Ed Morrisey posted awhile back about some new information uncovered by the intrepid Jay Epstein concerning the Iraq/Qaeda connection. Ed's post refers to an article by Laurie Mylroie on Frontpage, which in turn refers to an Epstein piece on the topic. Well I'm a believer, but this new information doesn't offer very much. Even worse, the way it's presented tends to undermine the credibility of the whole case. But if some critical information is missing from the case, I wouldn't necessarily give up hope.
First though, when I read the following statement it didn't exactly fill me with confidence:
The discovery of the notation in al-Ani’s appointment calendar about a meeting with a “Hamburg student” provides critical corroboration of the Czech claim.
Epstein's contention, and Mylroie's, is that the Hamburg student was Atta, and the entry in the appointment calendar fits into an elaborate and dynamic problematique that Jay has constructed, involving Atta and Iraqi Intelligence. Mylroie's article also suggests that both Czech and US Intelligence are holding other critical evidence. Fair enough, but as one who would love to see this proved, but also as one who knows a bit about proper methodology, I have to say that the above statement is extremely sloppy. Looked at from the perspective of someone already convinced of the connection the appointment book entry might seem like obvious and critical corroboration. But from the perspective of a skeptic, which is the proper perspective for a researcher attempting to falsify the "null hypothesis" that there was no link, the obvious problem is that there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day. Plug the probability into the equation, and multiply it times the other known probabilities, and it doesn't really enhance the case against the null hypothesis very much. Now, the case may be very good without that piece, but what I'm saying is that making such a statement betrays what could be a fatal bias that taints the entire problematique. Indeed, for all we know the appointment wasn't even kept.
[Update: A rather lengthy rebuttal by Blixa, in a comment below, compelled me to reconsider my position a little. If the Czechs have synched up the appointment in the calendar, with the meeting that was eye-witnessed, then it is important corroboration. That's because people can't be in two places at the same time, so we've linked the two accounts such that they can't be unlinked. That means we don't need to consider the independent probability that the appointment in the calendar was Atta. We can consider a conditional probability, because we know two things about the same person. We know he was a student from Hamburg, and we know he looked like Atta. But if the times aren't synched, we don't have that luxury, and we're back to square one. The student in the appointment book could be practically anyone from Hamburg.]
The problem with proper methodology is that it seems inconvenient and overly demanding to most people, because they're focused on "proving a positive" rather than "falsifying a negative." But there's a very good reason for such an inconvenient approach. There just isn't any other way to make reliable inferences without fooling yourself, because we humans have a tendancy to do that a lot. It's one of our "key traits." We're experts at it.
That said, I agree with one of Ed's commentors, that we may see some rather surprising and extraordinary information coming to light as the months progress toward November, involving the two main theses upon which the anti-war folks and Democrats have hung their fortunes: no WMD in Iraq, and the Iraq/Qaeda connection. They don't seem to have noticed, but without both of those pillars the doorway will collapse into an impregnable wall, and if it happens late in the campaign there just won't be time to build another doorway.
Ask yourself this question: If you were G.W. Bush facing the sort of implacably tenacious opposition he has been confronting since 9/11 wouldn't you like an ace in the hole, that you could pull out after the inevitable late-day muckraking of a Sy Hersh expose', or a Zarqawi-ordered tet-like-offensive in Iraq just before the election? If I were in that postion, holding something that was a genuinely critical piece of the puzzle (assuming that holding it back didn't jeapardize national security), I'd be inclined to play it pretty close the the vest, right now. In fact, I'd probably be baiting the opposition to raise the stakes.
Posted by Demosophist at May 11, 2004 07:54 PM | TrackBackthe obvious problem is that there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day
??? Not that I have a problem with most of what you wrote, but this counter-explanation seems inapt to me, are you sure you're saying what you mean to say (or meant to say what you said)?
Unless I'm missing something, the "Hamburg student" meeting does not imply merely, or even at all, that there was some Hamburg student was "contemplating a trip to the CZ Republic" and we (therefore, somehow?) think he's Atta. That seems to be a complete misreading of the Epstein-cum-Mylroie data point.
First of all, to clear the air, the meeting was *in* the Czech Republic, not Hamburg. Second, it was a scheduled meeting *with the Iraqi consul*, so "contemplating a trip to CZ Republic", or anywhere really, doesn't have anything to do with anything. (I doubt one would have to have a "meeting" with the actual consul himself even to get a visa to visit Iraq... I suppose I could be wrong about that.)
No, what you really need to supply is a counter-hypothesis for why a "Hamburg student" or purported "Hamburg student" (of which there are presumably *far fewer* in Prague than in Hamburg, by the way) would have been interested in meeting with the Iraqi consul in Prague, CZ. I don't know that it's so obvious that the Iraq consul in Prague would be meeting with "Hamburg students" all the time, so your argument from there being lots of Hamburg students, let alone lots of Hamburg students who want to visit Prague, just misses the mark. I don't care how many there are, the fraction of them who would conceivably have business with the Iraqi consul there, is still presumptively miniscule.
Now obviously, I agree with you that there *are* conceivable possible counter-explanations but "there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day" doesn't even begin to resemble one.
For the record I agree with you that the Mylroie article is stated too strongly (and a bit weirdly). To me the word "confirmation" is what is misleading people here, because let's be sober, what the "Hamburg student" document "confirms" is not SADDAM BEHIND 9/11!!! but, rather, it confirms the possible veracity of the prior evidence we knew about - of a meeting between Atta and Ani in Prague. (That evidence being: the eyewitness testimony of presumably a Czech agent.)
Previously, that testimony was all that we had. A truly determined skeptic could huffily say "Hmmph, I don't even think that Ani met with ANYONE on that day, I think the Czech agent was on drugs and mis-identified both persons", and I couldn't really have rebutted. But now I can, at least to an extent: now, we have *further* "confirmation" (how about "corroborating circumstantial evidence" instead?) that the event attested to happened and that Atta was indeed the person seen there. Why? First because of Ani having recorded a meeting with a "Hamburg student" in his notebook on the same day as the alleged meeting (thus SOME meeting took place with SOMEONE, it's only reasonable to think), plus (in addition to the agent testifying that that SOMEONE was Atta) toss in whatever "spooky coincidence" factor you think is incurred by the phrase "Hamburg student" appearing both in that notebook and on Atta's visa application (to Czech republic!!) of a year earlier.
