. This sounds too much like people calling Bush or Obama stupid when they disagree. I suspect that you have a strongly opposing political ideology to Coulter's, and this is biasing you against her. I am aware of factual errors in some of her books, especially the ones about evolution, obviously. But assuming that you don't like Coulter and that you disagree with her values, I don't think you can objectively comment on her rationality.
That would seem to make being politically polarized a bulletproof protection against people noticing that you are completely irrational. I don't buy that.
Do you extend this distrust of statements made about people who disagree with you on politics, to the field of religion as well? Do you expect creationist Christians to be as rational as scientific atheists who accept evolution? Coulter is not only "conservative," she's also a creationist. My problem with Coulter is not that she's conservative. It's that she doesn't think about issues independent of her ideology. There are those on the left who are similar.
Of course not, because that's not a two-sided debate. Politics is. But in that case, you could equally well say, "The first thing I saw was that X is convinced that Amanda and Raffaele are innocent. I immediately moved my belief in their innocence way down. When X takes a strong position on a controversial issue, she is almost always wrong," where X is Coulter's left-wing opposite. If Coulter and X always write rhetoric from one ideological position, I can agree that you could say they were equally irrational, in the sense that they can't think outside their ideology. But I don't see how you could come up with a useful truth-heuristic from that. Ideological does not imply incorrect. Your truth-heuristic of always opposing Coulter seems to be another way of saying "Coulter's ideology is always wrong."
I'm curious what you mean by this exactly. Do you mean that politics is something on which rational people can agree to disagree whereas the truth of evolution is not?
Politics, unlike evolution, isn't really a factual matter. Take abortion, for instance: some people have the goal of reducing abortions, some people have the goal of reducing the population, some people have the goal of maximizing women's reproductive choices, and some people have the goal of maximizing fathers' control over whether or not they have kids. Which goal you have depends on your personal preferences and values, though you may have to bite the appropriate bullet for whichever goal you accept. And even if two people or groups agree on the goal, there isn't enough information for us to know what the most efficient way of getting to that goal. We might have different ethical or pragmatic constraints as we work towards that goal (for instance, not wanting to give the government too much power on our way there). Even if we agree on the probabilities of success for each legislative choice, say, we may have different tolerances for the risk we as a society assume in getting there. The world of politics and legislation is vague enough that there's not always only one right answer (though there may be definite wrong ones).
I think it's probably true that some political disagreements come down to differing preferences and values. It doesn't strike me that the majority of policy debates which account for much of the noise that passes for 'political discourse' can be seen as debates over fundamental values and preferences however. A significant amount of political debate seems to revolve around what at least appear to be questions of fact about how to best achieve certain broadly agreed upon aims. I tend to think that a lot of politics and political debate is best understood not as a search for truth but as serving other goals not related to the surface appearance but the more thoughtful and better intentioned participants in the debate do generally seem to believe that they are to some extent debating a factual matter, albeit one that is not as clearly settled by the available evidence as the question of evolution.
You can doubt k3nt's ability to recognize such cases correctly, but there's no internal inconsistency in his description. I know that many people - certainly public media or political personas - are, in fact, ideologists of this kind. I can't say about Coulter because I'm not from the USA and know nothing about her...
They can do that?

Amanda Knox guilty: .01

Raffaele Sollecito guilty: .01

Rudy Guede guilty: .99

I've become highly familiar with this case since the verdict came down. Over the last 2 years, I've heard bits and pieces about it and all along had assumed Amanda and Raffaele were guilty. I'm a little embarrassed to admit how much time I've spent reading up on the case recently - I think I'm motivated to learn more because I'm perfectly appalled at the amount of misogyny (not necessarily anti-Americanism) I see from the prosecution and the European media with regards to Amanda Kno... (read more)

For Knox, Sollecito, and Guede in order: 30, 30, 95, but didn't find either website or Wikipedia enough to feel like I had sufficient information. I think you probably thought the same.

EDIT: After looking at everyone's comments, I'm revisiting to 20, 20, 95. Anyone else want to edit their comments to say how they updated in light of everyone else's opinions?

I was unfamiliar with the case. After checking out both links for quite some time, but prior to reading the comments, I estimated:

  1. 80% (Knox)
  2. 60% (Sollecito)
  3. 95% (Guede)
  4. 90% (confidence in coincidence)

After reading the comments, I was a little surprised that the consensus seems to be decidedly against Knox's guilt. The simplest explanation is that I'm just not a very good rationalist, but I don't find that very satisfying. The four parts of the story that I felt were inconsistent with Knox being innocent were:

  1. Knox's initial account of the night.
... (read more)

The body was covered. This is inconsistent with the actions of a rapist/murderer, but very much what you would expect of someone who had a close relationship with the deceased.

I find the above incredible. I'd give it almost no weight.

1.What I gathered was that the police saw Knox's text message to the bartender and then coerced a confession involving him. The fact that they got this confession when there is no way it could have been him suggests to me that much of the confession could be totally fabricated. For that matter, why would Knox name the wrong accomplice if they knew they didn't cover up Guede's presence at the crime scene? 1. One of my problems with the supposed cover-up is that if S and K were intoxicated during the crime and likely during the cover-up they A) wouldn't have been able to distinguish between evidence implicating them and evidence implicating Guede and B) wouldn't have been nearly as successful covering up physical evidence as they apparently were. 2. Perhaps Guede liked her and felt guilty. Part of the suspicion re: Knox was that she was insensitive after the fact. But this would be inconsistent with her covering up the body. 3. Knox was almost certainly seriously hung-over and not in the mood to go near fecal matter. Also, her roommates testified that she didn't do a lot to keep the place clean in general. I'm also not sure how not flushing suggests guilt. Marijuana does not undermine rationality to that extent. Reefer madness PSAs considerably overstated that :-)
I have to generally agree with you (and I'm also surprised that the majority here seems to believe in K+S's innocence. The other piece that seems strange is why Kurcher's clothing was in the wash that morning. Just seems like something strange to do... a generally messy person doesn't wash someone else's clothing the morning after partying. Who else might have run the washer otherwise? It's questionable exactly how involved Knox and Sollecito were, but I don't believe that they are completely innocent. I was unfamiliar with the case but spend about 2 hours reading the two provided links.
My understanding was that the clothes washing was a rumor and never used as evidence. That said I find the amount of misinformation surrounding the case incredible problematic.

This was my first contact with this story. I still don't feel informed.

Wikipedia was the best of the resources. The site arguing defendants are guilty was the worst. My probabilities on Konx and Sollecito are "high" because I feel I still have not found an argument against them that was properly constructed. Before lowering my probability to 1% I would like to hear a better explanation of why the court found them guilty.

  1. Your probability estimate that Amanda Knox is guilty. p = 6%
  2. Your probability estimate that Raffaele Sollecito is guilty
... (read more)

Just skimmed the two sites. First:

  1. Your probability estimate that Amanda Knox is guilty.
  2. Your probability estimate that Raffaele Sollecito is guilty.
  3. Your probability estimate that Rudy Guede is guilty.
  4. How much you think your opinion will turn out to coincide with mine
  1. p = .45
  2. p = .45
  3. p = .60
  4. We probably agree on which side of .5 these numbers should be, i suspect you are far more confident, which is perhaps understandable if you have a better understanding of the evidence

It's hard to be very confident after skimming for 30-40min. I don't have th... (read more)

Any thoughts on why that might be the case?
I'm more or less a libertarian.

