posted 18 Apr 2005 in Volume 8 Issue 7
Zen and the art of taxonomy maintenance Part IV
A masterclass covering the creation, implementation and maintenance of taxonomies in a corporate context. Part four: implementation and maintenance. By Jan Wyllie.
Not only did these mechanics not find that sheared pin, but it was clearly a mechanic who had sheared it in the first place, by assembling the side cover plate improperly… It occurred to me that there is no manual that deals with the real business of motorcycle maintenance, the most important aspect of all. Caring about what you are doing is considered either unimportant or taken for granted… When you want to hurry, that means you no longer care about it and want to get on to other things. I just want to get at it slowly, but carefully and thoroughly, the attitude I remember was present just before I found the sheared pin. It was the attitude that found it, nothing else.Pirsig, R.M., Zen and the Art of Motorcycle Maintenance (1974)
Now all the creative work – the thinking and the design of an organisational taxonomy – has been done, it is time to undertake what many people regard as the hard and unexciting bit; that is, all the tasks and processes involved in implementation and maintenance. According to this way of thinking, a system needs to be budgeted for, set up and run as quickly and as cheaply as possible. It then must be kept going. It can even be improved using techniques such as monthly updates, form filling and protocols for making decisions.
The great danger with this kind of attitude, which effectively separates people from the processes that constitute their working life, is that it tends to leave out what Pirsig calls “the most important aspect of all”, which is caring. Pirsig’s recommendation in Zen and the Art of Motorcycle Maintenance is that all motorcycle riders should understand how their motorcycle works and how to fix it, and should care about how well it is running. People, such as the narrator’s friend John, who take their motorcycles for granted until something breaks (at which point they hand it over to the all-too-often careless mechanic) not only live dangerously, but also miss out on an essential aspect of the motorcycle experience, the perception of the motorcycle as “a system of concepts worked out in steel” that is brought into its physical being by the human mind.
Riding a system of concepts
An excellent description of a taxonomy would be as ‘a system of concepts worked out in words’. Users of taxonomies are also like motorcycle riders in the sense that the purpose of all three types of taxonomies – information retrieval, collaboration and intelligence – is to take people where they want to go in the virtual worlds of information and knowledge, fast and reliably. If the taxonomy fails to do that, then it is unlikely to ever be successfully implemented, let alone maintained.
If they are to work in any meaningful way, people need to learn to ride taxonomies to their knowledge destinations. In this sense, the quality of the signposting is the most critical factor. Last month, learning to use a taxonomy was compared to learning a language, which, like learning to ride a motorbike, is more about feel and familiarity than about memorising hierarchies of vocabularies and comprehending systems diagrams.
However, the necessary prerequisite for learning a language or how to ride a motorbike is that the person doing the learning needs to really want to do it. The onus is on those responsible for implementing a taxonomy to give them sufficient motivation to make the intellectual effort required of them. There need to be persuasive answers to the two key questions that everyone has: what will it do for me? And, what will it do for the organisation/workgroup?
Communicating the benefits
What will it do for me?
An information-retrieval taxonomy will take you to what you are searching for significantly faster, more intuitively and more reliably than current experience.
An intelligence taxonomy will provide you with what you need to know in a form you can assimilate quickly and easily without you having to ask or search.
A collaboration taxonomy will enable you to both participate in, and keep track of, group work more effectively, saving on work and travel time.
What will it do for the organisation?
An information-retrieval taxonomy will make employees more productive, less frustrated and better informed.
An intelligence taxonomy will give the organisation competitive advantage by alerting it to key developments first.
A collaboration taxonomy will improve the quality of communication between employees, improving decision making and morale, while reducing political and personal conflicts.
On this basis, most potential participants would at least make an effort to learn to use a taxonomy. The answers to these questions are more than goals, objectives or expectations. They are promises. If they are not kept, then people will feel betrayed and resentful, and the cost to the organisation could be considerable.
It is therefore prudent to test taxonomies and their utilisation extensively on closed focus groups before making any promises to anyone. Also, there is no need to promise immediate results. Taxonomies are not magic potions; they take good, old-fashioned work and discipline to use well. It takes practice, and practice takes time. Taxonomies need to be applied with care, and as Pirsig points out, caring cannot be hurried. False deadlines create pressure, while disrupting and delaying completions. Deadlines can work if tasks have been done many times before. They can be fatal when groups are doing something for the first time.
