As computer people, scientists etc you should all be aware of the dangers of untrained outsiders who believe that they can not only understand a complex field with a minimum of logic, rationality and commonsense but understand it better than the trained professionals in that field. As a practicing lawyer, this analysis really sounds to me like the analysis of an intelligent climate skeptic or an intelligent arts graduate sounding forth on AI.
Criminal justice systems are hard. They're right up their with the hardest things humanity has to deal with. They deal with ambiguous facts, necessarily inadequate rule systems and with intelligent human actors constantly trying to game the system - what's more the stakes are terrifyingly high, up to and including the life or death of the participants.
Sure, errors tragically occur. But we should really pause before pronouncing --- based on Bayes, blogs and journalists --- that the cursory judgement of a non-profesional who wasn't in the court is superior to the considered judgement of judge and jury made in full cognisance of the likely outcome of conviction after listening to all the evidence including first hand witness testimony.
And no, throwing in an offhand comment about how Occam's razor disproves God and therefore a little logic wielded with "courage and a certain kind of ruthlessness" enables superior thinkers to do better than the hidebound judge and jury (who were probably irrational Christians anyway) doesn't make the argument any more convincing.
The point of the post on lesswrong.com is that since it's agreed that Guede murdered the victim, you have completely explained away the murder. So now you must start over, from scratch, as if the victim were still alive, and ask how likely it is that Knox and Sollecito killed somebody based on the other evidence: that both lied repeatedly to the police, that there was only one good fingerprint of hers within her own apartment, that she uses bleach to clean her apartment, and so on. Taking that evidence by itself - without any dead or missing person, mind you - would you even accuse them of anything more serious than obstructing the police?
And yes, the priors - like the probability that a 20 year old female college student with no history of violence or mental disturbance of any kind would commit a murder - are certainly very relevant. That is one of the least violent categorizations of humans you can create. (In fact a more accurate prior would be even more miniscule: the probability that such a person would lead an elaborate team sex murder, an act so rare there is no term for it.)
(However, there is some other evidence that is hard to deal with in this model, like the crime lab saying that there is Knox DNA on the bloody bra clasp of the victim. That evidence seems hopelessly entangled with the murder. How can this model be used to reason about it?)
Anyway, all I know of the case is what other commenters here and the OP on lesswrong.com wrote, so my facts are suspect. I've just tried to explain what the OP is saying. Did I succeed?
Hmmm. So using this model, what would the OP say if there were Knox DNA under the victim's fingernails? If her roommate weren't a murder victim (more precisely, that we don't have any reason to think that there's been a murder, since that's what we're imagining to keep ourselves honest) these things aren't even suggestive of murder.
Here's another question: what a friend of the victim had testified that the victim had called sometime around the believed time of death and choked out, "Guede and Knox did it!"? If we're imagining that we don't know if there is a murder, then someone doing that isn't all that damnning, since hey, "it" could be any silly little thing? Yet it sounds pretty bad. Aha! Guede's name there associates something he did with Knox and it's agreed that he only did one thing of note in her apartment that day. OK, can we make a fact that would be more entangled? Yes: what if she said in that dying phone call, "Knox killed me"? Then aren't we saying that we only add the additional evidence of a roommate who isn't thought to be dead uttering that sentence? People say things like that all the time using "killed" as a metaphor for all sorts of innocent things. Yet it seems pretty damning. Can we really untangle the murder from the other evidence? I mean, the probability that the victim would at that moment use the word "killed" to mean something other than "stuck a knife in my throat" seems really low. And I think there is some similar entanglement with the other facts I've mentioned, too. The fact of the murder seems to make this technique faulty.
I think you did, and it's important to say that the Less Wrong OP has just made one of many different cases for why reasonable doubt exists. This approach was not used at the trial because the defense argued the version of the truth that Knox and Solecito tried to create.
>> The point of the post on lesswrong.com is that since it's agreed that Guede murdered the victim, you have completely explained away the murder. So now you must start over, from scratch, as if the victim were still alive, and ask how likely it is that Knox and Sollecito killed somebody based on the other evidence:
To me, I think this first line should read ".. since it's agreed that Guede was found guilty for his involvement in a murder that, due to extensive crime scene analysis, was clearly committed by multiple assailants". So I'm not sure why we must "start over, from scratch, as if the victim were still alive".
Furthermore, keep in mind that the evidence includes an initial written statement from Knox that implicates her in the murder.
You're right, if the crime was clearly committed by multiple assailants you need to find another killer. But the OP throws out the evidence that indicates multiple assailants.
False written statements are often signed in police stations.
e.g. A bunch of people debate "rationality" using a few "priors" based on flimsy evidence from the internet. Just because you can use Bayesian nomenclature to snow others, doesn't make it rational. This whole thing hinges on the priors, and the author has given very weak evidence for them.