As for "proper" way to form inferences about all of this, well heck I don't know, and I must admit that you've confused me here by seeming to switch from "alpha" to "beta" (or vice versa?), for reasons I can't determine. What we have is some Czech agent saying "I saw a meeting between Ani and someone, and that someone, looked just like Atta". Despite the best tap dancing efforts of guys like Isikoff (parroting their FBI sources), as far as I know as of now we have no reason whatsoever to think that Atta having been in CZ on that day was any kind of logical impossibility, indeed it meshes nicely into a gap into his known schedule. And we even have precedence of Atta visiting Prague, twice, for unexplained reasons, so it's not like we can fallback to "Atta would never go to *Prague*, that's absurd". Now what we have here is a reasonably objective reason to believe at least a *part* of Czech intel's testimony: that *some* meeting between Ani and *someone* identifying as a "Hamburg student" (like Atta the *Egyptian* did on his visa application!) did indeed take place. What we're left with is contemplating the probability of some *other* self-described "Hamburg student" having had some kind of business with the Iraqi consul in Prague, CZ, which required them to meet somewhere. (Remember, Ani was seen and identified, right?) I admit that this probability does not seem high to me, but at this point I couldn't tell you whether I'm using an alpha or a beta method.
How about this: if that "Hamburg student" whom Ani met with in Prague wasn't Atta... then who was he? Why doesn't the *real* person who met with Ani on April 8, 2001 (because all indications now are that *someone* falling into the category of "Hamburg student" did) step forward and clear all this up for us? I won't hold my breath ;-) Best,
Posted by: Blixa at May 11, 2004 09:21 PMI had a very lengthy, and somewhat aloof, response. But after thinking about it here's the bottom line: Do they have the times synched? Is the appointment in the calendar for the same time that the eyewitness says he saw the two of them, the Atta-like guy and Ani, together? If they have that then, yes it's important corroboration. If they don't, if it's just a "same day" thing, then it just isn't worth much. But I'll wait before explaining why, because it might not be relevant.
Posted by: Scott (to Blixa) at May 11, 2004 11:25 PMRe: time-synch; I don't know any more than you, I'm looking at the same article(s) you are. By my reading, (1) Czech intelligence believes they observed a meeting on April 8, 2001 between Ani and someone. (2) Ani's notebook shows a meeting with a "Hamburg student" scheduled for April 8, 2001. Is that enough time-synch or are you talking time of day? Are you suggesting that Czech intelligence observed the Iraqi consul meeting with unknown young man at time 1 (*not in the consulate* I remind you, but on "the outskirts of Prague" - is that normal for a consul to schedule meetings outside the consulate?), but that wasn't the "Hamburg student", because the "Hamburg student" meeting of the notebook was really at time 2? That's the counter-scenario you're (kinda sorta) proposing. Sure, it's *possible*.... anything's possible. However now we get down into probabilities again.
In your Update you say, "The student in the appointment book could be practically anyone from Hamburg." In PRAGUE remember. You do realize that these events take place in PRAGUE, CZ, and not Germany, where Hamburg is located? I don't understand why you consider "Hamburg students in Prague, meeting with the Iraqi consul" to be such common everyday events that they are unworthy of notice... See my problem is, you seem to be willfully ignoring much of the prior information we have about this event. If you're *really* seriously suggesting that Ani met with some *other* "Hamburg student", ok I guess, but why are you so blase about the whole Iraqi-consul-meeting-with-student-from-Hamburg event, on its face? Seriously, I ask you: what *possible* business could a "Hamburg student" have with the Iraqi consul, *in Prague*? (...which, perhaps, required them to meet *outside the consulate* on the "outskirts" of town?)
I have to say, the issue of "time-synch" (beyond what we already have: *same day*) is a bit of a red herring. You're seriously raising the counter-scenario of two meetings on April 8? one with Atta look-alike outside the consulate (and observed by Czech intel), and then *another* one with a "Hamburg student" (to take care of notebook entry)?? I have to say, a more plausible counter-scenario is just to posit that the notebook and Czech intel both refer to the same meeting (the one "on the outskirts of Prague"), and the "Hamburg student" just happened to look like Atta...(who just happened to use that phrase "Hamburg student" on his visa application a year earlier...) i.e. that it was all just a big coincidence. At least *that* is a rebuttal I would understand..
What's more disturbing is that it's still unclear that you realize just how inapt "there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day" is as a response to these data points. Seriously: it just misses the mark; everything from the Number of Students in Hamburg, to the existence of Students contemplating a trip to the CZ Republic on that day, is *irrelevant* to constructing a counter-scenario fitting the facts we are given.
I suspect you simply didn't read them (Epstein's data points) that carefully, or perhaps you went with Mylroie's account, and what this all just proves is that her article was indeed poorly written and confusing.
Again, don't get me wrong, I'm not here to say the meeting Definitely took place (or rather, that the young-male subject observed at that meeting was Definitely Atta.. it seems clear that *some* meeting with *someone* took place), but the probabilities are leaning far more in that direction that you're acknowledging, and some of your counterarguments leading you to the opposite conclusion are just so inappropriate, even in some cases irrelevant, that I can't believe you stand by them.
I do think skepticism is warranted here but the data should still actually be understood and addressed. Best,
Posted by: Blixa at May 12, 2004 11:07 AMI don't know any more than you, I'm looking at the same article(s) you are. By my reading, (1) Czech intelligence believes they observed a meeting on April 8, 2001 between Ani and someone. (2) Ani's notebook shows a meeting with a "Hamburg student" scheduled for April 8, 2001. Is that enough time-synch or are you talking time of day?
