Evidence and the very unlikely

OK strange question, how much evidence would you need to prove that an human alien hybrid killed Meredith Kercher?  Assume that were the prosecution's theory.  Assume they had picked someone up and they believed that person was a human alien hybrid.  Assume they were using his alien traits to explain evidence, like he had been able to get through the window because hybrids can jump 12 feet easy, he hadn't left any evidence because hybrids can make their fingers not secrete oils....  This was the prosecution's theory and you as a juror had to rule on the case.  What sort of standard should you hold them to?  You might stay that's a really stupid question, and your reaction is precisely where I want your head at, so bear with me please.

Is it impossible that human alien hybrids exist or just very very unlikely?  You might say impossible.  OK what if there were families of human alien hybrids known to exist, towns which people could visit full of them.  Some had undergone analysis in various biological laboratories and the results were public.    You even knew people, who had met some and see them change shape.  Then you might say, "well then yes I'd believe in them".  In other words, its not impossible its just a question that there is nowhere near enough evidence to believe in something so unlikely.  Very much like Russell's Teapot.  There are lots of good reasons to believe that an animal like the platypus didn't exist, when it was first described and for several years there were debates when it was discovered if it was a fraud or a fluke.  But the evidence overwhelmed the skepticism.  And I think that this is a similar case, you don't really mean "impossible" what you mean instead is highly improbably, that is to say something for which you are going to need lots of high quality evidence.

So given any murder there is a certain percent chance it was done by a human alien hybrid.  There is a certain chance it was done by random quantum effects.  There is a certain chance is it done by a monkey like The Murders in the Rue Morgue.  Those are doubts in any case, they just aren't reasonable doubts, because they are so unlikely.  But remember this situation is different, this is the prosecutor's theory.   Which means the prosecution not only has to prove the crime but because they are using this theory to explain away the counter evidence like the 12 foot jump via. the alien hybrid theory they actually have to provide enough evidence to justify the existence of alien human hybrids.  And of course you are beginning to see where I'm going; while Mignini and Massei's theories may not seem up there with alien human hybrids they are still incredibly unlikely.  So lets work this hypothetical a bit before jumping back to the main case.

There is one more condition.  Maybe even a tremendous amount of evidence doesn't cut it.  The possibility that the pieces of evidence correlate and thus all or most of it is together wrong, that your analysis is wrong, that my analysis is wrong vastly overwhelm the likelihood of those scenarios.  One of the things that will strike you immediately if you read old trials is the sorts of scenarios that are considered likely or unlikely.  Something like an insect disease leading to a local significant shift in a particular insect population (like a bee) leading to a crop failure if it is considered at all, and not in that language, would be treated as unlikely while witchcraft or direct divine intervention are likely explanations for this natural phenomena.  Its hard to account for these variables but they exist with most evidence.     The Structure of Scientific Revolutions, bu Kuhn talks about how science deals with the incredibly unlikely.  Once it shows up it provides it disproves the paradigm of probably, which requires a paradigm shift, and that shift is undertaken only when the evidence becomes truly overwhelming.

Just to put this in perspective lets do a quick through experiment.  Assume you have to decide between just two options A and B (with B being guilty).  Assume that you have pieces of evidence each of which is 70% accurate and fully independent, to help in picking between A and B.  If A and B are equally likely, and your standard of evidence was more likely than not you would just count up the evidence and side without whatever letter had more "evidence" behind it.  For most everyday decisions even 2 pieces of evidence would be 90% and thus good enough.   If OTOH you standard was "beyond a reasonable doubt" say 98% and A and B were still both equally likely, you would only need to go up 4 pieces of evidence.  So this 70% evidence is great stuff for making day to day judgements.

But what if A and B weren't equally likely?  Assume that B were something like a human alien hybrid conjecture and A were something like "drug killing, gang killing, robbery, x-boyfriend, honor killing combined" in other words a grab bag of the alien human hybrid didn't do it.  Lets say that the one in ten billion murders at most are caused by an alien human hybrid. So to meet the reasonable doubt standard we would need B to be 500 billion times more likely based on evidence alone than A.    Which is to say if we have to pick between A and B we are often going to pick A even when most evidence points to B. If it were a pure 70/30 shot then it would take about 23 pieces of non correlating evidence each agreeing,  to make the odds less than 1 in 500 billion.

Ah but happens if I have good quality evidence?  Say 98% evidence like a videotape of my alien human hybrid or a repeatable blood sample that shows him his cells producing a silicon based sugar.  1 piece of evidence for reasonable guilt if A and B are equally likely and only 6 for human alien hybrid.   So its a linear factor of 4.    So if you think one of the pieces of evidence is overwhelming, certainly not as good as a video of the crime.  Go ahead and count it twice.

