David Snowden Keynote Tags, Categories and Knowledge Sharing @ KMWorld

November 9, 2007

in Knowledge Innovation

The crowd is assembling to listen to Dave Snowden (A bio here on David Gurteen’s site) on tags and categories. I have a feeling this will be a provocative and thought provoking session. I had the pleasure of being at a bloggers dinner last night organized by Jay Cross which brought two conferences together. A mashup of elearning and KM. It was a great group of people which you can visit here; including Jay, Dave, David, Jon. Dave shared a few comments with me and others on KM. I hope this morning I’ll have an opportunity to capture his vision for how things are being shaken up! Dave also spoke on Tuesday.

Tags, Categories and Knowledge Sharing.

Jane Dysart apparently asked Dave Snowden to be controversial; Starting with four blog posts. These are worth reading and following as background.

1) Paradigm Changes Reference Nick Carr Categories are broken
2) Hubert’s error 10 Feb 2007 A blog is not a publishing tool!
3) Weltanschauung for Social Computing 12 March 2007 – design for ecology
4) Natural numbers networks and communities 31 December 2006.

Close your eyes…. which is the odd one out. How many think cow is the odd one out. Raise hands. Nobody for cows. (no one being the awkward bastard! he says..joking!) Around the world two thirds of the world they will eliminate the chicken… cow has a relationship with grass. He’s trying to illustrate that part of the world is wired to see things first in terms of relationships. While many of us in the western world see things in terms of categories. Social computing has evolved primarily as a social categorization device. We like boxes and little tests. Tribal cultures think in relationships. Basically we must see Social computing being more about relationships than categories. See more:

The Cow, chicken and grass test is a simple one. Nisbett reports that the majority of Americans will choose grass, while Asians choose chicken. The reason suggested is that American’s come from the Aristotelian tradition of categorisation and grass is a vegetable while the other two are animals. In contrast Asians see things in terms of relationships, and the cow has a relationship with grass. Cognitive Edge: East is east and west is west?

He’s saying Corporations haven’t realized the volatility of social environments and are failing to look at them like an ecology. We are also seeing the old patterns and ways of working sustain themselves longer than we expect. Example Process ReEngineering to Six Sigma! He contrasts what happens in the ecology of blogging. I think Dave’s been blogging about 18 months. A few comments captured about what he has learned blogging.

  • If you blog you have to be intimate.
  • You have to say something about yourself. I.
  • Blogging punishes the ego. All bloggers check their technorati score.
  • If you try and impose yourself they stop linking.

Moving right along… brief look at the nature of complexity and the nature of systems..

  • Ordered systems: have repeated relationships they are algorithmic in nature. They are predictive the system constrains the agent.
  • Chaotic system.. lots in independent agents. Then we can use math tools to control and create predictions. No boundaries.
  • Complex Adaptive Systems: The essence of a complex system is the system constrains the agent and the agents constrain the system which creates a highly volatile system. These system constantly adapt. This is not new. This study has shown that simple rules can create highly complex behavior. CAS are are highly sensitive to starting conditions. Like the blogosphere there are principles of locality and constant adaption at work. Thus in the blogosphere people react to proximate relationships, rather than via a wider system. Same thing in a company. Very small things can produce a catastrophic shift a few years later. Complexity theory gives us the basis to understand these conditions.

Distributed cognition.
Looking at a children’s party story. Imagine organizing a party for 12 year old boys. My notes are a little cryptic on the description here. First how will you manage? A chaotic approach buy drugs and alcohol so they can go on a journey and so what if the house is destroyed. A social approach will have a mission statement, a project plan for the party , and clear milestones.. and the senior adults should start the party with a video and then use ppt to demonstrate their personal commitment etc. so it will conclude with an after action review and mandate future….

Let’s approach it as if we create an environment that we know has both good attractors and bad attractors. We can create an environment and see what they play. Then we have to have a scanning capability to see what happens. If you manage an ecology you can amplify and disturb but you most stand aside from the system and watch. You cannot have a fail-safe design approach. You must take a safe-fail approach. This is very similar to serious play and prototype to action and beta type approaches.

Dave shares the magic roundabout story. This isn’t an ordinary roundabout (and totally foreign to Americans) In this picture you see five main arteries coming in and then five mini roundabouts at each entry point. The roundabout serves five main arterial routes. It never ever jams up. Foreigners go round the edge scared to death, the locals go though the middle and shoot straight through. Perfect example of a complex adaptive system. Apparently it took them a few years to get it right. Tweaking it a little at a time. This is the type of environment and behavior we need to create. You can read details of the roundabout in Wikipedia (including the diagram). By contrast he says most companies are a pile up of totally congested traffic signs. A total mess. It’s a wonderful comparison to make; contrasting the command and control / the silos and containers with an ecology (the roundabout) that has adapted.

Dave makes quick reference to David Weinberger’s book “Everything is Miscellaneous“. He reframes it with Everything is Fragmented. There is a consistency of message in both of these. The difference is that Dave Snowden is really focused on how we do the sense-making. He says the human brain doesn’t make decision on the basis of a rational ordered process. People scan at most 2 to 8 percent of what they view and then the first patterns they see they lock into. The brain holds many hundred thousand patterns. Example a radiologists matches against patterns. We evolved to make decisions to form patterns and make rapid decisions. Identified how we deal with danger from lions.. you don’t look at the case studies for how to cope on lions. Dave is really negative on case studies as a method for dumbing down your response and effectively containerizing stupidity. He says we want to recognize patterns and jump to these from fragments. You are wired and want to recognize danger (the lion again) very fast and then run and escape. We evolved to make decisions very quickly on the basis of partial data fragments.

