= 27d567bc2d48d9f40e9ccc20ea370b15)
komponisto comments on The Amanda Knox Test: How an Hour on the Internet Beats a Year in the Courtroom - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (632)
You're right, this is the part I have trouble understanding. Partly because I have trouble separating the facts of this particular case from the theoretical point you're making. So if you'll humor me for a second: let's say you're the chief of police somewhere, and one of your detectives comes back from a murder scene and tells you "we've arrested a prime suspect and we've got very strong evidence against him, including an ironclad DNA match that ties him to the murder. but we don't think he acted alone, because of X, so we want to keep looking for another suspect."
It sounds like for almost any X, your response would be "Don't waste your time. Case closed. Your hypothesis that there was another killer is unnecessary to explain the victim's death." Is that correct? What kind of X would suffice for you to let your detectives keep investigating?
Two examples that come to mind immediately:
-Similarly incriminating DNA from someone else in addition to the prime suspect.
-Information acquired from the prime suspect himself that points to accomplices.
Maybe. But remember that reality is consistent: If two people were at the scene committing the crime, why would there be vastly more evidence of one than the other?
Once you have the suspect, you can interrogate him to find out who he knows.
Come to think of it, why would a further (unknown) DNA sample even be enough? You've got an explanation for the crime. Surely there are far more likely priors than a double murder that would explain a second DNA sample, right? "Similarly incriminating" might not really be possible; much of what made the first sample so incriminating is that it matched a guy who you already had reason to suspect. Or did you mean that an unknown DNA sample would not suffice, only one that matched another suspect?
If the victim's body and surroundings contains DNA from two different individuals, that would suggest multiple attackers.
OK, fair enough. I think I understand your standard better now. But let's go back to the actual case. Here's a quote from that truejustice site's summary of the Micheli report:
At what point exactly do you part ways with Judge Micheli in that chain of reasoning? To me, his reasoning starts to sound stretched around the middle of the second paragraph. But it sounds like you might have a problem with this logic from the very beginning ...? Are you saying the evidence that the body had been moved is just not worth paying attention to?
If, starting from the premise of Guede's involvement, there are reasons to infer the involvement of someone else, then that sort of thing may very well be worth paying attention to. If the trail through Guede goes utterly cold, however, there comes a point where you just have to declare that Guede's actions + there's-something-we're-missing-about-that-other-"evidence" is a more parsimonious explanation of the data than Guede's actions + someone-like-Amanda-Knox-is-guilty.
In this situation, we should suspect that, if we bothered to investigate further, we would find that we were missing something. And sure enough, by golly, that's what often seems to happen.
OK, but why do you keep saying "if"? The judge is making an argument on your terms. He is trying not to privilege the hypothesis. He is starting from the premise of Guede's involvement, and he does find a reason to infer the involvement of someone else. He does not conclude that the trail goes utterly cold, but instead that it leads convincingly to Raffaele and Amanda.
Now, you may disagree with this argument, but I still haven't heard the substance of your disagreement. All you've done is gainsay it.
Don't get me wrong -- I think your original post was a very good explanation of some huge conceptual problems in the way the case against K&S has come together. If I came in believing they were guilty, you would have raised massive doubts in my mind. And that's no small accomplishment on your part. But it doesn't follow that no case against them remains. In order to convince me, or anyone else who's around the average of 35%, that we should lower our odds to your 1-10% range, I think you have to address the facts more directly. It's not enough to say "there comes a point when you just have to declare..." or that certain DNA evidence "doesn't count."
If you were a defense attorney and we were jurors, then you're right, you'd have your acquittal. But it's not a juror's job to distinguish between a 1% and a 35% probability of guilt. To make that case, I don't think you can just point out weaknesses here and there in the prosecution's argument -- you need to lay out the strongest version of the prosecution's case, even if you have to put it together for them, and then show step-by-step why it doesn't lead to a probability higher than 1-10%.
I don't blame you for not doing that, because as I've been saying, the primary sources aren't available to do it -- at least not in objective, quality English translations. I'm just not sure how you can get to such low odds without taking a more granular approach. You've suggested that the correct prior is the (very low) odds of someone like Amanda committing homicide, and others have suggested that the correct prior is the (very high) odds of a convicted defendant in a modern legal system being guilty -- but these are two ends of a spectrum, not binary alternatives. Finding the correct point on that spectrum is equivalent to assessing the strength of the prosecution's case.
No he isn't! He's starting from the premise that some investigator found the condition of Meredith's clothing and bloodstains to be unusual given the hypothesis of only one killer. As far as I can tell, he has failed to update properly on the lack of connection between Guede and anyone else who might be a suspect -- not to mention the lack of other evidence (e.g. DNA) that would indicate two or more killers.
I'm starting to suspect that we may just have a disagreement about how strong the anti-Knox evidence is. Yes, I agree it isn't literally zero. But that's not the point. The point is that it is utterly dwarfed by the other evidence. Exactly how strong of a dwarfing is this? Well, that's what seems to be the point of contention. I claim the net evidence of Knox's guilt yields a probability of no more than 0.1; you're uncomfortable going below 0.35. The only way to resolve this would be to do some sort of rigorous calculation of the inferential power of clothing-mechanics-analysis evidence -- something which I think would take us too far away from our main topics here.
I suppose I can console myself with the fact that it's good news for Amanda and Raffaele (and bad news for the prosecution) if what are probably the most intelligent and sophisticated discussions of their case on the whole Internet consist of vociferous arguments about whether the probability of their guilt should be 0.35, 0.1, or even lower.
That's setting the bar pretty high. What about witnesses who claim to have seen two men running from the scene, or heard multiple voices at the time of the murder?
those would be the witnesses that appeared months later and whose testimony is doubtful. The running feet and scream women's story cannot be true... a reenactment later showed them to be impossible and they were actually not sure if it was Halloween the night before when there would have been a lot more rowdy students around... And the eyewitness was also discredited.