Interesting conversations in Washington a week ago, some traffic on ActKM and the odd email brought the question of national KM (especially in the context of security) to mind late last night and I resolved to blog it this morning, although it will be in two parts. I have a minor paid advisory role in respect of the subject by the way to declare an interest of sorts. One of the main triggers to write the news of a well intentioned but sadly misguided proposal which has outlined a prescription for a consistent, centralized application of KM across US Federal government. In part the justification for this is a running claim that failures such as 9/11 and Katrina are in effect failures of knowledge management. Words like standards are also being bandied about which is always a bad sign and one gets a very real sniff of a bureaucracy in the worst sense of the word and some nice comfortable jobs that will up people's profile on the conference circuit.
Now I remember when the 9-11 Commission Report came out we had similar talk and it all involved re-organisations, data sharing etc. all based on the benefits of hindsight. When you look backwards its fairly easy to see what should have been done, but as I have said many times hindsight does not lead to foresight. Now I remember a conference call at the time the report was published in connection with out DARPA work and several of us had noticed that the conclusions of the Commission Report were similar to those of the Challenger Disaster (1986), namely a failure of knowledge (well really information) management. However the report on the Columbia Disaster (2003) basically said that all of the right information was in front of the right people, put they simply didn't pay attention to it. The buzz on the conference call was that the same mistakes, based on the same false & idealistic assumptions about the capability of knowledge management, were (and have been) made post 9-11 as were made post Challenger.
In effect the argument, which is common one in knowledge management, was that the failure was one of not connecting the dots, not realising the significance of key data items early enough. The idea is that we create bigger and bigger databases with more search algorithms, centralise functions, standardise procedures, appoint an obergruppenführer and somehow or other no future errors will be made. Now anyone who argues that dots can be joined up in a human system in this way is either a poor mathematician or has failed to read the right stories when young, Let me share with you the story of the Emperor's Chess Board, courtesy of Eurekas and Euphorias.
According to legend, a Chinese Emperor asked a sage what reward he would require in return for an important service. The sage named his price: nothing more than some rice, two grains to be placed on the first square of a chessboard, four on the second, eight on the third, and so on. A modest demand, the Emperor thought, and happily agreed; but he had failed to grasp the principle of geometric progressions. The entire rice crop of the empire would have had to go on a single square, long before the sixty fourth was reached.
The actual number if you want it is 2⊥64/1 1.844674E19 and the same point is made well by Boisot who makes it more explicit in this paper. He points out that four dots have six linkages, which means a total of sixty four patterns that can form. ten dots gives 3.5 trillion and twelve 4,700 quadrillion. How many dots are there in a human system? How many possible patterns? OK hindsight is a wonderful thing as the significant dots and the linkages are now visible, but in respect of foresight? Forget it. Remember Lincoln? He said in his second annual message to Congress in 1862 As our case is new, so we must think anew, and act anew. We must disenthrall ourselves, and then we shall save our country. Never a truer word more greatly ignored in government and most certainly ignored by those presenting the same old tired solutions of centralised knowledge management?
So what should we do? Well I will address that in part II tomorrow
Comments (6)
This is terrific stuff. That chessboard analogy is really helpful in fingering the essence of the mistake being made.
I'd add that KM is doing what so many other industries do: frame everything in their own narcissitic way. I daresay there are people in the marketing world dining out on their explanation that 9/11 was a failure of branding somewhere or other.
Posted by Johnnie Moore | August 3, 2008 10:42 AM
Posted on August 3, 2008 10:42
That seems a little confused to me - both in that "An arithmetic increase in the number of “dots” to play with — high quality or otherwise — leads to a geometric increase in the possible connections or links that one can establish between them and to an exponential increase in the number of patterns that can be generated from connected dots." There is a quadratic increase in the number of dots - n(n-1)/2 - and 'geometric' and 'exponential' are synonymous. Secondly, from an engineering perspective, once the number of combinations exceeds the sample size it's not the limiting factor (in effect you shift from radix sorting to one of the O(log N) methods), as you compare data to each other. Most large data sets are also sparse, so once you're working on something like the netflix prize set (a bipartite graph with 480189 and 17770 nodes), you compare 480189 vectors each 17770 values long to each other rather than all possible graphs - (6 pow 17770) pow 480189 (connections are weighted with an integer from 1 to 5). The simpler case is still a big enough number to make the problem interesting, but the numbers for the number of possible patterns are no longer relevant.