His point of view is information models don’t apply to humans. We are pattern recognition devices. We recognize and act based on partial data sets. We know that this is often true. We basically find that people like environments that are fragmented. Context and data is increasingly fragmented. We like google for just that reason. In fact we like blogs for much the same reason. It is both ordered and chaotic. The blogosphere is an ecology that that thrives on fragments. Personally, it’s one reason I believe blogs aid recognition or the stimulation and creation of emergent patterns. My own observations suggest that the blogosphere learns faster.

So if your model for your organization is to take things into a structured case study then you will find no one wants to look at a number of cases. People will go to the fragmented stories. The problem is the immediate moment in which you codify something, you make it relevant to the time. This is very consistent with the data presented yesterday in the Parc session on tagging. Containers and cases lose their context; just like tags depreciate with time. Today you fine the oral histories have more interest to you.

The lesson here is if I go for low levels of fragmentation and high level of context interdependence then it’s hard to create much value or stimulate information / knowledge creation. It simply takes too much effort. I think this is the same as saying the codification model is dead. Oh and the replicative model…. create a knowledge object.. well by the time you have created it… it too is obsolete.

So if I go to extreme fragmentation and a loose adaptive contextual framework then I can create a more flexible adaptive system… Obviously, a system for pattern recognition. It will also create a more flexible cost system. Unfortunately, what happens is management likes to move to the left to reduce ambiguity. Great conceptual chart. It illustrates how traditional management information systems drive creativity and adaptability out of the organization. It’s no wonder then that more and more people are forced to DIY their tools in order to make progress. And the organization is thus threatened by this. Letting go has never been more important.

Approaches to metadata.

  • Traditional: Professional or author created. First agree all the key words. Remove ambiguity. Key word search. In formal work groups this is too static.
  • User Created: keyword tagging. most are really cool but like Newtonian physics they work well within boundaries. language changes subtle in context. anything that you search is useful within boundaries. In effect we have a flat structure without grammar. It is too hard to keep up or be consistent with tags and categories. We are moving away from taxonomic search to serendipity. “The tale of two princes of serendipity… each time they found something new… In an uncertain world you want serendipity not structure. Again this point was substantiated yesterday by the Parc guys via their research.

Currently we have scalability problems. As tags migrating from techies and early adopters there is too much ambiguity in the use of tags. Context depreciates. We are already approaching the time where we need something new.

New: A semistructured tagging approach.
Requires a language of coherence that is neither structured nor fragmented. We’ve found that humans need to understand why. If we introduce more than five factors they wouldn’t buy the forecast. Part of our wiring means we have to have some understanding, some sense of scale. This requires some sort of structure without going into rigid taxonomies. What we are trying to do is embed fragments into cognitive models so we can replicate how humans create constructs. We’ve learned that the names (tags given) are often more significant that the content. In fact the words used rarely appear in the content. names, options, key words, free text , filters. when asked.. key words.. aren’t in original content. What we do is add additional meaning to the content. After free text we use filters to scale using natural English language. The result provides a way of handling ambiguous structures.

Dave Snowden’s Advice and the Future

  • Imagine you had 10k stories coming in each day around your operation. Now you see a landscape and how the stories are coming in. Then you dive into areas of interest. You are now scanning the system, you are thinking about how to disturb the system.
  • The death knell or the redefining point for KM. You have to think differently. Don’t go and create a portal or COP. There are no recipes. If there are any successes they happened by accident not by design.

The recipe (which I’d say is not new / been arguing elements for years which does work)

  • Install blogs: Install wikis. Use MediaWiki (wikipedia has trained them already, fastest adoption,(good advice!)) then get key personnel who influence people. Then set in motion a blogging program, sit with them every friday and build a habit. If hou can habituate people then it starts to spread.
  • Then based on uses.. add utilities, html training, common co document system, encourage links etc. Find a project, start with blogs and then consolidate to wiki and then publish. give them one month of free blogging and then take the content and preload the wiki… it is not given. Giving them a wiki upfront will provide no structure, easier to move them across when it is populated people so can see how it is amended. Start using narrative, links and connections.
  • Stimulate networks, use natural numbers 5, 15, 150. Social network stimulation. Objective everyone to 2 or 3 degrees… tightly connected. At this point KM will look after itself. It is not hard to get this type of symbiosis.
  • More ways to get radical.. Ban attachments on email. There is no reason for attachments if you have hotlinks. This creates a barrier that you can enforce. Then use search engines… Yes ban attachments and your security can go up and your firewall can be more open. No reason to have anything collaborative that happens within the firewall.
  • Concurrently free the masses… let people do what they want. . respond to novelty… they won’t all use the same colalboration software.

Three Heuristics he began working with a few years ago:

  1. knowledge can only ever be volunteered not conscripted
  2. we only need to know what we need to know when we know it.
  3. we always know more than we can say and we will always say more than we can write down.

Closing type of message / sense-making:
Knowledge about the way the brain makes decisions. The brains makes sense based on fragments.. Blogs are closer to the way are brains are supposed to work. Ergo! If you don’t have a blog strategy you don’t have a chance at a knowledge strategy. If you don’t have a knowledge strategy you don’t stand a chance of getting out of the cell you have constrained yourself to!

Lots of great themes here. It was good to effectively hear Dave’s stuff twice. It all boils down to common sense!… Now is that pattern recognition or not?

Previous post:

Next post: