Continuation Part 14: Amanda Knox/Raffaele Sollecito

[Grinder]

Someone should ask Ms. Schneps what Bayesian theory says about the chances that the ISC ruling was the result of a conspiracy of masons.

 

[LondonJohn]

Wouldn't you mathematically model whether a witness is believed in exactly the same way as you'd explain in language whether they're believed? Same for Guede breaking in: if you can explain your reasons in language, as you have to do in a motivation report, then those same reasons should be expressible in figures as well.

Again, it's as if you're saying there's this very complicated process that can't be communicated in any way. Could be, but then the problem is with communication in general, not with Bayesian reasoning in particular.

AAAAAAGHHHHHHHHHHHH!!!!!

I repeat: there is absolutely no problem with Bayesian reasoning in theory.

But there is a problem in trying to use it to model incredibly complex (though often superficially simple-looking) areas of human reasoning.

I can absolutely guarantee you that it would be astonishingly complicated to construct a mathematical model to replicate the process that drives whether one human being believes that what another human being just said is reliable and true.

For example, take Capezzali on the stand in the Massei trial in this case. How does a real-world human assess whether her testimony was reliable and true? He listens to her words; listens to (and visually observes) her demeanour; looks for any clues in the way she responds to certain questions in certain ways. Does she become defensive at times? If so, why? Does she sound (and look) like a "reliable witness". Does she pause at odd times. Are there any slight contradictions in her testimony that might not be explainable? Does she appear to be the sort of person who might honestly believe that she heard those screams, even if in fact she heard nothing that night? (and the answer to just that last question involves a stupendous array of factors, including an assessment of her mental health, of her recent life history, of the general phenomenon of people hearing imaginary noises, of how that phenomenon might increase with age and/or reduced mental health).

I'm somewhat amazed, frankly, that you cannot see that while it's clearly possible theoretically to construct a model that takes all these factors into account and renders them somehow in accurate numerical form (and that's just a small subset of all the factors that would feed into a human reasoning on Capezzali's testimony), in practice it is ASTONISHINGLY DIFFICULT TO DO. In fact, it might be only fuzzy logic, AI and quantum computing that would ever even stand a chance of being able to model these sorts of factors.

 

[LondonJohn]

EXACTLY THE POINT!! That reasoning is crap. If he'd had to give logical reasons then maybe he'd have reached the right verdict back then.

Whether it's crap or not is entirely moot. It's unchallengeable. And it's massively difficult to model what the "correct" judgement on that issue might be (which is exactly what you'd want to use a mathematical model for, after all).

 

[katy_did]

AAAAAAGHHHHHHHHHHHH!!!!!

I repeat: there is absolutely no problem with Bayesian reasoning in theory.

But there is a problem in trying to use it to model incredibly complex (though often superficially simple-looking) areas of human reasoning.

I can absolutely guarantee you that it would be astonishingly complicated to construct a mathematical model to replicate the process that drives whether one human being believes that what another human being just said is reliable and true.

For example, take Capezzali on the stand in the Massei trial in this case. How does a real-world human assess whether her testimony was reliable and true? He listens to her words; listens to (and visually observes) her demeanour; looks for any clues in the way she responds to certain questions in certain ways. Does she become defensive at times? If so, why? Does she sound (and look) like a "reliable witness". Does she pause at odd times. Are there any slight contradictions in her testimony that might not be explainable? Does she appear to be the sort of person who might honestly believe that she heard those screams, even if in fact she heard nothing that night? (and the answer to just that last question involves a stupendous array of factors, including an assessment of her mental health, of her recent life history, of the general phenomenon of people hearing imaginary noises, of how that phenomenon might increase with age and/or reduced mental health).

