Continuation Part 14: Amanda Knox/Raffaele Sollecito

[LondonJohn]

She was on IIP for awhile. She ran into a couple of scientists who work in PCR labs, who understand the problem of contamination in any experiments that involve microscopic quantities of biological material.

They tried to clue her in, and she flatly rejected their expertise.

I also called her on an egregious error in her book, where she conflated the account Raffaele gave to police with a spurious account from a UK tabloid. I showed her court records that proved what she had written was false. She simply insisted she was right without attempting to explain the discrepancy.

She gives the appearance of being rational and fair-minded, but she's not. I doubt she will ever change her thinking one iota.

Thoughtful has now made the same mistake a number of different times. She's isolated one specific problem that can be addressed with a mathematical analysis, and then she's applied that analysis to it. But (and in many ways just as with Balding's work on the DNA) she's abjectly failed to understand that in order to answer properly the question she's looking at, she needs to consider many more elements than the small number she puts into her analysis. It's a problem that's not at all uncommon when pure mathematicians attempt to analyse real-world situations. In fact, it's probably the main reason why there's a whole different discipline of applied mathematics.

Thoughtful's inability to even realise that her model is significantly imperfect (and that therefore its outputs are inherently wrong and unreliable) reminds me of the famous Monty Hall Problem, when even Nobel Prize-winning mathematicians (and Nobel physicists) embarrassed themselves by getting was was (even more embarrassingly) a pure maths problem utterly wrong:

http://en.wikipedia.org/wiki/Monty_Hall_problem

 

[Kestrel]

The LessWrong Blog has had some interesting discussions of Baysian analysis of the Knox case. Including this summary of predictions after the first acquittal was announced.

 

[LondonJohn]

That's really nice to know. Of course, it wouldn't matter if she was a heinous bitch, she'd still be innocent. But it's nice to know she isn't.

You've made an important point. I think it's incredibly instructive and revealing that a central tenet of pro-guilt beliefs is that they need Knox and Sollecito to have been unpleasant (to the point of being demonically evil) people in pretty much every facet of their lives and personalities. While at thew same time in their minds they need the victim, Meredith Kercher, to have been a flawless diamond, who was destined to make a huge mark on the world with her brilliance and astonishing character.

In fact, this issue has a very strong personal resonance for me. The first forum I discovered which discussed the case (I thought at that time that it was the only place the case was being discussed, in fact) was PMF. And aside from the fact that I wasn't getting anything like satisfactory answers to some of the questions I had (e.g. regarding the 5/6 Nov interrogation and the "confession/accusation"), I also quickly started to get red flags going up over this "Knox/Sollecito evil, Meredith amazing" mantra that seemed to be so deeply necessary to so many posters there.

And so I made a post which said, in effect: would it make one jot of difference to a rational, objective assessment of the case if Meredith Kercher had been (to go to extremes to illustrate the point) a crack whore? Would that have made her death any less or more of a tragedy, and would it have made it any more or less likely that Knox and Sollecito had participated in her death? I wanted to try to find out why it seemed so very important to them to construct this (frankly a little creepy) avatar of a perfect Meredith Kercher - it seemed to me that they needed this in order to build their own internal narrative and justification. Especially when what we did know of Meredith Kercher was that she was in many ways a very average British red brick university student, one who clearly had some intelligence and who clearly had many endearing, kind and inspiring qualities - but one who also had less desirable facets, including bitching about her housemate and (apparently) getting incredibly drunk (to the point of being barely able to walk) on regular occasions.

Sure enough, huge umbrage was taken at my comment, even though I'd clearly signposted that it was a deliberate exaggeration in order to illustrate a point. And it really was at that point that I started to understand in earnest that PMF was largely populated by kooks and damaged weirdos with ulterior motives for needing to campaign against Knox and Sollecito. It was a big turning point for me.

 

[LondonJohn]

The LessWrong Blog has had some interesting discussions of Baysian analysis of the Knox case. Including this summary of predictions after the first acquittal was announced.

I ought to make it clear perhaps that I absolutely believe it's possible - in theory - to construct a Bayesian model in a case such as this. I fully agree that Bayesian analysis can properly model the decision processes.

But..........

I think it's near-impossible to construct a Bayesian model that is sufficiently detailed (and complex) to act as an accurate representation of the full set of issues driving human reasoning in a case such as this one.

I also think that certain discrete elements within this case can usefully be approached with simple (or even slightly complex) Bayesian models, and that the outputs can be useful within certain limitations. For example, I don't doubt that Bayesian maths were very likely used by Balding in his work on assessing the probability of Sollecito's DNA being physically present on the bra clasp, given the set of data he was handed to work with. And Bayes could indeed work in answering that narrow question (is Sollecito's DNA physically present on the bra clasp?). Note though that Bayes cannot answer the question that's actually the important one in this area: what's the probability, as shown by the evidence, that Sollecito had primary contact with that bra clasp, as opposed to the DNA getting there through contamination (or worse)?

But once you start claiming (or trying) to use Bayes to calculate some sort of overall probability of guilt - as Thoughtful was so crassly and wrongly doing in her example from today - then I believe you're in deep deep water. Certainly if you're trying to output overall probabilities of guilt using just one Bayesian variable to move the model (as Thoughtful was), you're wrong to the point of being laughably wrong - if you're actually trying to produce a meaningful answer about the real-world probability of guilt. What's more, it's intellectually dishonest to claim (as Thoughtful did) that the complex question of the probability of guilt can be treated in such an over-simplified (and, frankly, cavalier) fashion.

 

[komponisto]

Do mathematicians do peer review?

Yes.

 

[Ammonitida]

Or far less than 10. Just one egregious example:

In Ireland a man named Eamonn Lillis bashed his wife's brains in with a brick on the patio of their lovely home in Howth in late 2008. He lied to the police and tried to pretend that an intruder had done it.

Incredibly, he was convicted of manslaughter instead of murder, even though there was plenty of evidence to show that he planned the crime.

He has served 5 years, and now he will be released any day to take possession of the 1.3 EUR estate he inherited from the wife he murdered.

http://www.independent.ie/irish-new...thin-a-month-with-13m-in-pocket-31110254.html

His wife's family and his own daughter have campaigned to change the law in Ireland so a person who is convicted of killing another person cannot automatically inherit the victim's property. The law was recently changed, but it only affects future criminals. Not Eamonn Lillis.

Actually, when it comes to spousal murder, it's usually the woman who is treated more leniently, even when the woman is not alleging self-defense (and judging by the recent high profile cases where women alleged self-defense to beat a spousal murder conviction, it appears that many if not the majority of these self-defense claims may not be true -- see Jodi Arias, Marissa Devault, Ana Trujillo, Tracy Grissom, Susan Polk, all of whom claimed self-defense but were convicted of murder because their self-defense claims were determined to be lies).

http://www.bjs.gov/content/pub/ascii/SPMUREX.TXT

"Wives received shorter prison sentences than husbands (a 10-year
difference, on average) even when the comparison is restricted to
defendants who were alike in terms of whether or not they were
provoked--"

A previous poster cited the Jodi Arias case as an example of where the murder victim is vilified by anti-victim hate groups. This is rather common in cases where women murder their spouses IME. Another example would be Hunter Grissom, who was stalked and gunned down by his ex at his workplace.

https://www.facebook.com/JusticeForHunterGrissom

Also check out the case of Deidra Lane;

http://murderpedia.org/female.L/l/lane-deidra.htm

Sentenced to 8 years, but served only 5 for a murder that was premeditated. The woman had a violent criminal history.

 

[Grinder]

What would the Bayesian theory say about the probability that a homeless heroin addict would be a witness in three criminal cases?

What would it be for one cub reporter to find 3 of the "super" witnesses as opposed to the police finding none on their own?

What would it be for no chyme moving to the duodenum before the murder at 9, 9:30 or 9:45?

What would the probability be for not one video capture on the streets between Raf's and the cottage? or that no "normal" person saw them running around bloody with a large knife?

What would the probability that no blood was found on the kids, their clothes, Raf's flat or the knife but a microscopic piece of dna was found?

What is the probability that Amanda told the police what they knew to be true on her own, when in fact it wasn't true?

What is the probability that the only DNA found of Raf would be on a piece collected 45 days later and none on all the other items?

Come on thoughtful give it a whirl.

 

[Bill Williams]

Can anyone actually explain her argument for guilt?

It seems to be that the DNA evidence is dodgy, but that there's all the other evidence.

If someone can point to a better analysis, I'd like to see it.

 

[anglolawyer]

What would the Bayesian theory say about the probability that a homeless heroin addict would be a witness in three criminal cases?

What would it be for one cub reporter to find 3 of the "super" witnesses as opposed to the police finding none on their own?