No, it's not enough. You have to be able to take advantage of the fact that the same person can't be in two places at the same time to link the two pieces of information. If you can't uses that "Bayesian" or conditional approach, you're stuck using some modified "independent probability" approach in which the probability of one event is multiplied by the probability of another event and then the result is added in. Multiplying two fractions, especially if one is very small (as is the case with the "Hamburg Student's" known identity) results in a very small fraction.
Another way we might make the connection is if we know that all of the agent's appointments have been placed on the calendar. But I don't think we know that either.
re you suggesting that Czech intelligence observed the Iraqi consul meeting with unknown young man at time 1 (*not in the consulate* I remind you, but on "the outskirts of Prague" - is that normal for a consul to schedule meetings outside the consulate?), but that wasn't the "Hamburg student", because the "Hamburg student" meeting of the notebook was really at time 2? That's the counter-scenario you're (kinda sorta) proposing. Sure, it's *possible*.... anything's possible. However now we get down into probabilities again.
Well, you're always at probabilities. If you don't know they took place at the time time, that's the way you have to look at it, yeah. I mean, you can modify the approach by calculating some sort of independent probability that the calendar entry refers to the observed meeting. But do we know enough to make that calculation? What are the parameters? The "coincidence" between the Hamburg reference and the meeting that, I'll bet, forms a big part of your certainty, assumes a dependent probability that isn't really justified. It's a coincidence. The odds of that coincidence are actually pretty high, in relative terms, because the calendar reference could refer to quite a lot of people. We know the fellow at the meeting looked like Atta, and we know that the agent probably had a meeting that day with someone who was a student from Hamburg. What that does is "fail to falsify" the hypothesis that the person at the meeting was Atta, and it adds a small increment to the probability that it was Atta. But it's very small. As in .005, or something like that.
I know that's inconvenient, but the silver lining is that it provides a very clear way to connect the dots, and one that's completely unambiguous and impervious to misinterpretation. By demanding a rigorous link you can close that .005 probability to 1 pretty quickly.
So, I think, the relevant question is: Why don't we have that critical piece of information? There are only two options: either no one has it, or it's being held back.
And why would it be held back? I can think of a very good reason. And I can think of a very good reason why the Administration isn't using any of Dan Darling's excellent arguments either. It's simply a better strategy to connect those dots at a more convenient time.
Seriously, I ask you: what *possible* business could a "Hamburg student" have with the Iraqi consul, *in Prague*? (...which, perhaps, required them to meet *outside the consulate* on the "outskirts" of town?)
You see what you've done here? You've assumed they were the same meeting, and only then do you ask the question about what other Hamburg students might be doing there. Your assumption is an imposed condition on the probability without justification from the data.
Proper methodology in a case like this is definitely not second nature, which is probably why you assume I'm "blase'," etc. Emotionality is totally irrelevant. We know what we know, and that's it.
I have to say, a more plausible counter-scenario is just to posit that the notebook and Czech intel both refer to the same meeting (the one "on the outskirts of Prague"), and the "Hamburg student" just happened to look like Atta...(who just happened to use that phrase "Hamburg student" on his visa application a year earlier...) i.e. that it was all just a big coincidence.
Why is it "more plausible?" What does "more plausible" mean, other that that you're willing to make the assumption and just proceed? That's what a conditional, or Bayesian probability is, except that it requires an actually fact rather than a back of the envelope calculation. That would be justified in a "beta method" if we were concerned about a risk of assuming no connection. But that risk is neglible at this point, so we have to abandon the beta method, and adopt the more familiar "innocent until proved guilty" approach. And, if you do that, it conveniently puts you on the same wavelength with the rest of the public. Connect the two dots with an actual piece of data, and you're done... pretty much. Don't connect them, and you have a lot of work in front of you.
Again, proper method is not second nature. It's tough. It's one of the most difficult things to learn about research. For instance, it's something Seymour Hersh never learned.
Posted by: Scott (to Blixa) at May 12, 2004 01:56 PMNo, it's not enough. You have to be able to take advantage of the fact that the same person can't be in two places at the same time to link the two pieces of information
Obviously that's true if one wanted to say that this data point PROVES that the meeting observed by Czech intel (with Atta lookalike) was scheduled as being with a "Hamburg student" and, therefore, possibly Atta himself like Czech intel insists. But I have never said that. You seem to think I'm saying something I'm not. I'm just saying, this is strong circumstantial evidence, due to what I consider to be the unlikelihood that an Iraqi consul in Prague would have been observed in an external meeting with a young man and Atta look-alike, AND ALSO have some scheduled meeting with a "Hamburg student" on the same day, which was NOT the Atta lookalike meeting.
It seems unlikely to me and I can't quite believe that it doesn't also seem unlikely to you.
Another way we might make the connection is if we know that all of the agent's appointments have been placed on the calendar. But I don't think we know that either.
Look, there's no denying that having extra data can only help. As it stands we know that (1) he was observed in an external meeting with an Atta looklike and (2) he had a notation indicating a meeting with a "Hamburg student" on the same day.
Those two meetings not being the same meeting, would imply two (what I consider to be a priori) highly unlikely events *both* happening on the same day. (1) The Iraqi consul goes to the "outskirts of Prague" to meet with some young, Arab-looking Arab-speaking man whom Czech intelligence can't identify, and which makes them suspicious, the purpose of which meeting al-Ani won't reveal when asked by Czech intel, and he's (therefore) tossed out of the country over it. (2) Meanwhile, some student from Hamburg, a different person, up and decides he'd like to meet with the Iraqi consul in Prague, for some unfathomable reason, on the same day.