I'm somewhat amazed, frankly, that you cannot see that while it's clearly possible theoretically to construct a model that takes all these factors into account and renders them somehow in accurate numerical form (and that's just a small subset of all the factors that would feed into a human reasoning on Capezzali's testimony), in practice it is ASTONISHINGLY DIFFICULT TO DO. In fact, it might be only fuzzy logic, AI and quantum computing that would ever even stand a chance of being able to model these sorts of factors.

Didn't you read where I said that the problem is obviously with communicating a person's reasoning in general, not with Bayes in particular? If you can explain your conclusion in language adequately then you can do it in figures as well.

In the context of a court case I also think it's a bit different, in that intuition should matter less than things which are more concrete (length of time it took for the person to come forward etc). Things which lend themselves better to a mathematical model. You're talking in an abstract, philosophical sense - why one person believes another - not in terms of proof BARD.

 

[katy_did]

Whether it's crap or not is entirely moot. It's unchallengeable. And it's massively difficult to model what the "correct" judgement on that issue might be (which is exactly what you'd want to use a mathematical model for, after all).

No it isn't, because this is a court case and crap reasoning gets overturned. Like I said, you're talking in the abstract, not in the context of proving something beyond reasonable doubt. You don't have to include all of human behaviour in a mathematical model of why you believe or disbelieve a witness, you include the concrete factors which led you to do so, which are usually much less woolly than "he seemed like a nice chap". That's why there are studies and guidelines on what factors make a witness reliable/unreliable.

 

[Dougm]

Remember when the Italian Supreme Court acquitted Amanda Knox and Raffaele Sollecito?

I recall, in my college days, leaving the stadium after a major college football rivalry game. My school's team had just finished beating our top rival, rather handily at that. As we left the stadium, some of the more obnoxious fans of the other team felt it necessary to run up and yell things like, "you cheated", and "you paid off the refs", etc. To which we simply replied, "SCOREBOARD"!

I enjoyed your post as a similar sentiment.

 

[anglolawyer]

So if the judge said of Curatolo, "I just believed him", you'd be OK with that?

No. Course not! I'm not OK with the reasoned version either, though. I was just saying that in our system, which we use to find people guilty all the time, sometimes the verdict rests on nothing more than not believing someone (e.g. a shoplifter who claims they 'forgot' to pay) and that disbelief may be a question of 'feel'. Like a good chess player can make a move based on feel or a typist whose fingers 'know' where the keys are on the keyboard even though she herself does not.

 

[HumanityBlues]

I recall, in my college days, leaving the stadium after a major college football rivalry game. My school's team had just finished beating our top rival, rather handily at that. As we left the stadium, some of the more obnoxious fans of the other team felt it necessary to run up and yell things like, "you cheated", and "you paid off the refs", etc. To which we simply replied, "SCOREBOARD"!

I enjoyed your post as a similar sentiment.

I mean, what else can a guilter say? Amanda Knox and Raffaele Sollecito were acquitted, but we should still feel bad because....something about a holocaust denier? Really?

PMF got their a@$es handed to them on a silver platter. It's not so much that they didn't get the result they wanted. It's the absolute arrogance they had that there was no chance whatsoever of acquittal--rubbing it in peoples' faces, bragging and gloating before the fact. I'm sure it was a gut punch, but it might have hurt a little less without all the before-the-fact cheering.

And maybe, just maybe, if they were all so wrong about the certain outcome of the case, they should reexamine the way they have been looking at things the whole time (they won't, because it really sucks to be wrong, especially with the way they've conducted themselves). At least as far as the judicial result goes, they can't walk away from that one. They were wrong. They said the Supreme Court would convict them over and over ("justice is coming"), and they were wrong. So maybe they're wrong about everything else too.

 

[LondonJohn]

No it isn't, because this is a court case and crap reasoning gets overturned. Like I said, you're talking in the abstract, not in the context of proving something beyond reasonable doubt. You don't have to include all of human behaviour in a mathematical model of why you believe or disbelieve a witness, you include the concrete factors which led you to do so, which are usually much less woolly than "he seemed like a nice chap". That's why there are studies and guidelines on what factors make a witness reliable/unreliable.

