What your doctor doesn’t know could kill you

Boston Globe article on PKC (via Jorn)
And Jorn’s page of Lawrence Weed’s Problem-Knowledge Couplers biographical links.
Another interesting link from Jorn. A variation of the medical expert diagnostic system which, rather than following pre-structured diagnostic sequences of questions, suspends analysis until after asking many questions then seeks to find patterns and relationships. Interesting evidence of effectiveness, but even more interesting evidence of human nature response (from real doctors accused of making wrong diagnoses.)

From the AI perspective, seriously suggesting evidence that the system is more “intelligent” (more capable of analysing and diagnosing a problem) than an experienced doctor. Also an angle on simply capturing more “raw” facts without any rationalisation or analytical conditioning provides for much better use of the information later – hence to focus on “record-keeping”. “Taking 20 minutes to get the facts straight from the start, he says, saves time and money down the road.” (see my angle on the “fraudster” thread below.) “Straight Facts” – no interpretation, or context free interpretation only – the main issue.

It gets better (it’s a long three page article) – also a line on dangers of “averaging” and statisical probabilities in a complex situation. A line on suspending judgement in analysis. Focus on “couplers” – things which might relate symptoms, not the symptoms themselves. Emperor’s suit of clothes or Galbraith / DeLorean / Argyris conspiracy of silence too “In the field of medicine, for many, many years, we have all been pretending we do not know that everybody has been pretending.”. Even scientific method – a doctors response “…. science is our tool – how can a computer do that?” Its not the technology, it’s the method / the model – simple rational logic (scientific method) is just not the best way to know about a human condition. Know it.

The guy is running a company selling his PKC sofware tool, but the rhetorical evidence here is very convincing stuff. A gold mine article. (Thanks again Jorn).

IQ vs Tacit Knowledge

New Yorker arcticle via Jorn.
Interesting comparison of individual IQ vs tacit knowledge and inter-personal relationships and correlations with “performance” of organisations, with some powerful evidence from McKinsey and connections with Enron / Andersens downfall. No suprises to discover it’s not what you know but who you know that matters, but its good to have better than anecdotal evidence of common sense.

In this hiatus in proceedings, still reading several different threads. Almost finished DNA’s Salmon of Doubt – a good compendium of existing published articles, plus 25 / 30% of an unfinished Dirk Gently novel (stangely – reviews by DNA and others of the work in progress suggest it was better plot material for an H2G2 novel, but I have to say it carries on where holistic detecting left off for me.) Confirms the guy really was a genius.

Have put Quine’s Word and Object to one side in the meantime, but still very interesting. Cogitating on Quine’s rather scathing analysis of scientific method, I’m also forming a new thread of complexity vs simplicity. Complex systems theory and the like are used by many in describing both the nature of true knowledge and the nature of human organisations, and of course, many people warn of applying “scientific method” to analysis of human organisations, along the lines that social science is not a science. Blindingly obvious I guess, but simplicity is at the heart of scientific method – logical induction by virtue of controlled experiments, with minimum numbers of variable changing at any given time in order to able to claim evidence as “proof”. By definition, complex systems are not like this. Quine doesn’t mention Occam’s Razor, but he’s talking about the same principle. If knowledge is about tacit understanding and relationships between people, then we should not expect to find any simple rational model for knowledge. Can I go home now ?

Pragmatic (Humorous) Philosophy

Every time I mention a philosopher (as I often do on Psybertron) I am reminded of one collaborator (Daniel Rivers-Moore) who confided that he was in fact a philosopher by training, and almost regretted sowing (early phase) Wittgesteinian seeds in the minds of a modelling group. I am also haunted by this witty correspondence of 21 June 2001 from another collaborator (Ian Bailey) on the edi.epistle.framework exploder.

I have to say it’s great fun to get involved in all this cod philosophy again, but it’s also got me worried.

There seems to be a dogged pursuit of the “perfect model”, and philosophy seems to be the branch of “science” that is being used to pursue this. If there’s one thing we can learn from philosophy, it is that there is NO perfect model for the world. Philosophers have been trying to get this stuff right ever since the dinner party was invented. Fashions change regularly in philosophy, and opinion is always divided. It makes for interesting reading, and simply super conversation over a bottle of Petrus in a Hampstead dining room. However, I really have to question its applicability to solving real-world engineering problems.

Set theory is rooted in rather more concrete logic (yes, I know it was a pet subject of quite a few philosphers too). But again, pragmatism is more important than theory to me as an implementor….. What really concerns me is that one of the modelling team is going to buy a new philosphy book to read on his summer holiday and come back and change the whole model again. What will it be this time ?

Albert Camus ? – Things should only be modelled when they have been personally experienced. Anything which has not been experienced by me is irrelevent and so should be handled by exception.

My point is that philosophy is fine, but utterly subjective. We may as well just base our entity diagrams on an artistic movement as trust all our models to philosophy. De Stijl lends itself well to data models, I find.

LOL. De Stijl and Mondriaan were two of the themed meeting rooms in the European conference in the Kurhaus Hotel Den Haag the previous year I seem to recall. Small world.

Anyway, no argument with firm separation between pragmatic implementation and any underlying philosophical basis for a model. Probably where I might not agree is in choosing set theory / concrete logic as the best basis for a model – too rational / objective, even for “concrete” engineering – the ongoing Psybertron agenda.