But here is where it gets tricky notice I keep saying independent.  What if they are not?  Well if they are even slightly dependent on one another that doubles the amount of evidence, moderate and I'm up around 100 pieces of the alien human hybrid.  And If you think about it that feels about right.  You would probably need about 100 anecdotes to believe this murder was committed by an alien human hybrid.  That is to say you believe this evidence correlated about 50% there is some overlap.

Ah....  but you might say.  "Wait a minute, CD!  Nothing in the Meredith Kercher murder theory is as unlikely as an alien human hybrid.  People get into squabbles all the time and someone ends up dead.  Domestic violence is frighteningly common not uncommon".  And you would be right.  You would also be rewriting the prosecution's theory of the case.  And boy is it tempting.  Their theory of the case is tremendously stupid.  It requires us to believe multiple highly low probability things.

Its hard to know exactly what is needed to prove the case and what is rank speculation.  But just starting on a particularly bad part of the report:

Meredith Kercher, returning home around nine in the evening, and without
anything in mind other than having a rest (the night before, Halloween, she had
stayed up very late) and doing some studying. Like her English friends, she thought
she had a class at 10 the following morning, and would not have had any intention
of acquiescing to the demands, held to be of an erotic-sexual nature by what has
already been observed, of whoever entered her room.
Besides, she felt attached to Giacomo Silenzi, with whom she had just started an
intimate relationship, and she was serious young woman with a strong
temperament.

1. How would you know what's on her mind as she is returning home?  What do you think you are writing a novel?
2. Do we really know enough about Meredith's sex life to know whether her 10:00 am class would or would not have had any impact on whether she wanted to have sex.  Heck there would have been many many years without sex if I had to wait till days I could sleep in till noon.
3. How do you know she felt attached with Giacomo Silenzi?  We know she had just started  boffing the pot grower downstairs.  Maybe she just liked his pot?  Maybe she liked his availability.  Maybe she liked the fact that Amanda and Laura had both wanted him and she just wanted to be queen bee?
4. And even the statements themselves, "serious young woman with serious temperament" -- Who is dating a pot dealer and helping him grow the stuff
5. Most people when striking up a conversation with a girl hint around the erotic sexual part a bit.  It might not have been entirely clear.  

And on and on and on goes the rank speculation needed to make this murder work out.   And mind you this is key.  This is paragraph is the evidence that Meredith wouldn't have opened the door and thus someone else let Rudy in.   The fact that someone else let Rudy is the evidence that Amanda had to fake the break in.  Amanda having to fake the break in is one of the key pieces of evidence that Amanda is the murder.

Reading this "evidence" does it sound sufficient to prove beyond a reasonable doubt that Meredith didn't say:

• sure come on in lets smoke some weed and then I gotta go to class or 
• sure my boyfriends away I'd love to play tonight
  or maybe going with different theory, a food theory
• you know I'm already getting sick of pastas, stracciatella I'm totally in the mood for some boiled cabbage, maybe some chips.  I'd love some company!

We are talking a guy who played basketball outside her school and hung with her boyfriend.   This is called the Massei conclusions report, perhaps the Massei wild guesses report would be more fitting?  So lets assume its 80% likely that Meredith didn't open the door herself for Rudy and assuming she didn't open the door that its 80% likely that Rudy had someone let him and given that 80% chance that the someone had to be one of the 4 girls that no one else had a key to a rental apartment that they had had made over the last decade  and 80% that given all this Amanda would have realized she needs to fake a break-in....  well the whole scenario then is only 1 in 3.  And the same way the evidence adds the rank speculation takes the odds down and down and down.  And by way of example, if I used 60% for that the chances would have been just 7.75%.

So you go through the Massei report counting conjectures.  How many of these 80% do you need to make the case.  20, 100,1000?  At 20 its less than 1% likely that things came down the way Massei speculates.  At 100 you are around the one in ten billion we used as a  placeholder for the human alien hybrid.  That's the power of compounding.    It doesn't sound crazy because it is a huge collection of more likely than not theories strung together.  But just multiple them out and you get something incredibly unlikely.  To prove this kind of a sequence, in practice you would need to:

1. Collect evidence
2. Construct a fixed single sequence sequence
3. Collect evidence independent of your evidence in step (1) to confirm / disconfirm your sequence

otherwise you need an astronomical amount of evidence to show you aren't just fitting a conjecture to the facts rather than confirming a conjecture with facts.

The fact is Massei has no idea what happened, because an investigation was never done (see my article on prosecutorial abuse as to why it was never done).  They think they probably have the right people and the rest of the report covers:

a)  Stuff they did investigate
b)  Wild conjectures to tie those scattered pieces of evidence into a case.