Woah, I disagree. And even where there are hard qualitative factors that can be listed for use in an assessment, how do you attach quantitative values to them for the purposes of mathematical modelling? One might be: "is the witness/suspect dressed as you would expect, and as would be appropriate?" And even that simple evaluation, while it's obviously possible to model in principle, would be very hard to convert into justifiable and defendable numbers for the purposes of inputting to a model.

In short, I repeat once more that it's possible to model pretty much everything - including abstract human behaviours and abstract human reasoning - using numbers and logic. But it quickly becomes incredibly, incredibly difficult to do so.

Remember, as an example (though a somewhat unrelated example), I pointed out that you and I can walk, run and jump without giving it a single thought, and we can maintain perfect balance throughout. Well, in reality there are a gigantic number of different brain functions going on in order to enable us to do that, requiring many different inputs (vision, balance, spatial awareness, feeling of things like wind, an incredibly complex understanding of the surface beneath our feet - its coefficient of friction, its stickiness, its resistance to force, its stability. And then it requires an amazingly complex series of outputs, including minutely-adjustable levels of muscle control across a large number of different muscles, and a whole number of feedback control loops.

And we do all of this without even thinking about it (apart from generating the initial desire to run, jump etc). Yet even this - a seemingly "simple" set of logical issues, without anything at all related to abstract thought etc - has proved incredibly difficult for mathematicians, computer scientists and engineers to replicate accurately through a model (i.e. in this case a robot that can run and jump just like a human).

So when it comes to more abstract, obtuse reasoning that's not so much based on logic per se but on a very complex fabric of life experiences, that's an area that is exponentially more difficult to model than "simply" a human-shaped form that can run and jump accurately.

Therefore, my ongoing belief is that it's probably beyond the current wit (and tools) in practice of mathematicians or computer scientists to construct mathematical models which would accurately and reliably take a whole set of inputs from something like a criminal trial (i.e. all the evidence and testimony), convert all those inputs into numeric forms, perform mathematical operations on those inputs, and output something like a "guilty or not-guilty" verdict. I think it will probably take quantum computing, AI and fuzzy logic to make that sort of thing even possible. Therefore, I conclude that as of now, it's impossible in practice to build a proper model of a whole trial.

I do though (as I've said) think that its eminently possible - and of some (though limited) use - to construct mathematical models of certain small discrete elements within a trial.

 

[LondonJohn]

I mean, what else can a guilter say? Amanda Knox and Raffaele Sollecito were acquitted, but we should still feel bad because....something about a holocaust denier? Really?

PMF got their a@$es handed to them on a silver platter. It's not so much that they didn't get the result they wanted. It's the absolute arrogance they had that there was no chance whatsoever of acquittal--rubbing it in peoples' faces, bragging and gloating before the fact. I'm sure it was a gut punch, but it might have hurt a little less without all the before-the-fact cheering.

And maybe, just maybe, if they were all so wrong about the certain outcome of the case, they should reexamine the way they have been looking at things the whole time (they won't, because it really sucks to be wrong, especially with the way they've conducted themselves). At least as far as the judicial result goes, they can't walk away from that one. They were wrong. They said the Supreme Court would convict them over and over ("justice is coming"), and they were wrong. So maybe they're wrong about everything else too.

Yes. I've made this same point before. I think it's very revealing that pretty much none of the pro-guilt commentators (with perhaps one or two exceptions) has shown an ounce of humility, self-reflection, or any kind of reference that they might have been even a tiny bit wrong. They pretty much all seem obsessed with reinforcing the mantra among themselves that the SC made a horrible mistake - whether through massive corruption or massive incompetence - and that Knox and Sollecito really did murder Meredith Kercher but they "got away with it".