What would it be for no chyme moving to the duodenum before the murder at 9, 9:30 or 9:45?

What would the probability be for not one video capture on the streets between Raf's and the cottage? or that no "normal" person saw them running around bloody with a large knife?

What would the probability that no blood was found on the kids, their clothes, Raf's flat or the knife but a microscopic piece of dna was found?

What is the probability that Amanda told the police what they knew to be true on her own, when in fact it wasn't true?

What is the probability that the only DNA found of Raf would be on a piece collected 45 days later and none on all the other items?

Come on thoughtful give it a whirl.

maybe this Bayesian **** isn't so bad after all.

 

[Kauffer]

You've made an important point. I think it's incredibly instructive and revealing that a central tenet of pro-guilt beliefs is that they need Knox and Sollecito to have been unpleasant (to the point of being demonically evil) people in pretty much every facet of their lives and personalities. While at thew same time in their minds they need the victim, Meredith Kercher, to have been a flawless diamond, who was destined to make a huge mark on the world with her brilliance and astonishing character.

In fact, this issue has a very strong personal resonance for me. The first forum I discovered which discussed the case (I thought at that time that it was the only place the case was being discussed, in fact) was PMF. And aside from the fact that I wasn't getting anything like satisfactory answers to some of the questions I had (e.g. regarding the 5/6 Nov interrogation and the "confession/accusation"), I also quickly started to get red flags going up over this "Knox/Sollecito evil, Meredith amazing" mantra that seemed to be so deeply necessary to so many posters there.

And so I made a post which said, in effect: would it make one jot of difference to a rational, objective assessment of the case if Meredith Kercher had been (to go to extremes to illustrate the point) a crack whore? Would that have made her death any less or more of a tragedy, and would it have made it any more or less likely that Knox and Sollecito had participated in her death? I wanted to try to find out why it seemed so very important to them to construct this (frankly a little creepy) avatar of a perfect Meredith Kercher - it seemed to me that they needed this in order to build their own internal narrative and justification. Especially when what we did know of Meredith Kercher was that she was in many ways a very average British red brick university student, one who clearly had some intelligence and who clearly had many endearing, kind and inspiring qualities - but one who also had less desirable facets, including bitching about her housemate and (apparently) getting incredibly drunk (to the point of being barely able to walk) on regular occasions.

Sure enough, huge umbrage was taken at my comment, even though I'd clearly signposted that it was a deliberate exaggeration in order to illustrate a point. And it really was at that point that I started to understand in earnest that PMF was largely populated by kooks and damaged weirdos with ulterior motives for needing to campaign against Knox and Sollecito. It was a big turning point for me.

Yes to all this. But a question that remains - and really one central to the garnering of support amongst the public in miscarriage of justice cases - is, if, using alex's colourful descriptor, Amanda Knox had in fact turned out to be a 'heinous bitch', but still as equally clearly innocent, would so many people have 'gone to bat for her' and in such a variety of different ways?

So was it enough, actually, that she was so obviously and demonstrably innocent? Or are the pro innocent supporters in a sense really a mirror image of the pro guilt supporters? Did we need to establish that she is good in order, actively to support her as much as they needed to conceive her as bad in order to attack her?

 

[katy_did]

I ought to make it clear perhaps that I absolutely believe it's possible - in theory - to construct a Bayesian model in a case such as this. I fully agree that Bayesian analysis can properly model the decision processes.

But..........

I think it's near-impossible to construct a Bayesian model that is sufficiently detailed (and complex) to act as an accurate representation of the full set of issues driving human reasoning in a case such as this one.

Given that these factors influence our thinking anyway, isn't it a good thing to bring them out into the open, determine whether or not they're valid and how much weight they should be given - i.e. create a Bayesian model?

I don't really understand why you're saying it's unrealistic to do this, given that surely this is what any jury discussing a case actually does (or should do) even if they don't assign probabilities to each factor. Isn't creating a Bayesian model just doing the exact same thing in a different form? Saying it's not practical suggests there's some mysterious factor influencing human thinking that can't be put into words (or figures).

I do agree that Leila's approach to this is flawed because she leaves out any inconvenient factors, but obviously that doesn't mean the whole approach is flawed.

 

[Charlie Wilkes]

Can anyone actually explain her argument for guilt?

With regard to the DNA evidence, she insists that contamination is not a realistic possibility, and she seems to think she can prove this with math.