Not quite. The author's point is that, at least according to released information and any unavoidable biases involved, the physical evidence strongly suggests Guede while hardly even slightly suggesting he had help.
More strongly the author is trying to argue that its an easy fallacy to get drawn into psychological motives because we're "great" at modeling other people's thoughts. In light of this implicitly perceived strength we tend to ignore the real likelihood of the evidences.
I don't know if it's something I agree about, however. I haven't personally reviewed the evidence and don't particularly put a lot of trust in any of the sources including the Less Wrong contributors.
The whole thing doesn't hinge on the priors: the priors should be rather easy to agree to (i.e., the observable rate at which the general population in Amanda Knox's relative bucket commit murders). The author's entire point is that all of the evidence presented against Ms Knox is irrelevant because the bar for it to be incriminating is way high given the evidence against Guede (and it doesn't meet that bar).
(a) It isn't zero-sum. Guede was convicted -- he is guilty, he did it. What we are determining is whether he acted alone.
(b) The priors are crap because they are weak, not because they are agreed upon. This is just a model, and it feels to me that the author is overfitting. The prior you mentioned as an example is so far away from the case at hand to be virtually useless.
So, you're right in that Guede being guilty doesn't imply that Knox isn't. But, the evidence against him affects how the evidence against Knox has to be handled: for example, the lack of physical evidence against Knox becomes a big deal when Guede leaves tons of it. Re: priors being far away: a prior is the probability of an event occurring before any evidence has been observed. Priors are always necessarily that far removed; those are the correct priors.
How exactly does the evidence against Guede affect the regarding of evidence against Knox? Assuming that everyone involved (if it was multiple parties) were equally involved equally competent and hiding their involvement and left equal evidence of their involvement is just that an "assumption". It's not necessarily true and is in fact highly unlikely to be true.
In the article actual physical evidence is dismissed because it's "small in quantity". The amount of blood is far less important than whose blood it is. You only need a little dna to prove contact. And you only need a little blood to prove prescence at the scene of an injury. It sounds to me like the jury was doing it's job and the article's author was just looking for a point to make.
But it isn't just that it's "small in quantity" that means that there is a problem. Amanda was a housemate, it would be incredible if her DNA couldn't be found in her housemate's room. The simplest explanation for Amanda's DNA being in the room is that she had previously spent some time there talking to Meredith.
To put this another way, have a think about where you would expect your own DNA to show up. At home, at your partner's appartment, at the office, at the cafe around the corner from the office, in your car, in your partner's car, at your best friend's home, your best friend's car. If a murder ever took place in one of these locations, how would you feel about the presence of your DNA being used to implicate you in the murder?
Fair enough, I could see relative evidence being used as evidence. :-) I think that negates my first point, you are correct.
But I strongly disagree on your understanding of the impact of priors. There are several ways to look at it, all equivalent. The easiest is the single prior: Did she do it? None of us are psychic, so that isn't helpful, 'cause we cant estimate it at all. We have to break that down into some of its components.
In fact, your argument (which I totally agree with, BTW) is based on a very specific prior: that in cases with this kind of relative imbalance of evidence, there is a high prob that one party is guilty and the other isn't (or some such, I may not be stating it as specifically as I should be).
All the evidence presented has it's own implied priors, and the right combination of evidence has it's own prior, etc. These all need to be combined with weights based on other priors. Since noone can compute all of these priors in the space of one hour, that makes it all a judgment call, not a rational decision.
And this is where I think the post's pontificating gets it terribly wrong. There's nothing rational at all about the process, other than you're attempting to blindly apply the Bayesian formula.
Ok. I agree with you that the post was confused and kind of incoherent at times. It was somewhere deep in it I ended up seeing what the author was talking about (taking his statements about physical evidence on faith for the sake of discussion). I'm definitely not saying that Ms Knox is innocent -- I don't know anything about the facts of that case, so we're only talking about the mathematical methodology here.
I do think your definition of a prior is a bit off though... like, the choices around priors wouldn't be whether to use "did she do it" vs the imbalances of evidence or such. We're trying to estimate the likelihood of "did she do it," so the question is what do we use to do that? The prior probability of "she did it" has to be the probability that some random member of the 20/female/american/student general public will commit the murder, because we don't have anything more than that (as you said). Then, we update our estimate by looking at the available evidence (such as, "was found standing by door holding bleach mop," which would the viewed in terms of the likelihood that that would be observed and the statement 'she did not do it' being true) to come to our posterior probability, which is the estimate informed by all of the available evidence.
As a side note, it's great when you can have a civil discussion on the internet. Maybe there's hope for HN after all.