Looking at (1) and (2) individually, they both seem to be small-probability events. But as you've just got through explaining, multiplying two small fractions results in....
the calendar reference could refer to quite a lot of people
This is where I have to disagree. The calendar reference refers to someone who self-describes as a "Hamburg student", evidently. You still seem to be stuck on that, saying "there's lots of students in Hamburg". But what we really have are a nested, descending sequence of sets, ALL of which the unknown subject belongs to:
-Hamburg students, or people who would self-describe as such
-"Hamburg students" who happened to be in Prague on April 8, 2001
-"Hamburg students" in Prague on April 8, 2001 AND had some sort of business with the Iraqi consulate (! seriously tell me WHAT this could be !)
-"Hamburg students" in Prague on 4/8/2001, who had business with the Iraqi consulate, which couldn't be handled just by showing up at the window (as can visa applications), but which required making a prior appointment
-"Hamburg students" in Prague on 4/8/2001, who had business w/Iraq consulate, which required an appointment, AND which required meeting with the consul himself (not just some consulate employee, as would be sufficient for e.g. visa applications)
-"Hamburg students" in Prague on 4/8/2001, who had business w/Iraq consulate, which required an appointment, which required meeting with the consul himself, BUT whose actual name the consul would not put down in a calendar scheduling the meeting, preferring the generic term "Hamburg student" for some reason.
See: the "Hamburg student" of the calendar entry, whoever he was, belongs to all of these sets, not just the first one.
Of course, there is every possibility in the world that it was not Atta. But you can't get to that conclusion by reasoning "there's lots of students in Hamburg", you see? (Let alone "lots of students in Hamburg who want to visit Czechia", which doesn't even have anything to do with the issue...)
What that does is "fail to falsify" the hypothesis that the person at the meeting was Atta, and it adds a small increment to the probability that it was Atta. But it's very small. As in .005, or something like that.
Apart from perhaps adjusting your .005 estimate of the increment, I am NOT saying anything other than this. I'm truly sorry that I seem to have given you the impression that I am.
[what *possible* business could a "Hamburg student" have with the Iraqi consul, *in Prague*? (...which, perhaps, required them to meet *outside the consulate* on the "outskirts" of town?)]
You see what you've done here? You've assumed they were the same meeting, and only then do you ask the question about what other Hamburg students might be doing there.
No, I really haven't. You are probably looking at my parenthetical remark, but please don't ignore the word "perhaps". Indeed, the "Hamburg student" meeting could have been a second, separate meeting the consul had on that day, and thus not the one observed by Czech intel on the outskirts of town; it could have been at the consulate, of course. That it was on the outskirts of town was just a "perhaps", get it? I must say however that I do reckon the notion of a "Hamburg student" needing to meet with the Iraqi consul in Prague personally, whether inside the consulate or outside, to be a small-probability event in and of itself. Seriously, don't you?
Proper methodology in a case like this is definitely not second nature, which is probably why you assume I'm "blase'," etc.
It's more that I'm wondering why you're sticking to "proper methodology" in this case and not others. For the record, I agree with you (since this is all you seem to be saying) that "Atta met with Ani in Prague on that day" is NOT a mathematically proven statement, either before this Epstein data or after. What I don't understand is why everything that falls short of mathematical proof suddenly fails to interest you.
[more plausible counter-scenario being one meeting, not Atta] Why is it "more plausible?"
It's more plausible according to how I reason, with logic combined with experience and intuition applied to known facts in this case. How I reason is an internal process that you really can't expect me to go into any more detail about. You can disagree with me about it being more plausible if you want but there's not a whole lot more to say about plausibility and anyway, I doubt that you do.
What does "more plausible" mean, other that that you're willing to make the assumption and just proceed?
what "assumption"? who's "proceeding"? This is really disorienting, you're assuming I'm saying a whole lot of stuff I'm simply NOT SAYING
That would be justified in a "beta method" if we were concerned about a risk of assuming no connection. But that risk is neglible at this point, so we have to abandon the beta method, and adopt the more familiar "innocent until proved guilty" approach.
Aha, I see your point, you're saying beta method is not needed here because there's no risk of assuming no connection (unlike Saddam-AQ connection etc), so you feel at liberty to (indeed apparently we "have to") proceed more cautiously. Ok, that's fine, I understand that aspect of what you're saying now.
I want to emphasize that (although I'm still not sure why we "have to" use alpha method) I haven't drawn any *conclusions* about anything, I'm not "proceeding" anywhere, and I don't think this is a sewn-up case. I'm looking at a set of circumstantial facts and drawing the most reasonable inference from them that I can. When I do that, I find that I'm left with about a .9+ probability that the meeting indeed occurred and that it was indeed with Atta. I'm saying NO MORE (and no less) than that. That means that you know how I'd place a finite bet (i.e. what odds I'd take), for example, but it DOESN'T mean that I'd stake my life on it, ok? For that I'd still use "alpha" method, like you....
Connect the two dots with an actual piece of data, and you're done... pretty much.
If this refers to the time-synch question, under "alpha" method I'd say you're still not done. All you'd have is that only one meeting took place, with a self-described "Hamburg student". It still could be simply an Atta lookalike and his "Hamburg student" phrasing on visa application just a coincidence.
Anyway, you've cleared stuff up a bit for me (i.e. why "beta" is not appropriate here), so thanks. I still don't understand, and am still disturbed by, the fact that you seem to so seriously misunderstand the nature of some of these data points, with non sequiturs about Hamburg students "wanting to go to the Czech Republic", and so on, but I'm content to leave it here if you are.
Posted by: Blixa at May 12, 2004 03:33 PMBlixa:
My time is somewhat limitted. No we aren't attempting to prove that Atta was there. Even if we get the events synched it wouldn't prove it. It would just be very strong circumstancial evidence, strong enough to allow us to reject the null hypothesis that there was no link.
It seems unlikely to me and I can't quite believe that it doesn't also seem unlikely to you.
Whether it seems likely to us isn't relevant. We're believers. I'm talking about what it would take to convince a doubter, which requires a dispassionate and methodologically rigorous approach.