When she advanced this line of thinking on IIP, she ran up against two scientists who have hands-on experience with PCR analysis of DNA samples. They told her contamination happens all the time. The knife is almost certainly residual contamination from the lab, and the bra clasp was contaminated before it ever got to the lab, so C&V got it right: both these test results are junk.

She didn't want to hear it. The more she argued with them, the more clear it became that she doesn't have even a layman's knowledge of PCR analysis. Hence she does not grasp the need for precision. She regarded Stefanoni's sloppy work as the maverick path of an innovator who is unfettered by conventional thinking.

I can't remember when this was, but I think it was probably 2012, and you might be able to find it in the archives over there. It is worth reading if you are interested.

 

[lonepinealex]

Yes to all this. But a question that remains - and really one central to the garnering of support amongst the public in miscarriage of justice cases - is, if, using alex's colourful descriptor, Amanda Knox had in fact turned out to be a 'heinous bitch', but still as equally clearly innocent, would so many people have 'gone to bat for her' and in such a variety of different ways?

So was it enough, actually, that she was so obviously and demonstrably innocent? Or are the pro innocent supporters in a sense really a mirror image of the pro guilt supporters? Did we need to establish that she is good in order, actively to support her as much as they needed to conceive her as bad in order to attack her?

Probably, in some ways. But actually, it doesn't really matter, because no human being is all bad or all good. So inevitably supporters on both sides of the coin will choose to see the good in their chosen champion and diminish the bad in their minds.

Although in this case, the guilters have created a false dichotomy, because no innocencisto hates Meredith. Our baddies are people like Mignini and Giobbi.

 

[Dougm]

I think they didn't hate Guede because he was quickly (and relatively efficiently) locked up - so they didn't have anything to push against in that respect with him. In addition, I think many pro-guilt commentators rationalised that Guede had sort of in a way "manned up" to his involvement (even though in reality he did absolutely nothing of the sort - he merely tried to invent a story to minimise his guilt which still fit all the solid evidence against him).

I think they didn't focus on Sollecito so much because a) his story and photos were featured in the non-Italian media only a tiny fraction relative to stories and photos about Knox; b) he almost exclusively communicated in Italian, which added an extra barrier to many pro-guilt commentators; c) maybe because of (a) and (b), Sollecito didn't appear to them to be "flaunting his freedom" to anywhere near the same degree as Knox.

And I think they directed most of their anger, bitterness and vitriol at Knox because of all of the above. And in addition, it was easy for them to pigeonhole Knox in the "mean girls" category - that was something that was culturally familiar to many of them: the entitled, spoiled, reasonably-attractive brat who is a bully and a manipulator, and who is always dodging responsibility for her actions.

And lastly, I strongly suspect that for many pro-guilt commentators, Knox formed a strong figure for the projection of their own inadequacies, rages against the system, feelings of disenfranchisement and lack of control, and even unresolved traumatic episodes in their own lives.

I think (genuinely) that sociologists and psychologists (and probably also psychiatrists) would gain some real insight from studying the pro-guilt community in this case.

One thing I have always wondered about the PGP thoughts on Guede. There have been continual charges of racism against Knox and her supporters. Yet the PGP are the ones that:

1) Take the position that one black man, Lumumba, was traded (by Knox, they claim) for another black man, Guede. As if their skin color makes them equivalent somehow.
2) Take the position that Guede only participated in the murder because he was under the spell of the pretty white seductress, Knox. I guess they think that he was too stupid or too lustful to not act of his own mind?

Those points of view are the racist ones. The idea that Guede was a small time burglar and drug dealer who broke in and killed Meredith is pretty straightforward, and does not judge Guede by his skin color, but rather by his actions. The idea that Guede was a rube who blindly followed the devious, sexy white girl plays to all the stereotypes -- the same ones that white plantation owners in the US south attributed to black slaves.

Guede's participation in the murder had nothing to do with his skin color.

 

[LondonJohn]

Given that these factors influence our thinking anyway, isn't it a good thing to bring them out into the open, determine whether or not they're valid and how much weight they should be given - i.e. create a Bayesian model?

I don't really understand why you're saying it's unrealistic to do this, given that surely this is what any jury discussing a case actually does (or should do) even if they don't assign probabilities to each factor. Isn't creating a Bayesian model just doing the exact same thing in a different form? Saying it's not practical suggests there's some mysterious factor influencing human thinking that can't be put into words (or figures).