Guesstimates based on quick reading without serious analysis:

(1) Probability that Amnda Knox is guilty: 5%

(2) Probablility that Raffaele Sollecito is guilty: 10%

(3) Probability that Gudy Guede is guilty: 60%

(4) Probability that my estimates are congruent with OP's: 50% (ie random, I can't tell what his opinion is)

I'm pretty familiar with this case having read a couple of books about it and followed both of the websites listed and a few others.

I would say there is 0 chance of Knox or Sollecito being involved in the murder at all. They have no motive. The prosecution completely failed to show that they had ever participated in "sex games" together or with different partners in the past. There is no evidence of either of them at the crime scene at the time of the crime. None of their past behaviors make them likely candidates to commit this type of crime. Ev... (read more)

Assigning probability zero to something is not in general a good idea [].
Why? What negative result follows when you assign probability zero to something? I haven't consciously experienced any so far.
It might help to read the linked essay in my previous comment. When one assigns probability 0 or 1 to something, one cannot update based on evidence. That is, if one really believes that, there should be no amount of evidence that will convince you. That's problematic if one is trying to be a rationalist. And if there is some amount of evidence that can convince you then you don't really assign probability 0 or 1 to the claim.
  1. ~40%
  2. ~20%
  3. ~60%
  4. I think with reasonably high confidence that you will be on the 'K/S are innocent, and G guilty'; perhaps 70% confidence

My sources are occasional articles in the NY Times, a few other mainstream media by way of Reddit, and 30 minutes of reading through the 2 linked sites. The blog one seemed very bad to me; I gave up reading it in disgust about halfway through the post on the telephone chronology, because I was tired of the breathless insinuations that if someone has a detailed phone-call, obviously they must remember it flawlessly hours ... (read more)

Great idea. I had heard the name Amanda Knox and knew that she was suspected in a murder of another exchange student in Italy. Thats it. I looked at both sites and much of the evidence. I glanced at the wikipedia page briefly and did a limited google search to try and find information about one more fact that I was hoping to find in the pro-Amanda page, but didn't.

My answers:

  1. 13% she was in the room and assisted in the murder. 9% She was involved in some other way, had prior knowledge etc..

  2. 15% in the room, 9% other involvement.

  3. 97%

  4. I'd say there is a

... (read more)

Amanda Knox being Guilty: 90%

Raffaele Sollecito being Guilty 90%

Rudy Guede being guilty: 90%

I hope you agree with me because no one else in the comments seem too. I'm gonna give the probability of you agreeing with me 75% based on my own arrogance and belief that I'm right and based and little else really.

Most people seem to believe Rudy Guede is guilty so lets skip that and look at the other two. They've changed their stories and have been proven to have lied quite a few times. For example Rudy at one point said he was at home surfing the internet b... (read more)

Do you really find it equally likely that Knox/Sollecito are guilty as you do that Guede is guilty? It seems like most of the weight should be given to Knox-Sollecito-Guede and Guede as possibilities, so unless you think the probability of Guede acting alone is very close to zero, this is sort of bizarre. In particular, it indicates that your P(KSG | G) is very close to 100%.
1. The statements were obviously coerced and Knox and Sollecito were intoxicated during the period in question, it isn't surprising they have given inconsistent and contradictory stories.The fact that no recording of the interrogations was ever released is incredibly damning. 2. My understanding from what I saw was that no evidence re: purchasing cleaning products was ever introduced. 3. Did you read the arguments countering the DNA evidence? Was this luminaled or visible? I might have missed this piece of evidence. Also, I wasn't at all convinced there was any clean-up afterward.
1. Just because they were intoxicated doesn't mean they shouldn't be able to tell us where they were. I have never been drunk enough in my entire life to not remember what house I slept at. Have you? So because the police were idiots and didn't record the interrogations that means they faked the evidence? What would be the police's motivation for faking the evidence? Evidence is held back for lots of reasons, doesn't didn't happen. It was on the wikipedia page so I'm taking it as fact. 1. No. I'd love a link to them. However I can't invision reasonable argument against DNA evidence, unless there's evidence of a massive police conspiracy.
Intoxicated people as a rule have fuzzy and missing memories. I've definitely forgotten events when I've been drunk and I've had friends lose a couple hours of memory and wake up not knowing where they are. Maybe this isn't sufficient to explain the inconsistencies but it is once you combine it with scary, potentially abusive police yelling at and threatening them... well, I'd say some inaccurate testimony is to be expected. That said I'm having trouble with their story of the morning the body was discovered. Initially the police said they had lost the recording and later said they had never made one. In fact, they didn't even have a transcript of the the interrogation just a signed statement in flawless italian. These things "get lost" when the police realize the confession was coerced and won't hold up on appeal once people see the video tape. And are you really asking me why police would manufacture evidence to get a conviction? Like, are you kidding? If you want a motive unique to this case you might read the criticism of the Italian prosecutor. From the wikipedia page: It doesn't mean it didn't happen. It may have, I'm not convinced of their innocence. But the fact that it is on a contested wikipedia article is a terrible reason to take it as fact. Wikipedia is great for reading about the history of cheddar cheese, less so for answering questions about current, controversial and biasing events. Btw, did you really read the defense's site? Quoting: Finally: This was in the main section of the defense's site []. Section 2.
Downvoted for poor spelling.

1) Knox guilty = 0.2, however 0.35 that she discussed something seriously or not with Guede before the murder. 2) Sollecito = 0.05 3) Guede = .98 4) Your opinion is probably similar to mine. 5) The wikipedia article.

I was totally unfamiliar with the case. My assessment is based on ~30 minutes of reading the pro-prosecution site and ~5 minutes of reading the pro-defense site.

  1. My probability estimate that Amanda Knox is guilty: 33%.
  2. My probability estimate that Raffaele Sollecito is guilty: 33%.
  3. My probability estimate that Rudy Guédé is guilty: 90%.
  4. How much I think my opinion will turn out to coincide with yours: same direction, mine less extreme.

For me, the red flag was the way Guédé replaced Lumumba in the prosecution's theory of the crime. An investigation... (read more)

  1. Your probability estimate that Amanda Knox is guilty.

5%. She was near the crime scene and doesn't have a coherent story about what she did that day. She might have bought bleach. Other than that they don't appear to have any evidence against her.

  1. Your probability estimate that Raffaele Sollecito is guilty.


  1. Your probability estimate that Rudy Guede is guilty.

95%. Handprints in blood. His blood mixed with the victims. His "DNA" "inside" the victim. If he is not guilty then he was framed in a quite advanced manner. And in... (read more)

Maybe this is a bit unsportsmanlike of me, but, for the question: "What probability your own opinion is accurate", my answer would be.... low enough to preclude me from participating in the other predictions.

Very little familiarity prior to this post, read both sites for maybe 5 minutes each.

  1. 50%
  2. 40%
  3. 90%
  4. Your username looks vaguely Italian, so I'm guessing that your estimates of the chances that Knox and Sollecito are guilty is significantly higher.
My username is Esperanto for "composer"; you wouldn't find the letter "k" in a native Italian word. As for what my own estimates are, see here [].

I have a different objection to the premise: the presumption of innocence in modern legal systems means that the job of the jury (and by extension the legal teams) is not just to arrive at a probability of guilt but at a certain level of confidence around that probability.