Look at this this way, if you can synch the two data points then what you have is really one data point with more information. If you can't synch them, then you have to build a bridge between them, and the bridge will be strong or rickety depending on the quality of the data you have. It would involves knowing where the agent was at other times during the day, what his habits are vis recording proposed meetings, stuff like that. Real data in other words, that relates to these two data points directly. You can't just assume that the two are the same because you think it's probably the case, because you have this whole big theory it all fits into. That's not appropriate.
Making the assumption that the two are the same, for instance, because you have a lot of other evidence that points to a relationship with Atta may be fine, but it doesn't make that calendar note any more significant than it is. You have to look at it separately, unless the other stuff you know relates directly in a material way to both data points. And even then it may improve the odds only a small amount. Building a solid bridge is extremely difficult if you have to use brute force, but it can be done, sometimes.
I can recommend a book that discusses scientific method in the social sciences, that might be relevant. The problem is that it requires having some other specialized knowledge, but you could probably thumb through it and get to a relevant passage. Suffice to say that everyone learning method rebels at this stuff. It just strikes people as unfair, until they get used to it, and know why it has to be this way. After you've seen a few serious mistakes taken as a result of not using approprate methodology you become convinced.
Anyway, I think you've got it now. Personally, I think the guy was Atta, but I wouldn't take it to the bank.
Posted by: Scott (to Blixa) at May 12, 2004 09:32 PMI'm talking about what it would take to convince a doubter
I see. I wasn't. Obviously you are right that more will be required to convince a doubter, but that was obvious to me. There are some doubters for whom videotape with full-audio and from at least two separate angles will be required. I'm beyond caring ;-)
Posted by: Blixa at May 13, 2004 12:05 PMP.S. Just realized videotape would never be enough. Could always be just a lookalike, still. We would still need DNA or fingerprint proof that the man shown on the hypothesized videotape(s) meeting with al-Ani was actually Atta. Even if al-Ani *thought* he was meeting with Atta, could just be an Atta body double posing as Atta to fool al-Ani. "beyond reasonable doubt" you see.
P.P.S. Also, pride won't let me not protest about something. You've continually misinterpreted my questioning *why* you were using alpha method (i.e. let "Atta never met with Ani" be H0 null hypothesis) for "resistance to" that method or failure to understand that method. I'm familiar enough with scientific reasoning and null hypotheses thank you very much. I just didn't understand why you were insisting on using it here, and not (for example) in the Saddam-AQ connection case. Now I do, even if I don't quite agree about the necessity of doing so. (But that's because "convincing skeptics" was never my primary motivation..)
P.P.P.S. Do you understand why "there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day" was a non sequitur, and " The student in the appointment book could be practically anyone from Hamburg" is actually way too broad, yet? Just curious... these things still bother me, usually you're much more careful and I still can't quite believe you've read Epstein's data carefully enough ;-)
Posted by: Blixa at May 13, 2004 02:13 PMWe would still need DNA or fingerprint proof that the man shown on the hypothesized videotape(s) meeting with al-Ani was actually Atta. Even if al-Ani *thought* he was meeting with Atta, could just be an Atta body double posing as Atta to fool al-Ani. "beyond reasonable doubt" you see.
I don't want to presume too much about your background, but are you familiar with the Logical Positivists versus Karl Popper, and the revolution in Scientific Method that the defeat of the L Positivists entailed? Essentially we aren't required to prove anything. We presume the null until we have a threshold-exceeding reason to refect or falsify it. After that we presume the non-null, until there's a threshold-exceeding reason to reject it. "Verification," or in other words "proof (positive)" is no longer the driving paradigm in method. Rather, it's "falsification."
You may know this already though.
Point taken that many people know about hypothesis testing, at least as it's used in statistical analysis.
Do you understand why "there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day" was a non sequitur, and " The student in the appointment book could be practically anyone from Hamburg" is actually way too broad, yet?
I'm not using this in reference to the overall thesis of Epstein's about whether or not Ani met with Atta. Viewed independently of all of that, however, who could the student in the reference be? Independently, that's the key. You have to ignore Epstein's whole case, and just consider the independent probability that this person referenced in the note was Atta. Only after you've been able to identify him as Atta do you get to fold it into the rest of the data. The only exception to this is if some data from Epstein is actually materially relevant to the calendar entry, such as time of day, etc.
No, it's not a non-sequitor. It's the heart of the matter.
Posted by: Scott (to Blixa) at May 13, 2004 03:08 PMWe presume the null until we have a threshold-exceeding reason to refect or falsify it.
And for some doubters that threshold is quite high. (which was my point) So high I'm beyond worrying about swaying them. In fact for some people the effective threshold is, "the hypothesis must not be rejected unless it is deemed that this will help get Bush out of office".
Viewed independently of all of that, however, who could the student in the reference be? Independently, that's the key.
Of course, but see the nested sequence in my previous comment (which DID view the "Hamburg student" issue independently BTW), in short he would have to be: (1) a self-described "Hamburg student" who (2) was in Prague (3) on that day (4) who had some kind of business w/Iraqi consulate (5) which required an appointment (6) with the consul himself (7) who would not use his actual name but rather just "Hamburg student" when scheduling the appointment. That's about seven qualifiers, not just one, which the unknown subject would have to satisfy.
Yet you had said "he could be just about anyone from Hamburg". This is simply too broad, like I said. By orders of magnitude in fact. The vast majority of people "from Hamburg" do not satisfy (2)-(7), even if some ~half of them could conceivably satisfy (1).
The only exception to this is if some data from Epstein is actually materially relevant to the calendar entry, such as time of day, etc.
Another thing you're missing is that we don't need exact time-of-day correspondence. There is a continuum of possibilities. Sure if the external meeting was at 8 a.m. and the "Hamburg student" meeting was scheduled at 7 p.m. then we have damn good reason to decorrelate the meetings and consider them pretty definitively separate events. But if the external meeting was observed at 3:15 (and concluded soon after) and the appointment book's "Hamburg student" meeting is written into the 2:00-3:00 time slot, we may have reason to increase the probability of them being the same meeting, depending on whether we think Ani could have had a meeting end at 3:00 and then rush to "the outskirts of Prague" site in 15 minutes in time to get observed by Czech intel. This depends on e.g. consul-"outskirts" distance, which we don't know. If he could've, then fine, adhere to the "different meetings" hypothesis, but what if the external observation was at 3:10? 3:05? 3:01? What if the Czech spook was tailing al-Ani most of the day (how else would he have been able to observe an unreported external meeting with someone)? The probability against H0 (different meetings) being true starts to increase when you fill in this kind of information... there's a continuum here, not a binary switch.