I do agree that Leila's approach to this is flawed because she leaves out any inconvenient factors, but obviously that doesn't mean the whole approach is flawed.

Like I said, I think it's possible to construct a model that's capable of outputting the overall probability of guilt. But I think it would be hideously complex, and that therefore it would be of relatively no real-world use, since it would simply appear to most people to be a mysterious black box.

And yes, I do think that the complex reasoning processes employed by real people in the real world really do involve complicated, highly-nuanced elements. Not only do those involve the type of logic which is more easily measurable, but also incredibly difficult-to-quantify matters such as accumulated human life experience, intuition and human memories.

To make an (imperfect) analogy: humans stand up and walk and run instinctively, but computer scientists and robotics engineers are only just beginning to build robots that can accurately replicate this full ability. And as another example, humans decide who of the opposite sex (or the same sex) they find sexually attractive all the time - yet outside of some very basic variables, it's still impossible for maths or computer science to replicate this sort of decision-making at the incredibly complex level at which it actually takes place in the real world.

 

[katy_did]

Like I said, I think it's possible to construct a model that's capable of outputting the overall probability of guilt. But I think it would be hideously complex, and that therefore it would be of relatively no real-world use, since it would simply appear to most people to be a mysterious black box.

And yes, I do think that the complex reasoning processes employed by real people in the real world really do involve complicated, highly-nuanced elements. Not only do those involve the type of logic which is more easily measurable, but also incredibly difficult-to-quantify matters such as accumulated human life experience, intuition and human memories.

To make an (imperfect) analogy: humans stand up and walk and run instinctively, but computer scientists and robotics engineers are only just beginning to build robots that can accurately replicate this full ability. And as another example, humans decide who of the opposite sex (or the same sex) they find sexually attractive all the time - yet outside of some very basic variables, it's still impossible for maths or computer science to replicate this sort of decision-making at the incredibly complex level at which it actually takes place in the real world.

Hmmm, well I take your point, but isn't deciding on whether to convict or acquit someone of a crime exactly the type of decision that should be based on clear factors which can be explained in words (or figures, if that's your preference) rather than on 'woollier' factors like intuition or life experience? It could well be that these harder-to-define things affect your decision, but in the end it should still be possible to explain that decision without reference to those factors.

Otherwise the judge could write a motivation report and say "Sorry, I know it's a bit crap but we made our decision based partly on complicated, highly nuanced elements that we couldn't put into words".

 

[Kauffer]

One thing I have always wondered about the PGP thoughts on Guede. There have been continual charges of racism against Knox and her supporters. Yet the PGP are the ones that:

1) Take the position that one black man, Lumumba, was traded (by Knox, they claim) for another black man, Guede. As if their skin color makes them equivalent somehow.
2) Take the position that Guede only participated in the murder because he was under the spell of the pretty white seductress, Knox. I guess they think that he was too stupid or too lustful to not act of his own mind?

Those points of view are the racist ones. The idea that Guede was a small time burglar and drug dealer who broke in and killed Meredith is pretty straightforward, and does not judge Guede by his skin color, but rather by his actions. The idea that Guede was a rube who blindly followed the devious, sexy white girl plays to all the stereotypes -- the same ones that white plantation owners in the US south attributed to black slaves.

Guede's participation in the murder had nothing to do with his skin color.

The PGP position managed to be both misogynist and misandrist at the same time. We have a negative portrayal of an inappropriately sexualised and amoral young woman AND a negative portrayal of young men manipulated by a human siren to act criminally.

 

[LondonJohn]

Hmmm, well I take your point, but isn't deciding on whether to convict or acquit someone of a crime exactly the type of decision that should be based on clear factors which can be explained in words (or figures, if that's your preference) rather than on 'woollier' factors like intuition or life experience? It could well be that these harder-to-define things affect your decision, but in the end it should still be possible to explain that decision without reference to those factors.

Otherwise the judge could write a motivation report and say "Sorry, I know it's a bit crap but we made our decision based partly on complicated, highly nuanced elements that we couldn't put into words".

Not at all. Jurors (and judges in their motivation reports) are only required to reason why they arrived at a certain verdict. And that happens by assessing and weighing the totality of the evidence/testimony. And within that there are a whole bunch of individual - but linked - assessments about the reliability and relative value of each piece of evidence/testimony.