And he does hit his conjectures to the facts at hand.  In the pages on the stab wounds and their order he has a high quality autopsy and thus lots of facts he has to fit to.  On the what Meredith was thinking as she approached her door he has essentially none so he is free to assert anything he wants.

There certainly is enough to indict Amanda based on the Massei report, but to convict?  Play a game.  Read the Massei report.  Each time you hit a piece of evidence cancel out 2 conjectures of his, which is being really generous with the evidence.  And that's not counting the fact of how silly some of the conjectures are.

And then there are places where the evidence is just wrong.  I'm not an expert on DNA, one can see extensive evaluation of the evidence all over the web and I don't have the background to know enough to evaluate it.  But this line is different:

Encase forensic analysis software determined, for such time period, that the only files created (last created) or written (last written) were generated, automatically, either by the computer’s operating system or the Firefox web-browser within its own cache: being files generated at regular intervals.

I'm going to get a bit techie for this paragraph, explain how this is total nonsense, feel free to skip it as just an example.  I'd like you if you are on a mac or a linux box to open up a terminal right now and type the phrase man touch.  If you are stuck on a windows box here is what you would have seen link.   Touch is a program designed to change timestamps because it is such a common activity.   Changing those dates are standard Unix activities, I have tons of scripts that modify those dates to things other than their defaults that run on my machine; for example when I push data to the TIVO I script changes to mtime so that it sorts the way I want it on my TV.   Mac's internally have 5 timestamps they associate with files: createDate, contentModDate, attributeModDate, accessDate and backupDate. accessDate implements atime, attributeModDate implements ctime and contentModDate implements mtime. The fsCatalogInfo attribute for a file in objective-C (the default language for system's programming on an Apple) has those five as variables (i.e. for example fsCatalogInfo.createDate), which is to say this is not some deeply hidden attribute, Apple invested money in making these timestamps alterable because programs need to do that so frequently.    Other than those 5 attributes there is no place any time information is stored about file manipulation.    You can in 3 seconds have a file on your mac that was last modified 10 years before you owned the computer, heck before it was created as far as the filesystem is concerned.  And there is no secondary record of these changes.   With an average user timestamps are obviously good evidence.   By definition average users are people who think about how to work their computer not how their computer works.  But Raffaele is a computer science graduate who is doing a degree in genetic programming, he's spending all day thinking abut how computers work.  For him,  the timestamp mean nothing more than the times he choose to assign to files.   He's probably 10x the programmer I am, I'm way over the hill, he's in his prime.  If this were Amanda's computer I'd think "determined that the only files created" was too strong I'd weaken it to something like "indicates that most likely the only files created".    For Raffaele's I'd say "a weak easily modifiable record which at the time of analysis showed..." is a fair characterization. [note added 1/12/11: Rose below translated from Italian what they actually used.  I'm leaving this unmodified for continuity, but the actual forensic method was much less reliable than the one I assumed they used]

I'm nitpicking the timestamp paragraph,  because that is one where I'm not quoting other experts.   I know for a fact that Massei is indicating that something that is only likely as an absolute certainty.     This is at least for me a perfect example of the basic problem with the Massei report, it replaces possible with likely, likely with almost certainly true and almost certainly true with tautologically true.  He takes weak evidence and argues that it shows things way beyond what it does in fact show.  Please google everything in these two paragraphs, check that everything I'm saying about timestamps is absolutely true.  Everyone does this with the inconsequential, "I'm sure I put gas in the car" as shorthand for "I'm usually pretty good about filling it once it gets below half full, and 3 days I remember it was less than 1/2..."  But if my job depended on it, I'd go out and check the car and  I'd still fill it up just in case the gas gauge wasn't working right.  If someone's life depended on it, I'd try and fill it and I'd make sure to have a spare gas container in the trunk.  And that's the level of certainty I would want before locking someone away for a quarter century, beyond a reasonable doubt.   And that is what the law requires.

 And then ask yourself did the forensic analysis really determine what happened on that computer or did it just provide a fallible piece of evidence about what happened on that computer?  And if it didn't then Raffaele can be telling the truth about what he and Amanda did during the time of the murder.  This little forensics was considered a major blow to their alibi.    And while you are thinking about that, read the report for yourself and go find yourself a dozen example like this of these unbelievable leaps of pure conjecture.   In the end there is one key question you absolutely must be able to answer before taking the awesome responsibility of destroying 3 children:

• What lethal acts do we know for certain that Amanda Knox performed?
• What lethal acts do we know for certain that Raffaele Sollecito performed?
• What lethal acts do we know for certain that Rudy Guede performed?

And I have yet to hear an answer to that question.

_____

See also:

• A similar argument was made by Raffaele's attorney's in his appeal (link), translated in the comments to this post here.