Part of what I've been saying is that this continually-increasing probability has increased away from 0 just by virtue of the fact of the two meetings/scheduled meetings known to be on the same day. Surely the "Hamburg student"/Prague outskirts meetings both being on the same day means their probability against their not being the same is higher than if they had been on different days. (.00001 vs. 0, if you like.)
No, it's not a non-sequitor. It's the heart of the matter.
Please understand (because I can only surmise that you do not) what I was saying was a non-sequitur: it was the following statement
there are probably millions of students in Hamburg, many of whom might have been contemplating a trip to the CZ Repubic on that day
And please, please try to understand why that is a non-sequitur. I'm actually a bit surprised it hasn't sunk in yet.
Seriously: ask yourself what "contemplating a trip to the CZ Republic" has to do with anything?
Whether you're a "Hamburg student" or not, you don't go visit the Iraqi consul in Prague if you're "contemplating a trip to the CZ Republic". The former will not help the latter, and the latter would not motivate one to do the former.
Among other things by the time you get to that meeting, you're already in the CZ Republic for crying out loud!! Or, if you really are a "student in Hamburg" you're going to have a hard time visiting the Iraqi consul in Prague if you don't visit the CZ Republic first, which may (or may not - not sure about EU rules nowadays? and is Czechia in EU?) require visiting the Czech embassy in Germany to get a visa, but certainly not the Iraqi consul, in Germany or anywhere else.
Seriously, you really don't get the problem with this???? You may as well have said "there are probably millions of students in Hamburg, many of whom might like soccer". Should I just give up on it sinking in? It's frustrating because this is a somewhat minor point but as long as it doesn't sink in, it's hard to have any confidence that you've thought about these matters sufficiently.
Posted by: Blixa at May 13, 2004 04:46 PMDo we place Atta in Prague on that day? Independent of the eyewitness acount, I mean? I can't recall whether we actually know that, although it seems to be implied. Well, it's taken for granted at any rate.
Posted by: Scott (to Blixa) at May 13, 2004 08:36 PMDo we place Atta in Prague on that day?
No. And you still don't get it. Sigh. We DO place the "Hamburg student", whoever he was, in Prague on that day. Or he at least *planned* to be in Prague on that day, at some point, given that he made an appointment with the Iraqi consul in Prague.
Independent of the eyewitness acount, I mean?
Just as a side note, an eyewitness account is not nothing. Sure, one eyewitness is not enough to convict for treason, but it can be for other crimes....
I can't recall whether we actually know that, although it seems to be implied. Well, it's taken for granted at any rate.
By whom? Certainly not me.
You did not read my post. I give up.
Posted by: Blixa at May 14, 2004 09:55 AMDon't be so testy. I read your post and had prepared a rather long response, but decided that there wasn't much point in going further if we know Atta was in Praque on that day. But because we don't know that (which is what I suspected) what we have is something analogous to the following sitution. (Sometimes it helps to think of an analogy with ridiculous characters, to get the emotional charge off of things.)
Lets say I was from Duluth, and I suspected my girlfriend was having an affair with an Iraqi consul in Prague on such-and-such a day. So I buy a plane ticket, sneak into his office, look on his computer, and sure enough he has a meeting with a "student from Duluth."
Well heck, that's gotta be my girlfriend, dammit! But what do I know about other people he might know from Duluth? I mean, actually, I don't even know that he even knows my girlfriend. I just know some guy who says he saw that consul with someone who looked like my girlfriend in a bar that day. So, I'm all over it.
I'm pretty sure that if I actually did all that stuff I'd have a few friends who'd tell me not to hypeventilate too much, especially since I don't even know my girlfriend is within a thousand miles of Prague.
So what real connections do I have? I've got a suspicion, based on rumor. (And here we're talking about the connection with the consul, not the general propsition that my girlfriend is cheating on me.) I have the testimony of someone who said he saw the consul with someone who looked like my girlfriend. And I have this appointment thingy?
I don't know where my girlfriend is. I have no material reason to think she even knows this consul, other than the rumors. And I have this note that, coincidentally, references someone from Duluth. How much do all of those things add to the eyewitness account? That's my strongest evidence, but it's far from certain so I can't use it to bolster my opinion that the calendar reference is to my girlfriend. That's an invalid inference. You can't use uncertain information to bolster uncertain information because the probability is less than 1.
Let's say that through other means I've determined that the eyewitness account is 40% reliable, and the independent probability that the reference in the calendar is to my girlfriend is 5%. .4 * .05 is .02, which added to 0.4 is 0.42 . So that's what the calendar reference gave me. It just isn't decisive. Not nearly as decisive as my "naked intuition" tells me, at least. And saying that the odds of the calendar reference being my girlfriend is 5% is a gift. It makes a lot of assumptions about how much I can narrow those odds.
Yes I did read your post. Let me deal with the crux of the matter:
Whether you're a "Hamburg student" or not, you don't go visit the Iraqi consul in Prague if you're "contemplating a trip to the CZ Republic". The former will not help the latter, and the latter would not motivate one to do the former.
No, but the only people you can even consider as potential appointments are those at least contemplating a visit. And without more information you have to consider all of them equally likely.
Why is this not getting through? What else do you know about the person that bolsters the odds? You know it refers to a student, so you can eliminate all the people contemplating a visit, who aren't students. But, potentially, all of those may have a reason to visit the Iraqi consulate. They may know someone in an Iraqi prison and want to submit an entreaty on their behalf, after having already visited the Hamburg consulate. Seeing two people increases your odds of success over seeing one.