Is that actually true, or are you assuming that is true?
The comments on the post relate further evidence that was missing from the analysis the poster provides. Why? Because the poster made a judgement call that it wasn't relevant. That alone makes this just so much blathering... sorry.
I think the evidence on the internet firmly establishes that there is a substantial probability that Knox and Solecito are guilty of lying about the events of the evening but that does not come to close to establishing that they committed murder.
There is a reasonable doubt.
What evidence is missing from the internet that makes them guilty? I guess we will find out in a few months when the judge and jury issue their written explanation.
Is anyone willing to state that they are certain Knox and Solecito are guilty of murder?
The media has truely outdone itself is mis-reporting this case, and sadly this article - surprisingly for one focused on rational thought - is no exception. There is a body of hard evidence, including forensic, such as (but not an exhaustive list):
- 4 mixed DNA samples of the Knox and Kercher. Bloody footprints that match. Blood on the tap put Knox at the crime scene that evening or in the early morning.
- A bloody bra in the washing machine
- Attempts to frame an innocent man for the crime.
- Repeated lying to authorities regarding her whereabouts. Still no fixed alibi.
- Knox and Kercher's house was cleaned with bleach.
- Sollecito's trainers were cleaned with bleach. A defect in the sole match a blood print on the scene.
The decision went through going through 19 expert judges and 6 lay judges. Conviction was unanimous.
Certainly, we can ignore some of the strange behaviour of the convicted (the phone calls, the lies, her past). But the case is much stronger than what this article makes out.
This is an English translation of the judge's summary:
edit: I left out a few other key points to the prosecutions case:
- Eye witness putting all 3 suspects together outside the cottage that night, and the night previous.
- Knox and Sollecito standing outside the crime scene when the police arrived with mop and bucket.
Rational thought takes anyone first to the conclusion that there was more than one killer. Guede left the scene soon after the murder (eye witness reports; evidence of cleaning, Guede's own statement). It is clear the crime scene was later returned to, and altered to cover evidence (faked break-in; footprints; removal of bra from the body). With all the additional evidence against Knox and Sollecito, physical and circumstantial, clear motive, and consitantly irrational behaviour, gives a very strong prosecution case.
My counter points:
1. No hard link shown between Guede and Knox/Sollecito
2. Lack of DNA evidence for Knox/Sollecito while plenty was present for Guede.
3. No evidence was shown that proved the houses were bleached after the murder
Regarding these points:
- Eye witness putting all 3 suspects together outside the cottage that night, and the night previous.
- Knox and Sollecito standing outside the crime scene when the police arrived with mop and bucket.
From what I read these were given by unreliable sources (drunk homeless man)
The police comes out pretty tainted in my opinion and it seems they were eager to find scapegoats early in the investigation.
Also, I don't find a clear motive for Knox/Sollecito. Not participating in a sex orgy?? It seems more probable that Guede raped and then murdered the victim. Also, in the initial testimony guede doesn't mention Knox or Sollecito being involved
Have you read the Judge's summary? It was very enlightening to me having read many mis-reports from media sources.
I will reply to your points in turn, but to be clear, I'm not specifically arguing for or against the conviction; I'm saying that this article grossly ignores many of the facts of the case.
It appears that large parts of the American media have painted this case as some loon prosecutor bent on convicting an American girl with flimsy flawed evidence and a fanciful idea of some Satanic sexy orgy gone wrong. There's been a lot of column inches devoted to discussing some of the things this article breaks down (the strange behaviour, the telephone calls). I can understand the media focusing on this (it's what the public find interesting, not the detailed crime scene details). But as this article is about digging for information and applying reasonable thought, it's disappointing they chose to ignore the large body of evidence presented at the trial that lead to their conviction.
To your counterpoints:
1. If I recall, there is good (not the drunk homeless man) eyewitness testimony of the 3 meeting prior to the night in question.
- There is fair amount of DNA evidence (see my parent post, and the links). Yes, some may appear potentially flawed (e.g. bra clasp). However the judge considered and admitted the evidence. Consider that a panel of legal experts considered and unanimously convicted based on this evidence.
- From truejustice.org regarding evidence of a cleanup: "The ONLY fingerprints of Amanda Knox in the entire house were found on a glass in the kitchen. Even in her own room there was not one print."
- Yes, the eyewitness isn't great. But again, the evidence was admitted by the judge for reasons given in the judge's summary.
- The police themselves were the eyewitnesses to them being at the crime scene with mop and bucket.
- Whether the police appear tainted is worth considering, but doesn't explain away a body of hard evidence that was admitted in court.
- Your theory that Guede was a lone-wolf rape and murderer was quickly dismissed by the judge, given the huge body of physical evidence suggesting other people involved (again, see my post and the report).
edit:
This is a great link that goes over some of your points (motive, DNA evidence etc:
Until I saw this thread on HN, I wasn't even aware the media were mis-reporting it.