I realize that these can technically be made equivalent by endless "priors" -- i.e. a juror walks into the courtroom with a certain set of beliefs, ie probability .8 that someone on trial is guilty, .01 that a sex crime would have been committed by a female rather than a male, e... (read more)

I don't think those are two separate things. What does it mean to be 50% sure that there is a 90% probability someone committed the murder? If you're not sure you should just lower your probability of guilt. I think Eliezer had it right that when answering the question you should give your best estimate of the probability that each suspect committed the murder. The question of what probability corresponds to 'beyond reasonable doubt' is a separate one and isn't actually raised in the original question. Personally I think you'd have to assign at least a 90% probability of guilt to convict but the exact threshold is open to debate.
I just went to Wikipedia and found a more articulate version of what I'm trying to say: I am not really a stats person and I'm not prepared to defend Garder-Medwin's model as being correct -- but right or wrong, it's a better description than Bayesian inference of most people's intuitive concept of the task of a juror. In other words, when I imagine myself as a juror I'm automatically more concerned about a false positive (convicting an innocent person), and I will intuitively try to answer the question "has the prosecution proved its case" rather than "is this person guilty." If asked to answer the second question and quantify my odds of guilt, I'm likely to understate them, precisely because I can't separate that estimate from the real-world effect of a guilty verdict. Or in your terms, the "question of what probability corresponds to 'beyond reasonable doubt' [or whatever the equivalent standard in Italy]" can't be completely excluded from the question when we imagine ourselves as jurors, only made implicit. This reminds me slightly of Eliezer's "true Prisoner's Dilemma" article, which I really liked. Just as you can't posit that someone is my confederate (in his case) and then ask me to consider them in a purely selfish, impartial way -- you can't tell me I'm a juror and then ask me to make a purely impartial assessment. I'm describing a much weaker effect than he was, and maybe it's more socially conditioned than inherent to human nature, but I think the general concept is the same. So ...better to say "forget the fact that there's even a trial going on, just imagine that tomorrow the absolute truth will be revealed and you have to bet on it now."
This is an interesting point and one where I think the legal system is wrong from a strict rationality sense but I can see the argument given that juries are human and so not very rational. It is common for juries to either not be given information which is very relevant to the prior probabilities of guilt or instructed to discard it. The debate in the UK over whether previous convictions should be admissible as evidence is a good example. From a strict Bayesian/rationality point of view, all information is potentially relevant and more information should only improve the decisions made by the jury. Information about previous convictions is very relevant when determining priors for certain types of crimes, particularly sexual offences which triggered calls for changes to the law in the UK. The counter argument is that telling juries about prior offences will bias them too much against the defendant, in recognition of the fact that juries are prone to certain kinds of irrational bias. The rules about keeping juries away from potentially prejudicial media coverage exist for similar reasons. The failure to avoid this in the Amanda Knox case is one of the criticisms leveled against the prosecution by her supporters.
From a really strict Bayesian point of view, more information can certainly make decision worse. Only perfect information (or, perhaps, arbitrarily close-to-perfect information??) necessarily makes decisions better. Of course, perfect information includes the bit of information saying whodunnit. Here's another way to put it. Given a jury, you (the court) can give them information that will cause them to decide guilty, or you can give them other information that will cause them to decide not guilty. In many cases both sets of information will be true - just different subsets of the complete truth. How do you decide what to tell them?
Not true. A perfect Bayesian updater will never make worse decisions in the light of new information. If new information causes worse decisions that is a reflection that the new information was not appropriately weighted according to the trustworthiness of the information source. In other words, false information can only make for worse decisions if it is treated as true. The only reason you would treat false information as true is that you placed too much trust in the source of the information. The problem is not the receipt of the new information, it is incorrect updating due to incorrect priors regarding the reliability of the information source. That may be a common problem for actual imperfect humans but it is not an indication that acquiring new information can ever lead to worse decisions for a theoretical perfect Bayesian updater.
That's not quite right. The provision of all true but biased information (e.g. only those facts that are consistent with guilt) without complete awareness of the exact nature of the bias applied can increase the chances of an error. Even unbiased info can't be said to always help. A good example is someone who has crazy priors. Suppose someone has the crazy prior that with probability .99999 creationism is true. If they have somehow aquired evidence that overcomes this prior but further information about problems with evolutionary theories would leave them with still strong but not convincing evidence that evolution is true then providing them with that evidence increases their chance of error. More generally, disagreement in priors forces one to believe that others will make better decisions if evidence that exacerbates the errors in their priors is provided.
Additionally to my first reply, even if all the data the updater is given is completely true, incomplete data can still lead to worse decisions. Here's a trivial example. This is a list of true facts (not really, just in the example): (1) B was murdered. (2) A was in town on the day of the murder. (3) A is green. Green people are less likely to commit murder than the general population. (4) A was B's close friend. Most murders are commited by friends. A perfect Bayesian judge is asked: given fact (1), did A murder B? He has some prior probability for this. Then he is given fact 2. His probability (of A's guilt) goes up. Then he is given fact 3; the probability goes down. Then fact 4; it goes up again. And so on. This works independently of whether A murdered B or not.
I think you're confusing two things here: a) is the judge making the best decision in light of the information available to him and b) in light of new information is his probability moving in the 'correct' direction given what really happened. The question of b) is irrelevant: the best we can hope for is the judge to make the best possible decision given the available information. A perfect Bayesian judge will do that. The problem in the real world of humans is not whether giving them new information will lead them closer to or further from 'the truth' but whether they will update correctly in light of the new information. The reason certain facts are withheld from juries is that it is believed they will not update correctly on the new information but rather will be consistently biased in a particular direction by it to an extent not warranted by the facts.
That's right. That's not right. Even if the juries always update correctly on the new information they may still become more distant from the truth. The jury may be performing your (a) perfectly, but we do really want (b). My point was that even with a perfect Bayesian jury, the disrepancy between executing (a) and (b) will cause us to withhold certain facts sometimes, because the partial presentation of the facts will cause the jury's (a) to be closer to the actual (b), the truth.
We might really want to ride a unicorn as well but it's not really an option. Well, at least riding a unicorn is logically possible I suppose, unlike making decisions better than the best possible given the information available... The only way to get closer to the truth when you already update perfectly is to seek out more information. The legal system is not designed around perfect updaters for the same reason it's not designed around unicorns. We wouldn't need judges and juries if we had perfect Bayesian updaters - we could just make those updaters investigator, judge, jury and executioner and tell them to deliver punishment when they reached a certain threshold level of probability. The idea of withholding certain information from jurors is predicated on the idea that jurors are less good updaters than judges. Whether that is true or not is another question.
This is all true about our system. But my point still stands: even with perfect updaters there can still be a reason to withhold information. It's true that it's usually an insignificant concern with human juries, because other problems swamp this one. You originally said: if "improve" means "bring their decisions closer to the objective perfect-knowledge truth" then that statement is false, as I have explained. I don't see what else "improve" can mean here - it can't refer to the jury's correctly updating if we assume that their updating is perfect ("strictly Bayesian").
The only way for a perfect Bayesian updater to move closer to the truth from its own perspective is to seek out more information. Some new pieces of information could move its probability estimates in the wrong direction (relative to the unknown truth) but it cannot know in advance what those might be. Another agent with more information could attempt to manipulate the perfect updater's beliefs by selectively feeding it with information (it would have to be quite subtle about this and quite good at hiding it's own motives to fool the perfect updater but with a sufficient informational advantage it should be possible). Such an agent may or may not be interested in moving the perfect updater's beliefs closer to the truth as it perceives it but unless it has perfect information it can't be sure what the truth is anyway. If the agent wishes to move the perfect updater in the direction of what it perceives as the truth then its best tactic is probably just to share all of its information with the perfect updater. Only if it wishes to move the perfect updater's beliefs away from its own should it selectively withhold information. 'Improve' for a perfect Bayesian can only mean 'seek out more knowledge'. A perfect Bayesian will also know exactly which information to prioritize seeking out in order to get maximum epistemic bang for its buck. A perfect Bayesian will never find itself in a situation where its best option is to avoid finding out more information or to deliberately forget information in order to move closer to the objective truth. An external agent with more knowledge could observe that the perfect Bayesian on occasion updated its probabilities in the 'wrong' direction (relative to the truth as perceived by the external agent) but that does not imply that the perfect Bayesian should have avoided acquiring the information given its own state of knowledge.
Not so. The agent in question has an information advantage over another, including information about what the intended pupil believes about aspiring teacher. It knows exactly how the pupil will react to stimulus. The task then is to feed whichever combination of information leads to the state closest to that of the teacher. This is probably not sharing all information. It is probably sharing nearly all information with a some perfectly selected differences or omissions here and there. Dan's point still stands even in this idealised case.
The updater may be perfect, but because the updater's knowledge is imperfect, the updater cannot correctly judge the reliability of the source of information, and therefore it may assign that information incorrect (imprecise) weight or even take false information for true.
We're talking about a hypothetical perfect updater that always updates correctly on new information. If it's updating incorrectly on new information due to putting too much trust in it given it's current state of knowledge then it's not a perfect updater.
What you call a perfect updater here is an agent with perfectly complete knowledge. That's the only way to always judge correctly the weight of new information. Of course, such an agent never needs to update, at least not about past events.
No, I'm not talking about an agent with perfect knowledge. I'm talking about a perfect updater. A perfect Bayesian updater comes to the best possible decisions given the available information. Giving such a perfect updater new information never makes it's decisions worse because by definition it always makes the best possible decision given the information. This is a different question from whether it's probability estimates move closer or further from 'the truth' as judged from some external perspective where more information is available. The concern with imperfect updaters like humans is that giving them more information leads them further away from the theoretical best decision given the information available to them, not that it leads them further away from 'the truth'. In other words, giving people more information can lead them to make worse decisions (less like the decisions of a perfect Bayesian updater) which may or may not mean their opinions become more aligned with the truth.
These are both concerns, and if we could replace humans with perfect Bayesian updaters, we'd notice the only remaining concern a lot more - namely, that given more (true) information can cause the updater to move away from the objective truth we are trying to reach (the truth that is only knowable with perfect information).
Who would decide which information to withhold in that case? The only way you could be qualified to judge what information to withhold would be if you yourself had perfect information, in which case there'd really be no need for the jury and you could just pass judgement yourself. The only way for a perfect updater to get closer to the truth is for it to seek out more information.
That's a strong claim. Is there a formal proof of this?
I don't think a formal proof is needed. An agent with imperfect knowledge does not, by definition, know what 'the truth' is. It may be able to judge the impact of extra information on another agent and whether that information will move the other agent closer or further from the first agent's own probability estimates but it cannot know whether that has the result of moving the second agent's probability estimates closer to 'the truth' because it does not know 'the truth'.
Point taken. If we assume the Court-agent can effectively communicate all of its knowledge to the Jury-agent, then the Jury can make decisions at least as good as the Court's. Or the Jury could communicate all of its knowledge to the Court and then we wouldn't need a Jury. You're right about this. But as long as we're forced to have separate Court and Jury who cannot communicate all their knowledge to one another - perhaps they can only communicate all the knowledge directly relevant to the trial at hand, or there are bandwidth constraints, or the Judge cannot itself appear as witness to provide new information to the Court - then my point stands.