However, the higher level argument holds - simply getting more data to compare doesn't automatically relate to making useful comparisons, and there's little hope getting good predictions on something as important as security when we can't even do something constrained such as movie recommendations accurately.
Posted by Pete Kirkham | August 3, 2008 1:43 PM
Posted on August 3, 2008 13:43
Well, of course the dots cannot be joined meaningfully by some mechanical automatic process -- anymore than 6 numbers between 1 and 45 can be chosen to win a lottery, except by chance. What is needed is well constructed hypotheticals that, by way of story (either of history or future conjecture), reduce variety and help sensitize the mind to the possibilities of emerging phenomena and their weak signals. Then these patterns may be more, or less, recognised as significant. The last decade of crazy 'dot-joining' has been proved a disaster in the hands of incompetent millennial dreamers. Using the "what if ..." effectively is an art (and perhaps a science) that few appreciate and less practice ethically. One needs a grand hypothesis, or three, to start with. A meta-systemic view perhaps. Some call these scenarios, but scenarios are more likely to be path-ways within these broader schema in my opinion. The ways of thinking (of your warrior friends) are far too culturally bound and systemically limited to see beyond or through their own neurotic fears and anxieties. Monty Python still comes to my mind when I see or read of these types of situation-reactions listed. Sorrow and outrage when I see how many suffer and die -- especially when it has human causes. No super computer is ever going to work through to what an enlightened mind can see. But developing enlightened minds does not appear to be high on the political priority list these days -- and, I would assume, they certainly would be very anxious about working for an insane industrial-military complex under the influence of strong eschatological thinking and fantasies. Of course, under these conditions they will 'see' what they fixate on and miss the more general and tidal influences occurring and emerging in the real world. And of course, in the last decade we have heard little except various misleading propaganda about the ‘war on whatever …’ – perhaps it’s time for some new thinking about the ways of peace. Could the recent signal from the 2nd most influential school of fundamentalist thinking, located in India, banning terrorism acts that affect the innocent (from a school reportedly more fundamentalist than the Saudi Wahabbi sect) have a more profound effect on this trend towards peace than all the ‘dot-joiners’ and their ink-blot tests in London and Washington combined? Simple really, when you think about it.(*) That is smart knowledge management and governance in my opinion: but there is no oil business in it ...
(* ref: Fareed Zakaria on LNL, http://www.abc.net.au/rn/latenightlive/stories/2008/2320605.htm )
[Of course I assume this will not be published. It's late Sunday night and I'm back in Perth, after 4 weeks holiday in the UK and Canada, and now trying to constrain the brain for a 'day-in-the-office' tomorrow.]
Posted by rc | August 3, 2008 2:30 PM
Posted on August 3, 2008 14:30
Ah, the knee-jerk reactions of the desperate. You can't repeat "Hindsight does not lead to foresight" enough - maybe carve it in stone over their door ?
Posted by Cheryl | August 3, 2008 8:28 PM
Posted on August 3, 2008 20:28
Great stuff! The Boisot article you linked to the word
this
had an error in the Number of Possible Patterns column that may confuse readers. It says p = 2L, when it should be p = 2^L. I revised the article will send you the pdf.Best,
Mark
Posted by Mark White | August 8, 2008 4:45 AM
Posted on August 8, 2008 04:45
You have hit the nail on the head ... i can now articulate why i have long believed that there should not be any pre-defined or best practice paths for knowledge flow! :-)
Posted by Atul Rai | August 11, 2008 6:48 AM
Posted on August 11, 2008 06:48