You can't even place Atta in the city on that day. You can't even verify that the appointment was actually kept. How, practically speaking, do you narrow that potential population down? You can't use the eyewitness acount, as I've said, because the only way you can combine those two observations is by multiplying their independent probabilities and then adding that new term in to the equation. Essentially you're asking your conclusion, or your assessment of the odds of their being a plot in the first place, to bolster your evidence of such a plot. You have a recursion in which the conclusion enhances the evidence enhances the conclusion enhances the evidence, etc., etc., And the result is that while your coefficient is getting large and larger so is your error term. As a result, you simply can't have much confidence in that coefficient.
The problem of recursion is serious. Under normal circumstances where you're making inference from data without allowing your inferences to asses the reliability of your data you have a static error term. It may be large or it may be small, but it's at least predictable. In the case of recursion you simply have no idea how big your error term really is, so you can't even have a confidence interval that makes sense. It could be anything.
Posted by: Scott (to Blixa) at May 14, 2004 11:10 AMBlixa:
You've made a commitment to the following idea, so that gives me a foothold on what the problem might be (whether it's me or you).
Among other things by the time you get to that meeting, you're already in the CZ Republic for crying out loud!!
Good example of reasoning back from your conclusion, which is the essence of the recursion problem. What justifies saying that a meeting took place? Do you have appointments in your appointment book that were never kept? And that's the only reason for considering everyone with the intention of going to Praque. If you know, or can be virtually certain, that a meeting actually took place then you can narrow the search down to those Hamburg students who actually made the trip to Prague and were there on that day. Fair enough?
And you obviously can't use the eyewitness account to bolster your certainty that the meeting was kept, because you'd have to assume the conclusion you're trying to test: that they were both with the same person.
Posted by: Scott (to Blixa, with an idea) at May 14, 2004 01:27 PM[Duluth analogy] Well heck, that's gotta be my girlfriend, dammit!
It's gotta be your girlfriend if you examine the pool of people who could *possibly* qualify as that "student from Duluth", and it contains only one person, your girlfriend.
If it contains two people (your girlfriend and someone else), then lacking other information to prefer the other person, the probability is less than 1, but still significant, that the "student from Duluth" was your girlfriend. If it contains three people, the probability is less still. And so on.
That's why examining the pool of people who could have been that "student from Hamburg" is important. And, like I have been trying to get through to you, that pool does NOT include all students in Hamburg. Far from it.
I'm pretty sure that if I actually did all that stuff I'd have a few friends who'd tell me not to hypeventilate too much, especially since I don't even know my girlfriend is within a thousand miles of Prague.
Apart from the eyewitness account which placed her there, that is. Which coincides with a completely-unexplained gap in her schedule. (And the fact that you have a gap in her schedule is actually more significant in this case given that after all this went down, you've searched her dwelling(s) and economic transactions and obtained intelligence about her movements based on car rentals, ATM receipts, phone records and the like, and don't have such gaps elsewhere.)
I have no material reason to think she even knows this consul, other than the rumors.
And the eyewitness account of them together.
And I have this note that, coincidentally, references someone from Duluth. How much do all of those things add to the eyewitness account?
More than a little, less than a lot.
That's my strongest evidence, but it's far from certain so I can't use it to bolster my opinion that the calendar reference is to my girlfriend.
The fact that it's far from certain doesn't mean you "can't" use it to "bolster" your opinion at all. You couldn't use it in a court of law or a scientific journal but that's not the same thing.
You can't use uncertain information to bolster uncertain information because the probability is less than 1.
What exactly do you think "bolster" means? Opinions can't be bolstered unless the probability is switched to 1? Do you have an opinion on Michael Jackson? Would you let him baby-sit your children? If not, why not?
No, but the only people you can even consider as potential appointments are those at least contemplating a visit.
Fair enough. Yes, "contemplated a visit to Prague" is a requisite precursor to qualifying as the "Hamburg student" who had an appointment with Ani. Essentially this is my qualifier (2) above, with some modification (that "considering" being in Prague on 4/8/2001, even if not actually there, would have been sufficient...). But it's not the only qualifier. What about (3)-(7)?
And without more information you have to consider all of them equally likely.
Like hell I do. Out of all "Hamburg students" who at least momentarily considered visiting Prague on 4/8/2001 I reckon only a tiny fraction of them would have
(4) had some kind of business w/Iraqi consulate (5) which required an appointment (6) with the consul himself (7) who would not use his actual name but rather just "Hamburg student" when scheduling the appointment.
Yet we KNOW that our "Hamburg student", whoever he was, satisifes ALL of (4)-(7). Sorry to be so repetitive but are you sure you read my comment?
Why is this not getting through?
Because it's false and willfully ignores additional circumstantial and contextual information that we have about these events.
What else do you know about the person that bolsters the odds?
(4)-(7), above
You know it refers to a student, so you can eliminate all the people contemplating a visit, who aren't students.
Well, I know the person *can be described as* a "Hamburg student", anyway. Which is (probably) a broader category, including ex-students, dropouts, wannabes, delusionals, etc. (Otherwise we might have to rule out Atta himself I think. ;-) Similarly Ani may have made a mistake when making the notation (maybe the meeting was with a Hamburger Salesman and he messed up..)
On the other hand, it could be a *smaller* category. There are probably at least some people who are students, in Hamburg, but would not necessarily describe themselves that way, to anyone, let alone when making an appointment with the Iraqi consul in Prague.
But, potentially, all of those may have a reason to visit the Iraqi consulate.
*snort*
I dissent.
They may know someone in an Iraqi prison and want to submit an entreaty on their behalf
A valid hypothetical example. This would apply to some small fraction of "Hamburg students", certainly not all of them.
You can't even place Atta in the city on that day.
Apart from the eyewitness testimony which places him there and is contradicted by nothing.
You can't even verify that the appointment was actually kept.
That is correct. I only know that it was made. That it was kept has some probability associated with it. We are for the most part examining the fact that it was made and discussing whom it could have been made *with*.