And of course this line of thinking gives them the cheeky bonus of being able to heap even more vitriol onto Knox and Sollecito, by imagining that they are now "flaunting their amazing luck in getting acquitted", and stating how incredibly disgusting it is for the two "murderers" to be acting in such a way (and what an insult to poor Meredith's family, etc, etc.....).

 

[acbytesla]

Woah, I disagree. And even where there are hard qualitative factors that can be listed for use in an assessment, how do you attach quantitative values to them for the purposes of mathematical modelling? One might be: "is the witness/suspect dressed as you would expect, and as would be appropriate?" And even that simple evaluation, while it's obviously possible to model in principle, would be very hard to convert into justifiable and defendable numbers for the purposes of inputting to a model.

In short, I repeat once more that it's possible to model pretty much everything - including abstract human behaviours and abstract human reasoning - using numbers and logic. But it quickly becomes incredibly, incredibly difficult to do so.

Remember, as an example (though a somewhat unrelated example), I pointed out that you and I can walk, run and jump without giving it a single thought, and we can maintain perfect balance throughout. Well, in reality there are a gigantic number of different brain functions going on in order to enable us to do that, requiring many different inputs (vision, balance, spatial awareness, feeling of things like wind, an incredibly complex understanding of the surface beneath our feet - its coefficient of friction, its stickiness, its resistance to force, its stability. And then it requires an amazingly complex series of outputs, including minutely-adjustable levels of muscle control across a large number of different muscles, and a whole number of feedback control loops.

And we do all of this without even thinking about it (apart from generating the initial desire to run, jump etc). Yet even this - a seemingly "simple" set of logical issues, without anything at all related to abstract thought etc - has proved incredibly difficult for mathematicians, computer scientists and engineers to replicate accurately through a model (i.e. in this case a robot that can run and jump just like a human).

So when it comes to more abstract, obtuse reasoning that's not so much based on logic per se but on a very complex fabric of life experiences, that's an area that is exponentially more difficult to model than "simply" a human-shaped form that can run and jump accurately.

Therefore, my ongoing belief is that it's probably beyond the current wit (and tools) in practice of mathematicians or computer scientists to construct mathematical models which would accurately and reliably take a whole set of inputs from something like a criminal trial (i.e. all the evidence and testimony), convert all those inputs into numeric forms, perform mathematical operations on those inputs, and output something like a "guilty or not-guilty" verdict. I think it will probably take quantum computing, AI and fuzzy logic to make that sort of thing even possible. Therefore, I conclude that as of now, it's impossible in practice to build a proper model of a whole trial.

I do though (as I've said) think that its eminently possible - and of some (though limited) use - to construct mathematical models of certain small discrete elements within a trial.

With all due respect LJ, I see this argument similar to "the anything is possible" argument. Both may be theoretically possible but don't represent an actual intellectual concept.

 

[Skeptic Ginger]

Just out of curiosity, how many people in this thread who said they trusted the court's initial guilty verdict because the court heard all the evidence and we lowly web surfers couldn't make a sound judgement now admit the woman is not guilty given a higher court has ruled in Knox's favor?

 

[The Big Dog]

Cart wheel girl got off, did she?

extradition would have been messy.

Can Anyone spare a 25k foot view of why the top court in Italy did what they did?

 

[Bill Williams]

Cart wheel girl got off, did she?

extradition would have been messy.

Can Anyone spare a 25k foot view of why the top court in Italy did what they did?

They did it because there was no evidence to support a conviction. They said that Raffaele and Amanda were innocent, not just "not guilty". Did not commit the act (the murder). Or on the ancillary charges, that the crime did not exist.

 

[Ammonitida]

They did it because there was no evidence to support a conviction. They said that Raffaele and Amanda were innocent, not just "not guilty". Did not commit the act (the murder). Or on the ancillary charges, that the crime did not exist.

Over at PMF, they're saying that the pair were acquitted under paragraph 2, which means insufficient evidence (equivalent to reasonable doubt).