Within all that, there are literally thousands of intangible reasoning processes going on in a complex interactive dance. Let's take just one element from the Knox/Sollecito case. Was the break-in real or was it staged by Knox/Sollecito? The Massei court reasoned that it was staged. They based that reasoning on a huge number of incredibly complex factors, including their opinions on why (or why not) Guede acting alone would have staged a break-in; why only Knox and Sollecito would have had the proper motivation to stage a break-in; whether the glass pattern and rock position in the room supports a rock being thrown in from outside the room or whether it supports someone breaking the glass from inside and placing the rock in the position in which it was discovered; whether the marks on the window shutters were consistent with a rock being thrown from the outside or whether they were consistent with a rock being used from inside in a staging; whether it was reasonable to suppose that if it were not a staging, Guede would have made repeated ascents of the wall to open the exterior shutters then throw the rock then climb up and in; whether the pattern of Romanelli's clothing in the room is consistent with someone who broke in with the intention of burgling the cottage. And so on and so on - this is only a small proportion of the factors that were considered.

And I'd defy you to construct any sort of non-enormous Bayesian model that could replicate the types of reasoning required to answer all of the above questions, just for starters.............

 

[carbonjam72]

Yes to all this. But a question that remains - and really one central to the garnering of support amongst the public in miscarriage of justice cases - is, if, using alex's colourful descriptor, Amanda Knox had in fact turned out to be a 'heinous bitch', but still as equally clearly innocent, would so many people have 'gone to bat for her' and in such a variety of different ways?
So was it enough, actually, that she was so obviously and demonstrably innocent? Or are the pro innocent supporters in a sense really a mirror image of the pro guilt supporters? Did we need to establish that she is good in order, actively to support her as much as they needed to conceive her as bad in order to attack her?

I do think this is an issue that is worth delving into a bit.

Gman Steve Moore wrote a blog entry just before the verdict titled, 'The Big Bad Wolf". The point of it was, that Amanda got trapped in large part by her own naiveté.

She thought she was "helping" the police, and they took advantage of her, but more to the point, they were able to take advantage of her, because she was an innocent.

If Amanda had been a mean person, I think she would have recognized it in others, like the police. None of the other housemates went near the police without a lawyer or responsible adult. That was prudent. Only Amanda and Raf allowed them selves to be vulnerable in this way.

They got framed by the police, targeted and intentionally framed by Mignini, because they were/are unusually gentle and compassionate people.

On the other side, is Mignini, who kidnapped two people from the scene whom he knew to be innocent, just because he needed to fill out his cast for an imagined satanic orgy, to prop up his career and defend himself from criminal charges for having made the identical claims in his Narducci trail cases. He borrowed their lives for 7 years. Mignini ought to answer for that. He really should spend some time in the same jail cell he sent them to.

But I do think its the stark contrast between the compassion and cruelty in the defendants and the prosecutor, that has hooked so many people. Mignini was so bad and so wrong, and even so literally crazy. (In his own criminal trial, he was described "as having fallen prey to a kind of delirium").The images in court were just bewilderingly out of place. The juxtaposition is what made it so compelling.

I wrote it better earlier on when I first joined this board. I said something like, 'let's leave the talk of angels and devils to the mad prosecutor of Perugia. In in the meantime, lets do what we can to reverse this injustice".

It came out right.

 

[anglolawyer]

Hmmm, well I take your point, but isn't deciding on whether to convict or acquit someone of a crime exactly the type of decision that should be based on clear factors which can be explained in words (or figures, if that's your preference) rather than on 'woollier' factors like intuition or life experience? It could well be that these harder-to-define things affect your decision, but in the end it should still be possible to explain that decision without reference to those factors.

Otherwise the judge could write a motivation report and say "Sorry, I know it's a bit crap but we made our decision based partly on complicated, highly nuanced elements that we couldn't put into words".

Is experience always expressible in words or figures? There was a conservative philosopher called Oakeshot who used to go on about experience and how there were things that could not be learned from a book but could only be acquired through experience, like making shoes or something.

Sometimes the task in front of a jury is incredibly simple: do they believe the defendant? They are entitled not to and not to have reasons that would not satisfy a Bayesian - 'I just didn't believe him', 'he came across as lying'. Every other shoplifting case is decided on this basis. The jury's collective life experience is what it brings to the party, not unlike this forum where we exchange our knowledge and incommunicable experience. 'Incommunicable' in the sense that my telling you mine does not equip you with it and vice versa. 'Untransferable' might be better.