I live near Leeds, where Ms Kercher was from, and the local media has been very detailed about the hard evidence, and painted the case in a way that has been very hard to misunderstand.
Now, admittedly the local media is likely biased as well, just in the opposite way to perhaps the American media has reported this case. But I wasn't aware there was this reporting of Ms Knox as being utterly innocent. If anything, it was clear from the reporting I've seen that she's a sociopath/psychopath trying to manipulate the court with her 'I'm soooo innocent' act.
They lied about the evening and may have been there before or during the murder, then panicked after Guede fled, so they cleaned up because they knew how guilty they were going to look. It does not mean that they are murderers.
Also note that Italian defendants are allowed to lie while they are being called as witnesses.
I don't think the quantitative probability analysis of the Less Wrong thread is that meaningful in the semantics of the analysis itself, but I think it shows that a reasonable person would have a qualitative reasonable doubt about the murder.
I think they were involved in it, but doesn't a murder conviction usually require proof that person X actually did the killing? Otherwise it's accessory to murder or something.
BTW, what a lot of people are saying in Italy is that she'll probably be let off on appeal.
I think they were involved in it, but doesn't a murder conviction usually require proof that person X actually did the killing?
That depends on your jurisdiction. I'm not familiar with Italy, but most U.S. states and the federal government do not require actually personally causing death as an element of the crime of murder. If you're a) engaged in the commission of a felony (sometimes, an enumerated list of "really bad felonies") and b) someone dies with cause proximate to the felony you're committing, you just earned yourself a murder charge, regardless of who pulled the trigger. This is called "felony murder". The most common use case is suspects A, B, and C attempt armed robbery on D, D dies in the attempt, and A, B, and C each admit to the robbery but not to the killing. Absent the felony murder theory, you might have to sift through confusing physical evidence and unreliable testimony of self-interested shady characters. With felony murder, your options are simpler: charge 'em all.
I think there is another reason for this practice:
Imagine that one criminal has good leadership skills and is able to manipulate another, weaker criminal. He could then convince the weaker criminal to pull the trigger and escape consequences himself, and then do it again and again with different weak criminals. By sharing the blame the system compels the leader, whoever he is, to restrain the underlings rather than push them to the extreme.
"doesn't a murder conviction usually require proof that person X actually did the killing?"
No. Murder, and Rape, commonly promote the crime from conspiracy/accomplice to the full crime (murder, rape). Presumably because the stakes are so high for these types of crime, and traditional sentencing for 'accomplice to...' or 'conspiracy to commit' cases is so much lower, that people could game the justice system somewhat.
For example, if you hire a hitman to take out your wife/husband, you WILL be tried as the murderer, as will the hitman. If you conspire with a lover to kill someone, you will be tried as a murderer, no matter who is holding the murder weapon.
Perhaps it's different in America, although I'm sure there have been cases I've heard about in the US that fall under the same rules as those above, and where the conspirators were tried as murderers themselves.
I'm not talking about conspiracy to kill someone, but about a situation like this: say 4 or 5 guys are picking on someone outside a bar, hitting him and so on. One of them pulls out a knife and kills the guy. No one talks. It's obvious that someone in the group killed him (let's say there was a camera or something that recorded things, but not well enough to pick out the killer). Do they all get a murder charge? I honestly don't know. It's evident they're all guilty of something, but only one of them is a murderer.
You can say everything about Italy, but not that our juridical system is broken in this way: it's much better than USA system in preventing innocent people from going to jail. There are three levels of giudice, and every level can only give you a penalty <= the previous one, and if there aren't many strong proofs it's likely you will not get marked as guilty. Also there is no such "popular jury", it's all made by professional people. We are mostly an irrelevant nation lately, but 2500 years of history of our juridical system are still in action.
Criminal justice systems are hard. They're right up their with the hardest things humanity has to deal with. They deal with ambiguous facts, necessarily inadequate rule systems and with intelligent human actors constantly trying to game the system - what's more the stakes are terrifyingly high, up to and including the life or death of the participants.
Sure, errors tragically occur. But we should really pause before pronouncing --- based on Bayes, blogs and journalists --- that the cursory judgement of a non-profesional who wasn't in the court is superior to the considered judgement of judge and jury made in full cognisance of the likely outcome of conviction after listening to all the evidence including first hand witness testimony.
And no, throwing in an offhand comment about how Occam's razor disproves God and therefore a little logic wielded with "courage and a certain kind of ruthlessness" enables superior thinkers to do better than the hidebound judge and jury (who were probably irrational Christians anyway) doesn't make the argument any more convincing.