 

[Strozzi]

And apparently after all that has gone on, they are still friends. One could not blame him for hating her.

Sollecito could hate the police, prosecution, and fraudulent judges, but I can't see any justification for him to hate Knox. Hate her for being a vulnerable victim to tough, deceitful cops in the midnight interrogation to obtain something condemnatory to arrest her? The police did essentially the same thing to Raffaele that night, only they began it with him a half-hour before they started in on Knox. Sollecito allowed himself to speak while confused. He signed a statement that they wrote which he said was a mash-up. The police played him too and used some of his confused statements on Knox to claim she went out the evening of the murder, when the fact is they both were together in his flat the entire evening and night.

I hope they remain friends. They have been through a terrible ordeal - one that was man-made by corrupt police, prosecutors, and judges who don't know the meaning of personal integrity.

Tell me one police player, prosecutor, or judge other than those on Hellman's court who themselves wouldn't have turned and lied to save their own skin for a foreign girl they knew only a week before this monstrosity engulfed them. You think Mignini or Stefanoni or Napoleoni or Giobi or Batistelli would have stood firm with the truth and suffered the unjust consequences? They would have lied to save their own skins to separate themselves from the co-victim, abandoning her to the Perugia wolves.

In later years, I hope Sollecito and Knox can look back at what happened to them in their twenties and realize that it was an evil that they survived. Much like others who in their 20's survive a terrible accident or near-fatal illness or war, to live on.

 

[Desert Fox]

Over at PMF, they're saying that the pair were acquitted under paragraph 2, which means insufficient evidence (equivalent to reasonable doubt).

If that gets them to lay off, great

 

[The Big Dog]

They did it because there was no evidence to support a conviction. They said that Raffaele and Amanda were innocent, not just "not guilty". Did not commit the act (the murder). Or on the ancillary charges, that the crime did not exist.

They said they were innocent? Is that typical in the Italian system? (Typical in the sense of something they do, as opposed to something they do often as I can't imagine it happens a lot)

 

[Desert Fox]

Sollecito could hate the police, prosecution, and fraudulent judges, but I can't see any justification for him to hate Knox. Hate her for being a vulnerable victim to tough, deceitful cops in the midnight interrogation to obtain something condemnatory to arrest her? The police did essentially the same thing with Raffaele that night, only they began it with him a half-hour before they started in on Knox. Sollecito allowed himself to speak while confused. He signed a statement that they wrote which he said was a mash-up. The police played him too and used some of his confused statements on Knox to claim she went out the evening of the murder, when the fact is they both were together in his flat the entire evening and night.

I hope they remain friends. They have been through a terrible ordeal - one that was man-made by corrupt police, prosecutors, and judges who don't know the meaning of personal integrity.

Tell me one police player, prosecutor, or judge other than those on Hellman's court who themselves wouldn't have turned and lied to save their own skin for a foreign girl they knew only a week before this monstrosity engulfed them. You think Mignini or Stefanoni or Napoleoni or Giobi or Batistelli would have stood firm with the truth and suffered the unjust consequences? They would have lied to save their own skins to separate themselves from the co-victim, abandoning her to the Perugia wolves.

In later years, I hope Sollecito and Knox can look back at what happened to them in their twenties and realize that it was an evil that they survived. Much like others who in their 20's survive a terrible accident or near-fatal illness or war, to live on.

The only thing I don't know about is if Amanda really should visit Italy.

 

[Bill Williams]

They did it because there was no evidence to support a conviction. They said that Raffaele and Amanda were innocent, not just "not guilty". Did not commit the act (the murder). Or on the ancillary charges, that the crime did not exist.

Over at PMF, they're saying that the pair were acquitted under paragraph 2, which means insufficient evidence (equivalent to reasonable doubt).

We will know for sure when the motivations report comes out. In the meantime most news outlets describe it as a finding of innocent, like the Hellmann decision.