I ended up spending a lot of time looking over the two provided sites, without having ever heard of the case before.

The parts of the "true justice" site that I found most helpful were:

Whereas the FoA site seemed very well put together.

The primary pieces of evidence that shape my current beliefs (... (read more)

Sorry, how do you get 99.9999% as a unit weighted combination of 99 and 1 ?!
the 99 and 1 are approximations. Although my primary motivation is that P(it mattering) is very very small. Also it's possible that there is crazy shit and he is still guilty.

For those of you who were not familiar with the case, which side's arguments did you read first?

Did it feel like this order influenced your reasoning or final judgment?

(I read pro-guilt ideas first, and feel now that this ordering contributed to my judgment that Knox and Sollecito are guilty. I wonder how conviction rates would change in the US if we let the defense go first, i.e. what are the relative influence values of the right to speak first vs. last?)

The principled position is to read the arguments in favor of non-guilt first because we generally believe in the presumptions of innocence and therefore should set up our minds with a bias to innocence rather than a bias to guilt when you have to have a bias but can't control it.
I flipped a coin, and it landed "read pro-guilt first". My take on its effect is basically the same as Cyan's.
I read 'friends of amanda' first, and was definite biased towards her innocence and guede's guilt for much of the rest of my reading. I didn't find anything particularly compelling against knox. The last thing I looked at was the wikipedia article, which knocked down my confidence in guede's guilt a lot. Overall, I think I compensated for the first impression I got. I'm not so sure I compensated for the fact that the pro-guilt site was stupidly 'think of the poor murdered victim, there has to be justice, and by justice we mean guilty verdicts.' The pro innocence side definitely struck me as more rational.
This. I was thoroughly annoyed by the sites layout and structure. If your main focus is on pictures of the victim and youtube videos, you don't really have a lot of arguments. Friends of Amanda looks a lot more professional and the main points are much more condensed. Estimates at three at night, very tired but sober. First tried to read the pro- and contra-sites, but I was too confused by the layouts and didn't know where to start. Then went on reading WP article and was pretty sure about Guede being the sole perpetrator. Read the other sites again and found the guilty-side very unconvincing, the not-guilty-side much better. The not-guilty-side reminded me specifically of the holocaust deniers tactics which I know very well. Couldn't force myself to read for longer than 30 minutes in total because I am very tired, might read something tomorrow but not expecting to change my opinion at all. Knox guilty: 25% Raffaele: 25% Guede: 75% Agreement: 60% After reading the comments, my estimates changed to 10/10/90 I hope this case goes to a European court or something. It really is a shame.
I read the pro-guilt side first. Whatever order effect there might have been was more than counteracted by the amount of crap I read through before getting to the actual evidence presented by the prosecution.


I haven't made up my mind yet - and if anyone's interested, this cbs take on it looks well done:

Context: I didn't know anything about the case; I think I've overheard something about it in the past, because I had a very slight sensation of déjà-vu reading your post, but I didn't really know anything. The info I used was reading the Wikipedia pages (one you linked and one on one of the trials), then skimming the sites you linked for less than a half hour.