But, how right he was, one of us did go off and read some new philosophy books, and now look what’s happened ! A good contribution which broke the ice in a modelling disagreement and, for me anyway, proved the value of humour in what on the face of it was a rational / technical debate. One for the “many a true word” thread.

[Post Note May 2012 – prescient on several fronts – philosophical pragmatism, direct (radical) experience and “science” in scare-quotes … re-read following this spoof Email Group IPO in a Daily Mash article posted by the same Mr Bailey via Twitter on LinkedIn.]

Minkow’s Fraudodynamics

I Was a Teenage Fraudster in the Guardian today.
Barry Minkow eventually imprisoned for a major wall street fraud, is “poacher turned gamekeper”, interviewed on BBC Today this morning. Describes small mis-representations leading to larger frauds, and an environment of collusion within business, even including ostensibly “independant” auditors. Makes explicit reference to the big accountancy firms and their independantly named consultancy arms. More evidence of Argyris / DeLorean / Galbraith effect.

Quote […not one of those companies went into business to defraud. Not one. Neither did I. What people don’t realise is that fraud is always a means to an end, never an end in itself. There is always a rationalisation ….] Unquote – That word “rationalisation” again.

Quote […shares soared from an initial offering of $4 to $18. Nobody was in the mood for asking questions – which was unfortunate, because Minkow might have had difficulty answering them …. Live for the now. Very existential. Jean-Paul Sartre. Camus.] Unquote. That’s the first time Sartre and Camus have crept into the philosophical thread.

Quote […auditors …. more criminally gullible than conspiring.] Unquote. Argyris calls it “skilled incompetence”.

Quote […’Things go from order to disorder …. This is one of the laws of what I call fraudodynamics. Let’s say Nick Leeson got the one good trading day he needed to cure the fraud. Well, he’s reinforced subconsciously the idea that he can do it – faced with pressure, he’ll do it again. Do you think Nick Leeson’s first trade was a billion? Of course not.] Unquote. Order, disorder, fraudodynamics – getting close to information as negative entropy view.

In my own experience they don’t start as “frauds” they are well intentioned simplifications to avoid hassle, embarassment, explanatory effort etc, but the effect is that “official” figures never tell the true story. I can think of many an occasion where I’ve had debates along the lines of why don’t we just capture the truth (say in manhour expenditure, or in bid tabulation) and put any allocation / rationalisation / interpretation (even spin if required) in the decision / recommendation / report. Unfortunately when the interpretation is reverse engineered into the numbers, to keep life simple (consistently rational) you no longer have honest information available. The right decision gets made locally, but any aggregate or historical view of the “facts” is in reality now misinformation. In the absence of continuity of people and experience, knowing the true circumstances, it makes knowledge management a very shaky prospect.

Kurzweil – Accelerating Intelligence

Ray Kurzweil (AI) Accelerating Intelligence
Overhyped AI site of “extraordinary”, “revolutionary”, “prophetic vision” (sic), but contributory big thinkers include the likes of Richard Dawkins, Arthur C Clarke, Richard Feynman, Daniel Dennett, Steven Pinker, Marvin Minsky, Roger Schank (of Jorn infamy ?), Murray Gell-Mann, Martin Rees, Roger Penrose, Steve Jones, Paul Davies, Stephen Jay Gould and more. Of course now I remember, I came across this many months ago and didn’t blog it then, but this lot originated with Kurzweil’s Edge / Reality Club / Digerati / Third Culture projects where he exploited the big names for many of whom he’d acted as publishing agent for their popular science culture books.

The hyping is deliberate – to get science into the public eye, to give “intellectuals” the same status as literary and artistic media stars, unfortunately the Third Culture manifesto comes perilously close to say this kind of popular science writing is replacing academic science entirely … “[ writers] in the empirical world who, through their work and expository writing, are taking the place of the traditional intellectual in rendering visible the deeper meanings of our lives, redefining who and what we are.”. I guess its the usual tension in any industry between the Marketing and Production departments – science is a big business.

(Thursday morning update – Spooky coincidence No328 in a series of thousands – BBC Today reports UK House of Commons committee concluding that the way science is taught in schools up to 16 yrs old is uninspiring and needs relating to the newsworthy science issues of the day. See Guardian story above too.)

Site Down – Drama

Site Down – Many Apologies
If you can read this it must be fixed, but as you may be aware my current ISP (ukgateway) was recently acquired by Tiscali. I guess they’re in a process of re-building servers and upgrading / replacing software, but this is the second unannounced outage – the last one a fortnight ago cut off my e-mail for two days too. Expectation is back on line mid-day Thurs 11 July UK time. Fingers crossed.
(Lousy timing as it happens, because I am in the process of migrating to a new domain anyway – watch this space for announcements.)

AI Reboots – Cyc / MIT

Some good links from the AI Reboots – MIT Article already blogged earlier.
(All very MIT-centric of course.)
Cycorp – exploiting Cyc – Foundation for common ssense knowledge (sic)
HAL’s Legacy – oft quoted in comp.ai groups.
Marvin Minsky’s home page “In recent years he has worked chiefly on imparting to machines the human capacity for commonsense reasoning” (I’d forgotten how good MM’s site was.)
And more. Looks like “common sense” needs a distinct new thread.