How, practically speaking, do you narrow that potential population down?
Potential population of "Hamburg students"? By using my razors (1)-(7) above, for starters. I don't know how "practical" that would be for actual calculation but in principle there's no problem here.
You do raise a good issue however. If we had a functional media, there is a highly relevant piece of news reporting to be done here. We have this "Hamburg student" calendar entry and suspicions (nothing more) that it was Atta with whom this appointment was made. If those suspicions are incorrect (which would be nice to know), then ipso facto the appointment was made with a different person than Mohammed Atta. That person is (or was) out there somewhere. A good reporter who cared about these matters, and was not ideologically blinded (or heck if he *was* ideologically blinded), might try to find that person (if only to parade him in front of Bush/the neocons and say "not Atta!!"). As a place to start I might even suggest Hamburg. :-) Why not a call to all "Hamburg students" - say, an advertisement or short article in a local paper - asking anyone who might have made this appointment to step forward? The potential pool here is large (if not as large as you seem to think), but not infinite. And if the REAL "Hamburg student" is just some innocent Joe Schmoe you'd think he'd (a) actually be aware of the Atta-in-Prague rumor ("hey, that's the same day I had that appointment with Ani...I wonder if they mistake me for Atta...") since it's been in the news and after all the "Hamburg student", whoever he was, must have had *some* interest in matters Iraq; and (b) be more than happy to clear this up by saying that it was he who was the "Hamburg student", and he is not Mohammed Atta, he just wanted the release of his friend or a visa or a job or whatever. (Unless he's a "neocon" and *wants* people to go on with the misapprehension that he was Atta visiting Ani? ;-)
Essentially you're asking your conclusion, or your assessment of the odds of their being a plot in the first place, to bolster your evidence of such a plot.
I don't think so. What "conclusion" by the way?
recursion... And the result is that while your coefficient is getting large and larger so is your error term.
Sorry but I haven't the slightest idea what you're talking about. Sounds like renormalization group theory, should I go crack open my old quantum mechanics texts? ;-)
[Among other things by the time you get to that meeting, you're already in the CZ Republic for crying out loud!!]
Good example of reasoning back from your conclusion, which is the essence of the recursion problem. What justifies saying that a meeting took place?
You've ignored context, which was something like.... You: the person who was the "Hamburg student" who made an appointment, made that appointment because he was contemplating a trip to CZ Republic. Me: you don't make an appointment with the Iraqi consul in Prague "because" you're contemplating a trip to the CZ Republic, that makes no sense. A meeting with Ani in Prague only makes sense after you're ALREAD IN CZ in the first place. You: AHA you're assuming he went to CZ Republic and that the meeting was kept!
NO, we were conditioned on that, at that point of the discussion. We were analyzing the situation *what could motivate a "Hamburg student" to make an appointment with the Iraqi consul in Prague* and your answer seems to have been "if he was contemplating a trip to the CZ Republic". This was a non sequitur, and I called it as such. Contemplating a trip to the CZ Republic does not militate towards the idea "well I guess I better make an appointment with the Iraqi consul in Prague then" any more than liking soccer militates towards that. I've made this point repeatedly and you've now missed it for at least the third time.
And you obviously can't use the eyewitness account to bolster your certainty that the meeting was kept, because you'd have to assume the conclusion you're trying to test: that they were both with the same person.
Again, like hell I can't use it to "bolster". I can't use it to *mathematically prove* the meeting was kept but I don't know why you think that's what "bolster" means. My certainty that the meeting was kept is less than 1 but greater than 0, and knowing that Ani had (and kept) a meeting with SOMEONE on that day (the guy observed by Czech intel) certainly gives me more confidence that he could have kept the "Hamburg student" meeting, than if he was known to have been at home sick in bed the whole day. Capisce? But please don't misinterpret, I'm not saying it makes me CERTAIN or ALMOST CERTAIN, just MORE CONFIDENT. You seem to blur amongst these concepts whenever it would make me look bad and my statements look the most unsupportable ;-P
Sorry, but it still seems like you're either not reading my posts, or we just have a serious miscommunication, either way you've given me no confidence that you've understood anything I've said, I'll go ahead and assume the fault is mine somewhere. I don't think I'm saying what you seem to think I'm saying, and meanwhile I'm making some fairly obvious points that don't seem to be reaching your brain. I walk away from this knowing loud and clear that you know about Popper, about recursion, about independent and conditional probabilities, and some statistical modelling, and that two small fractions multiplied together become smaller, and many other mathematical notions you appear eager to demonstrate your knowledge of, but I must say I'm a bit flabbergasted that you keep trying to shoehorn and distort my comments into some little box that lets you demonstrate that you have some sort of statistical training, none of which discussions are directly applicable to the points I was trying, and failing, to make. I should've stuck to my determination to give up.
Posted by: Blixa at May 14, 2004 02:14 PMHmm. Should have looked at additional sources before this discussion, would've mooted much (not all) of your analysis. Apparently the person observed at external meeting was also said, by an informant for Czech intel, to be a "Hamburg student":
"Then, on April 8, 2001, a BIS watcher saw al-Ani meeting in a restaurant outside Prague with an Arab man in his 20s. This set off alarm bells because a BIS informant in the Arab community had provided information indicating that the person with whom al-Ani was meeting was a visiting "student" from Hamburg—and one who was potentially dangerous."
So Ani met with a guy tentatively identified by a source as a "Hamburg student". Later an entry in his calendar also shows a scheduled meeting with a "Hamburg student".
But, sure, it's possible they were not the same meeting.. maybe Ani met with (or planned to meet with) *two* "Hamburg students" on that day, neither of them Atta.. Nope, the probability against Atta-in-Prague being false has not decreased one iota, right? ;-)
Posted by: Blixa at May 14, 2004 09:32 PMi think I meant increased not decreased there. ok sorry no more polluting comments ;-)
Posted by: Blixa at May 14, 2004 09:33 PM