I have reached a “decision” about it, but I didn't intend to post it. Then curiosity got the best of me and I read pretty much all the comments already here. My “estimate” wobbled a bit during the rea... (read more)

No, "X killed Kercher" is right. (Or perhaps better would be, "X participated in killing of Kercher".)
It's an academical point right now, given your later post, but even “X participated in killing of Kercher” is not entirely clear. Any action A, from cutting her throat, to holding her, to opening the door, to simply being in the room, can be considered as “participating” in the murder, in the sense that suitably complicated scenarios can be constructed where X not doing the action A would have avoided the murder, with X knowing it and deliberately choosing to do A in order to cause Kercher not to live anymore. I imagine what you actually wanted to ask was something like the probability that “X is guilty of whatever X was convicted”, though that only moves the uncertainty about what you ask to uncertainty about what Italian law says.

I've looked at this twice - first after reading the friends of amanda blog, wikipedia, and scanning the justice for meredith blog.

My initial probabilities were: P(AK=guilty) = .55, P(RS=guilty)=.5, P(RG=guilty)=.999, P(views coincide)=.5. Having read a few comments I initially revised the first two probabilities down - I realised I was guilty of having given a lot of weight to the rape story, and not given weight to the improbability of the "weird sex" story.

Having read more I find it hard to be sure of anything - it seems to be next to impossibl... (read more)

Keep in mind, this bias may not be entirely unjustified. The guilty blog quotes a major Italian newspaper (it says) which itself jokes about the Italian's system 'near biblical' slowness and forthrightly admits that it is the target of much legitimate criticism. And then there's the general black market economy of Italy, tax evasion, and dispect for the law. The Maxi Trial [] is an interesting example, without so far as I know, any American analogue: (My apologies for the lengthy quoting, but does this sound like a peaceful highly law-abiding nation, with an effective and uncorrupt judicature?)
I'd be careful about generalising from the south of Italy (Sicily) to the north - there's a famous division between the two parts of the country, to the extent that many believe in formally splitting the country. And I'm certainly not interested in which system is superior, American or Italian - the answer is clearly Canadian. What I think is interesting about this is that the decision comes down to whose judgement you trust least: * My judgement is clouded by lack of access to evidence and a lack of access to unbiased evidence. I feel I am unbiased because I have no axe to grind, but these websites expose me to every form of prejudice - I am sure it has an effect. * The jury convicted Amanda - a jury is only supposed to convict where guilt is proven beyond reasonable doubt. The jury has access to the evidence and hours to examine it. However juries do sometimes give incorrect verdicts where the victim is an attractive woman (e.g. the Jill Dando case in the UK). * The police worked hard to collect a lot of evidence. However the prosecutor appears to have a sexual obsession, and the police failed to record interviews with suspects. * The existence of a group like Friends of Amanda suggests that many people think the case is not robust. However Amanda is an attractive female, being tried in a foreign country. And of course mothers never think their children are guilty. * Meredith's parents and the British press are strongly against Amanda. The British Tabloids are not worth the paper they're written on. All in all, I think that I have no chance of making an unbiased and accurate judgement on the first hand evidence. Based on the fact that she was found guilty in a court of law, in northern Italy, and given that there was so much evidence, much of it from Amanda herself, I think she is probably guilty. However even with 99% probability I still wouldn't convict - 1 in 100 is a reasonable doubt. Check out Amanda's note by
Those are very high probabilities.

By guilty, do we mean "committed or significantly contributed to the murder"?

Or do we mean "committed or significantly contributed to the murder AND there is enough evidence showing that to satisfy the beyond-a-reasonable-doubt (or Italian equivalent) standard of proof for murder"?

The comments don't seem to make that distinction, but I think it could make a big difference.

3Eliezer Yudkowsky13y
probability = probability of having committed murder, not probability of sufficient evidence
I interpreted it as 'how would you vote if you were on the jury', which implies 'guilty beyond reasonable doubt' under the legal systems I'm familiar with. I don't know if the standard is any different in Italy.
I agree that that's one reasonable interpretation. I just want to emphasize that that standard is very different than the weaker "if I had to guess, I would say that the person actually committed the crime." The first standard is higher. Also, the law might forbid you from considering certain facts/evidence, even if you know in the back of your mind that the evidence is there and suggestive. There are probably other differences between the standards that I'm not thinking of.

ak guilty: 0.1

rs guilty: 0.1

rg guilty: 0.9

assessment of my opinion vs poster's opinion: can't guess at your actual probabilities, but probability similar general take identifying guede as vastly more likely to be guilty than the other two.

sources: small number (i estimate <5) of media stories covering the case that i've seen over recent months plus the wikipedia article as it read at a little after 4 pm today.

I spent less than an hour browsing the two links provided and randomly surfing elsewhere afterwards. I was unfamiliar with the case before.

My estimated odds of Raffaele Sollecito being guilty of murder are very roughly 1:1, of Amanda Knox being guilty of murder a little less, of Ruedy Guede being guilty closer to 4:1.

Unfortunately I couldn't entirely avoid having a peek at comments before coming to an estimate - there were some numbers in the right-hand "recent comments" box of the site.

The main driver in my estimates is that all three were convi... (read more)

Italy does not have voir-dire (juror screening) as in the U.S.
Are they chosen by simple sortition? (randomly)
  1. p = 0.5
  2. p = 0.5
  3. p = 0.95
  4. I have pretty high confidence (~95%) that you got the right answer, which doesn't tell me which side you're on for the first two, but means you'll likely agree on the third.

As a disclaimer, I did peek at other peoples comments before writing these down, but I had mentally committed to these numbers before doing so. I had heard of the case a while ago though I never looked into it. I had the impression that Knoxx and Sollecito were guilty (based solely on the fact that they tend to get the right bad guys) and didn't remember hear... (read more)

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.
2Paul Crowley13y
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: [] [] 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.
0Paul Crowley13y
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".
0Paul Crowley13y
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.
0Paul Crowley13y
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.
0Paul Crowley13y
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.

I'm putting myself in the "low familiarity" category.

  1. estimate Amanda Knox is guilty: 40%
  2. estimate Raffaele Sollecito is guilty: 40%
  3. estimate Rudy Guede is guilty: 90%
  4. not confident at all that my estimates correspond to komponisto's

I got most of my information from the "defendants are innocent" link and the Wikipedia article - I found the "defendants guilty" link hard to read or assess. I did not spend much time forming my opinion, just a quick browse.

  1. 5%
  2. 5%
  3. 95%
  4. probably generally the same - you might be more or less extreme.
  5. The sources are the two sites you linked to, plus the wikipedia article plus my impression that 1 man killing a girl in a sexual assault is a priori much more likely than three relatively recent acquaintances doing same.

NB - 5% and 95% are basically just proxies for 'pretty sure it's not true' and 'pretty sure it's true'. I wouldn't claim that they're anything like well-calibrated. I don't think I form opinions on stories like this often enough to be well-calibrated.

Oh, and I sec... (read more)

I had heard a little about this case before you brought it up, but I didn't want to read about it since I figured I'd get angry about either: 1. Yet another travesty of justice. or: 2. Yet another obviously guilty person being cast in a good light. It was the former. Unrecorded interrogations without a lawyer. No forensic evidence putting Knox in the room. No motive, no sequestered jury, tons of disputes about forensics. What a mess.

Your probability estimate that Amanda Knox is guilty.


Your probability estimate that Raffaele Sollecito is guilty.

0... (read more)

I'm moderately familiar with the case (have read a few media articles on the case over the last few months).

  1. 10% (weak evidence, lack of motive, low prior probability of female on female sexually motivated violence)
  2. 30% (evidence not much better, higher prior probability of male on female sexually motivated violence)
  3. 50% (seems to fit profile of this type of offence, evidence of association with victim and opportunity and circumstantial evidence)
  4. I expect your relative rankings to be the same, less confidence in absolute probabilities.
  5. Fairly mainstream m
... (read more)
After reading a bit more I'd revise my probabilities to: 1. 5% 2. 10% 3. 90% The evidence against Knox and Sollecito is very weak and there doesn't seem to be clear motive. The evidence against Guede is strong and it seems clear that had he been identified initially as the primary suspect there would have been no reason for the police to suspect anyone else was involved in the attack. I think there's a very high probability Guede is the only guilty party. There is a small probability that all 3 conspired in some way and an even smaller probability that Knox and Sollecito conspired together.
The spread between the base prior for male on female sexual violence vs. female on female sexual violence is nowhere near enough to separate the percentages for Knox and Sollecito this widely, let alone as widely as in your first cut. The general prior may be 2-3 times as likely, but that spread should not hold up after a lot of additional evidence to consider where they are essentially equal. Every piece of evidence that counts for or against them roughly equally should shrink that spread in ratio terms. And it should not expand it in absolute terms at all either. Given that these priors are extremely low (or should be), well under 1% for male or female, there shouldn't be more than a 1% spread in those simply from that prior. Only evidence which would amplify the male v. female disparity in prior, or evidence that specifically argues in favor of S's but not K's guilt, should get you to a spread of more than 1% between them.
According to this [] page, only 6.4% of offenders in sex related homicides are female (lower than the 11.2% of all homicide offenders who are female). Of all homicides only 2.4% are female on female. I can't see a simple way to derive the percentage of female offenders in sex related homicides with a female victim but it seems likely based on the other numbers to be lower than p(Female Offender|Female Victim) which I make to be 9.6%: p(FO|FV) = p(FV|FO)*p(FO)/p(FV) = 2.4 / (2.4 + 22.7) = 0.096. So given no other information, if you know you have a female victim in a sex related homicide it would be reasonable to assume that it is at least 10 times more likely that the murderer is male. About 91% of female murder victims are killed by someone known to the victim so it is reasonable for the police to start with her close associates. Given no particular reason to favour Knox or Sollecito as the murderer you would put much higher odds on it being Sollecito purely based on his gender. That was roughly the reasoning I used with my initial estimates. I hadn't looked up any statistics at that point, I just knew that a female murder victim was significantly more likely to have been killed by a male than by a female, especially if there appeared to be a sexual motive to the murder. In light of the statistics I think if anything I should have put a wider gap on the estimates in the absence of other evidence implicating one over the other.
I don't know anything about this specific case, but in general - we need to adjust this kind of statement for selection bias, and I would like to know how this is done. When killers aren't known to the victims, the police are less likely to find them both because they don't look for them (as you said) and because there are too many people the victim didn't know for the police to examine more than a small fraction. Therefore I expect the just conviction rate in these crimes to be lower, and the false conviction rate to be higher (the police are always likely to accuse and help convict someone the victim knew, but in these cases we know such an accused is innocent). However, your statistic (many murder victims knew their attacker) probably actually counts convictions. How should we revise this due to A) false convictions and B) a different-than-general rate of crimes solved (where conviction was achieved, or the primary suspect died or fled)? If the statistic already takes such considerations into account, how it this done?
Do you happen to remember from what country or countries these sources were? (Obviously most stories about this case have been from the US, UK, or Italy -- with, according to some, significantly different biases among them.)
I believe at least one of the articles was in the New York Times. Most of the blogs I read would link to US or UK sources and any coverage I saw on TV would likely have been from BBC World or from US sources.

I have never heard of this case before this post. After approximately 30 minutes of reading the two sites you linked to as well as the Wikipedia article, my current estimates are:

  • Knox guilty: 0.35
  • Sollecito guilty: 0.35
  • Guede guilty: 0.80

I think your opinion will roughly coincide with mine about Guede, but could differ dramatically about the other two.

Update: After reading the other comments here, I'm revising to * Knox guilty: 0.20 * Sollecito guilty: 0.20 * Guede guilty: 0.70 * None of the three: 0.20
I was curious, so I checked: if the 3 questions were independent (clearly they're not), your estimate for none of the 3 guilty should be .192 I assume from your similar .20 probabilities that you see Knox and Sollecito's guilt as highly correlated. This would have the effect of raising your p(none) higher than .192. But on the other hand, if Guede is guilty, then that should decrease the chance that the others are. So, it seems you at least thought about what it means to give p(none) I say just give all 2^3 probabilities (one of which is redundant) :)

[comment deleted]

  1. How much you think your opinion will turn out to coincide with mine - hard to define. If your respective answers are 10, 40, 90, how much did we agree? I'll guess that the sum of the three differences between my answers and yours is around sixty percentage points [out of 300 possible].

I'm going to have to distinguish here between guilt in the actual sense, and guilt in a legal sense. Do I think Amanda Knox did it? Somewhat likely. Do I think the prosecution proved that beyond a reasonable doubt? No.

I think my estimates of guilt for all three parties will be higher than most commenters, but here they are:

Probability that Knox participated in the murder: 15% Probability that Knox participated in or covered up the murder: 20% Probability that I would find Knox guilty of murder: 5%

Probability that Solecito participated in the murder 10%... (read more)

Although I'm invariably annoyed by this kind of (what seems to me like) weasly hedging ("just state your probability already!"), it might be a reasonable thing to say if your probability is somewhere between 50% and 99%. At 15%, however, I hardly see the point, and in fact it's downright misleading.
How is that 'weaselly'? Say there is a criminal who confesses to a crime, and quite obviously did it, but the police failed to properly Mirandize them, or otherwise unlawfully elicited the confession. Legally, you should find them not guilty, even if they likely committed the crime. Not guilty does not equal innocent.
That's a separate matter entirely. (Actually, my understanding of the way it's supposed to work in such a case is that you're not supposed to ever hear about the confession as a juror, and so may not have enough evidence to rationally believe they're guilty with the required confidence, even if they in fact are; as opposed believing them guilty beyond a reasonable doubt, but just "finding" them not -- a form of jury nullification.) This discussion here is entirely about one's rational belief about guilt or innocence, not anyone's legal opinion on the admissibility of evidence into court. You said your probability estimate that Amanda Knox killed her roommate was 15%. That's the information of interest, and it makes you a firm innocentista. Legal issues are a red herring. They would be a red herring even if your belief was 75% -- but in that case, it would at least be a legitimate discussion point to say,"thus, although I believe she probably did it, I would have to acquit if I were on a jury, because I don't believe it beyond a reasonable doubt."

How does one get to vote? I say, not guilty except Guede. What an embarrassment and nobody wants to admit their crooked government employees are nuts and criminals. Someone needs to look closely at Mignini and his motives, He sounds like a religious nut whose small mind saw one thing and then spoon fed the tabloids. And found guilty of "crimes in regard to evidence". This is a complex case. These two do not deserve to be in jail. Guede does and then the Perugian Magistrate, prosecutor and the tabloids should be held accountable for the crimes ag... (read more)

I have had a ringside seat to this debacle for three years and have never doubted the innocence of Amanda and Raff for even a day.

What is troubling is the number of folks out there who WANT to believe in their guilt without ever reading the information available to everyone reading this.

Amanda and Raff 0.00000% guilt Rudy 100.0000000% guilt.

The actual facts bear witness to my opinion. The Smear Campaign that occurred in the first three weeks after the murder continues to perpetuate the false opinion that these two innocent college students are depraved.... (read more)

[-][anonymous]12y 0

I have a ring side seat to this debacle and have never doubted Amanda and Raff's innocence. It is heartbreaking to see how many people WANT to believe they are guilty without ever taking the time to study the facts of the case. Most of which is at the fingertips of everyone reading this.

I find it a very simple case with tons of evidence and tons of lack of evidence that point to freedom. And to something SERIOUSLY wrong somewhere. I love America, I love Italy, I love truth and hate injustice. Fact over opinion. Show many facts...

[-][anonymous]13y 0

I find it a very simple case with tons of evidence and tons of lack of evidence that, to me, make any other scenario bizarre.

I'd love to know what Amanda and Raffaelle got up to that night but the lack of DNA in the room and on the body suggests that whatever they did, they weren't in the room or directly responsible for the death, and nor did they go back in the room to move the body around - that would require head to toe covering. But...

Did Amanda and Raffaelle sit in the flat egging Guede on, not realising the screams were real? Or, worse, did they laugh knowlingly when they heard screams?

What would they be guilty of? Would either scenario count as murder?

Did they feel so s... (read more)

Not very familiar -- had heard about it, but no real details.

I'll echo all those that have raised concerns about prosecutorial and police misconduct, releasing details that later turned out not to be true, or were entirely irrelevant. The group sex games focus also reminds me strongly of many other moral panic cases where people flip out for no good reason.

I don't find the "cleaning" hypothesis very well supported -- it's hard to selectively clean, which would be necessary to eliminate all of their own evidence, but still leave so much Guede evidence.

overall: P(Knox) ~ 0.02 P(Sollecito) ~ 0.02 P(Guede) ~ 0.98

I was aware of the case before, but hadn't looked into it in any detail. My reaction to the sites is that the site arguing innocence seems to be presenting facts and showing contradictions in the other side's arguments. I couldn't find any consistent argument on the other side. There were many scenarios, with inconsistent adherence to the facts, lots of innuendo and plausibility arguments for particular claims, but no coherent story.

  1. Your probability estimate that Amanda Knox is guilty. less than 30%
  2. Your probability estimate that Raffaele Sollecito
... (read more)

I basically agree with Lordweiner. 90/90/99.

It seems pretty clear that AK and RS were and are hiding something damaging to them. Their stories are just too changing and contradictory.

That damaging fact is almost certainly that they were involved in the murder. Because if it were something else, they would have to be insane not to come clean about it.

ETA: I came up with my estimate before reading anyone's posts.

I don't remember hearing about this case before reading this post. This is not strong evidence that I hadn't; it's the kind of news item I'd be likely to classify as "contains no valuable or interesting information" (no disrespect to the victim, her friends or her relatives is intended; but the global death rate is such that you just can't keep up with all the individual instances, and therefore you necessarily have to be extremely selective about which individual deaths receive any specific attention from you) and discard immediately from memory... (read more)

It turns out most commentors are of a rather different opinion on 1. and 2., so ... update time! New evidence includes reports that use of the involved drugs does significantly impair memory [], a reminder that absence of evidence is evidence of absence [], and fairly damning judgements about the Italien legal system. I also realized that my priors for P(group rape | rape) and P(group murder | murder) were insufficiently extreme (i.e. too high). All of this is fairly general evidence about such cases, so the jury could well have had access to it, as well. They're still in a position of having much better access to evidence, and the real question becomes how rational they are in evaluating it. Updated probabilities are: 1. 0.35 2. 0.35 3. 0.9

AK guilty: 90%; RS guilty; 85%; RG guilty: 99%. Probability that komponisto agrees that they are all probably guilty: 80%.

The more I read, the more it seems to me that the "pro-guilt" side has a case like the scientific case for the "pro-evolution" position, namely conclusive arguments that take a while to understand, while the "pro-innocence" side has a case like that for creationism, namely plausible arguments that do not stand up under scrutiny.

Evidence that komponisto agrees with this is another accidental similarity with ... (read more)

Does anyone else think that if Knox is innocent she only has herself to blame for changing her story so many times and for implicating an innocent man?

Maybe that's harsh of me to say and if she is innocent she obviously doesn't deserve jail but it's kind of hard to feel sorry for her.

She has herself to blame, but not ONLY herself to blame. Lying to police, despite being innocent, is obviously an incredibly stupid thing to do... but not nearly as uncommon as TV and the movies would have you believe. When people are accused of serious crimes, they freak out and do stupid things, like enhance their alibi or come up with stories to explain away anything that looks incriminating, or wildly speculate about alternate theories of the crime. Stupid yes, rare no.
I agree. I can't comment on how rare or common it is though. But I'm certainly going to be careful not to lie when I inevitably end up in court.

1:25% 2:25% 3:50% 4:60% on 1&2

I knew virtually nothing about this an hour ago.

I didn't know anything about this case until I read this post.

My first impression: Wow, there sure is a lot of information on what a nice person Meredith Kercher was on that one site, and what a nice person Amanda Knox is on that other site. (The latter is at least potentially relevant, if you think a nasty person is more likely to kill someone.)

Anyway: Knox : ~35%; Sollecito : ~35%; Guede : ~95%. And I'd say it's about 60% that all of these probabilities are on the same side of 50% as yours.

(I had a bunch of information here on how I came to my opinion, b... (read more)

Both sites are sources of information presented by people with extremely strong emotional investments in the case. The Guilty site seems to be populated by nothing but emotional arguments (even looking through the Micheli Report link only produced second-hand commentary couched in strong emotional terms, i.e., "our translators cried..."). The innocent site is less emotional. Neither provide objective third-party sources of information, so taking either side's arguments as "facts" seems to be undertaking a bias to begin with.

Even after killing an hour going through the sites, I'd have to abstain.

I attempted the test with zero familiarity with the case at hand. I also have very little knowledge of the Italian justice system.

One major problem in presenting a probability assessement is that the links presented in the post offer pratically no facts about the case. They are about a Washington Senator's reaction, instead. It would be ludicrous to answer the questions asked given only this information.

So, I went googling around for more information. I promptly hit a snag in that I do not what a fast-track trial is in the context of Italy, and searchi... (read more)

You must have only looked at the front pages of the two sites. You have to browse around somewhat to find the information. I suggest starting here [] on Friends of Amanda, and here [] on True Justice.

You should have sent the readers directly to that information, then. LW has thousands of readers, so putting in 10X work yourself to save thousands of readers X time is generally a good idea.

Well, as I said, the information is not all in one place (particularly on TJ; FoA is better organized), and I was worried about biasing readers via my selection of the first page to read. (In fact, as I indicated in the post, I was even somewhat worried about biasing readers via my selection of the sites themselves.) Most of the commenters seem not to have had problems. For the few that did, I don't mind giving a little more direction to them individually.
What makes you think that Sollecito is more likely to be guilty than Knox, and possibly more than 15x likely? (Which is quite the difference.) What likely scenarios are you envisioning that have Guede and Sollecito murdering her without Knox being involved? It seems to me that all the plausible or presented scenarios have Sollecito & Knox acting in concert or in some way covering up for the other, and so their guilts would be closely linked.
Simple. Sollecito's DNA was found there, but not Knox's. I think it highly unlikely that the DNA has any link to the actual murder, but its presence justifies an allowance of probability mass. Less than 0.1 and 0.15 are both very low probabilities.
Ahhhh... 0.15 isn't 15x 0.1 It's probably justified simply by priors re male/female committing rape/murder
<0.1 includes values like 0.00000001, y'know. So technically I'm not wrong... But you're right, I saw .1 and .15 and forgot that that was 1.5x, not 15x. (That would require .01)
  1. Your probability estimate that Amanda Knox is guilty.


  1. Your probability estimate that Raffaele Sollecito is guilty.


  1. Your probability estimate that Rudy Guede is guilty.


  1. How much you think your opinion will turn out to coincide with mine.

p(same direction) = 0.6

p(someone in the prosecution is guilty of actions I would give live imprisonment for) = 0.99.

It's good that you gave two separate links to look at both sides of the case. However, I really disagree with you suggesting people to look at wikipedia. Wikipedia is known to be horrifically unreliable when it comes to controversial topics. This much can easily be seen just by scrolling through the article's "discussion" subpage, which contains a lot of ouright allegations of impartiality from various wikipedia editors. I've seen a few people posting their assessments based on their reading of wikipedia, and this is a big problem.

For further il... (read more)

I haven't followed the instructions precisely, due to lack of time, but will answer anyway in the interest of avoiding self-selection bias. (I'm also not reading any of the other responses yet.)

I had heard of the case, but not seen any mainstream media reports. It was probably on one of the blogs I read.

I spent about 5 minutes looking at the "guilty" site, and about 1 minute looking at the "innocent" site. My first reactions to the two sites were:

(1) Wow, the "innocent" people aren't really trying very hard. Is this really the... (read more)

[-][anonymous]13y 0

I was unfamiliar with the case. I came up with: 1 - 20% 2 - 20% 3 - 96% 4 - probably in the same direction, but no idea how confident you were.

From reading other comments, it seems like I put a different interpretation on the numbers than most people. Mine were based on times in the past that I've formed an opinion from secondhand sources (blogs etc.) on a controversial issue like this, and then later reversed that opinion after learning many more facts.

(I didn't put any evidential weight on the jury's ruling. Conditioning on the simple fact that their ruling is controversial screens off most of its value.)

I think this is an interesting exercise which can be quite useful even if we can't find the answer in the back of the book afterwards.

It'd also be interesting to see what everyones second guess is, after conditioning on the beliefs in the comments.

I'm now "pretty sure" that Knox and Sollecito are innocent and Guede is guilty, though I'd have to think a bit in order to translate to numbers.

My familiarity with the case is low:

  1. p = 0.40

  2. p = 0.40

  3. p = 0.90

  4. I think we will agree with the general idea that Guede is likely to be guilty while Knox and Sollecito are unlikely to be guilty.

From what I could gather, the physical evidence against Knox and Sollecito is pretty weak and the force of the prosecution's argument is supposed to come from the inconsistencies between Knox's and Sollecito's accounts of the night. While there were more inconsistencies than I would have expected (judging just on the general fallibility of human memory), I felt... (read more)

I don't. She changed her story at least three times. And I'm sorry but if I was an innocent suspect in a murder trial you'd have to rape me up the arse with a barb wire dildo to get me to change my story from the truth.
But the question is, are you typical in that respect? That's not so clear.

One potential problem is that the first site is organized like a blog or forum, and thus it is hard to find a quick summary of the case there.

Orin Kerr thinks (or dreams) that's the future of legal briefs.

I didn't read the material and don't have an opinion. I had heard of the case before from sources pushing for innocence.

This might be a naive question, but why hasn't Guede cleared up any confusion about Amanda's/Raffaele's involvement? This seems important enough that the police/prosecution would somehow make him cooperate if he's refusing to (e.g. strike some deal or something).

His testimony doesn't appear to be very reliable. At least according to this New York Times reporter []
  1. Amanda Knox guilty: .85
  2. Raffele Sollecito guilty: .6
  3. Rudy Guede guilty: .2
  4. Similar opinion: .6
  5. Information: Your links
Not trying to enforce groupthink here, just sincerely curious: could you elaborate on your reasoning a bit? The pro-guilt site doesn't seem to have much in the way of coherent arguments - hopefully a fellow rationalist could do a better job of explaining the other side's thought process. I'm especially baffled by the 0.2 for Guede - it seems like the only possible explanation for his innocence would be a massive conspiracy/frameup by police... which, don't get me wrong, I would never be so naive as to assign p=0, but if that's what you think happened, how do you get from there to the other two being probably guilty? What's really confusing me I think is that your combination of the 3 values (guede innocent, knox and sollecito guilty) doesn't seem to correspond to any theory of the crime being proposed by anyone on either side. Oh, and since I'm posting I guess I should play by the rules and post my own opinions (which I arrived at before reading comments): 1. 0.1 2. 0.1 3. 0.95 4. 0.6 5. Provided links
After spending more time reading on the subject (damn you lw for sucking up so much of my time! :), I'm revising my numbers, in the direction of greater certainty: 1. 0.01 2. 0.01 3. 0.99 4. 0.9 My reasons? Many small things... probably the biggest being: why didn't knox implicate guede??? If all 3 were there at the crime scene, then she knew that there was a third person involved, someone who: * didn't have an alibi * had left gratuitous amounts of forensic evidence * was a generally disreputable character * was black (playing on prejudice) * she didn't know well enough to care about protecting So if under interrogation you're being pressured to point the finger at someone, wouldn't he be the obvious choice, rather than someone picked at random from your phone book? And why didn't guede implicate knox/sollecito? When multiple parties conspire to commit a crime, I would strongly suspect that the most common behaviour under interrogation would be to point fingers at each other, not at random 3rd/4th parties.... that is, if the "co-conspirators" are actually aware of each others existence. As for my revising upwards guede's guilt, that's due to the bloody handprint - guede's handprint, in the victim's blood.... that's about as unimpeachable as forensic evidence gets, as it proves not just that he was at the crime scene at some point in time, not just that he had sex with the victim at some point in time, but that the victim was bleeding profusely while in his presence. That seems more solid than a signed confession to me. As to my original question to bgrah449... well, its a shame he's not replying... my best guess is that he was confusing guede and lumumba - thats the only thing I can come up with that would make those numbers make sense.
[-][anonymous]13y -1

Sorry, but I looked at both websites, and determined there's no way I'm reading that much material just for a little dubious calibrating. I'd rather calibrate on something where the outcome can be determined.

No opinion on the case, but I notice a lot of people are giving guilt probabilities that sum to well over 100%. Unless they are figuring in the options (two of these three are guilty) and (all three are guilty), this seems a bit off. Also, these options are apparently being given quite high implicit probabilities, sufficiently so that they should perhaps have been listed explicitly.

The Italian courts decided all three are guilty which immediately suggests that people are likely to be considering that possibility. Any combination of two of the three suspects is also plausible. There is no particular reason to expect people's probabilities to sum to 100%.
Ah so. Serves me right for not reading the provided links. :)