Another advantage to starting with these promises is that they give the whole project clear measures of success. They provide direction, as well as the framework for a feedback system based on how much of the promise is being delivered from both the individual user’s perspective and the organisational perspective.
Assuming that everyone is on board for the project of realising a taxonomy’s promise, the next problem is how to teach both knowledge producers and knowledge users how to work with taxonomies.
Learning the basics
The first requirement is to provide learners with some context and some history, as well as a thorough demystification of the domain using examples of common taxonomy applications, such as the organisation of food and goods in supermarket aisles. Learners must also appreciate taxonomy design principles and techniques, see examples of different taxonomy types, and understand what automated classification systems are actually doing. They should also be told stories about organisational taxonomies in action, and how and why they did or did not live up to their promise. (For an excellent tutorial on this kind of business storytelling, see The Springboard: How Storytelling Ignites Action in Knowledge-Era Organisations by Steve Denning.)
Giving learners a copy of Taxonomies: Frameworks for Corporate Knowledge is also recommended (www.ark-group.com). One of the best features of this report is that its table of contents is also designed as a collaborative taxonomy. Organising e-mail, teleconferences and meetings according to its content framework will demonstrate the advantages of a multi-faceted common focus. People can also begin to add to and amend this taxonomy according to their own purposes, which shows them that taxonomies are malleable.
Providing some DIY experience
Before introducing the taxonomy to be learnt, users should experience taxonomy design and classification for themselves. Exploiting the knowledge that they have assimilated so far, groups of learners should be asked to spend a day developing a taxonomy using tactile, pile-making paper sources. The perfect material for this exercise is a largish pile of newspaper articles, preferably more or less about the same broad subject area, eg, the economy, the environment or management issues.
After deciding the purpose of their taxonomy, the groups should then divide into two. One group should be asked to develop its taxonomy on the basis of team members’ top-down, a priori understanding of the domain and the type of questions they think will need answering. The other group should use the source base to create piles of like-with-like articles. In this instance, the process is empirical; hierarchical structure can be introduced gradually, as desired.
Perhaps over lunch, the a priori and empirical groups can compare their draft outline taxonomies, after which the entire group would adopt a taxonomy considering both viewpoints and then apply it to the pile of articles. The paper experience, using physical file folders, will give learners a direct experience of how taxonomies are socially constructed, negotiated entities representing human perspectives that can be more or less fit to purpose.
Making the process second nature
Now the time has come to introduce the taxonomy. It does not matter whether classification is being done by humans or by software, the best way to learn a classification system is for users to tag documents with the classifications themselves. Start with the highest level classifications. Most people should find it relatively easy to make distinctions at the most general levels. For example, is this item about business or government, or maybe both?
Once the distinctions at the top level have been made, second-level and third-level categories can be seen and understood more easily because they are in context. Remember, user-friendly taxonomies do not list more than seven items per category and should have scope notes and descriptive material at all levels, too.
When learning a language, or a taxonomy, a certain amount of memorisation is required. It is much better to memorise a taxonomy while actually using it, both when classifying items and finding them. Giving people a couple of thousand (or even a couple of hundred) words to memorise up front is no way to treat them. In the beginning, learners will have to go through all the levels (ideally, starting at the top or most general) in order to find the best category. Sometimes they will know the most specific term, but lack knowledge of more general terms and how they fit into the overall structure of the taxonomy. Thesaurus software can be very useful in helping with the human processes of familiarisation and memorisation by making the two-way links between broader and narrower terms.
If automatic classification software is being used, both classifiers and finders need to be aware of the techniques being used. The onus is on vendors to explain the principles of how their software works to users in terms they can understand without resorting to false equivalents, such as the machine showing some kind of human-like, purpose-driven intelligence.
However, the most important category when learning and maintaining a taxonomy is the ‘don’t knows’ file. In a well designed taxonomy, the vast majority of items should fit clearly into categories, according to the agreed set of criteria. The most interesting questions about applying the taxonomy come up in the less obvious examples, which will turn up in the ‘don’t knows’.
Best is better than right
Emphasise that the goal is not to classify into the ‘right’ category, but rather into the ‘best’ category. Learners will quickly see that many items can be placed in more than one category, since documents are likely to be about more than one thing. Even with sophisticated multi-faceted hierarchies, rules must be set about the number of times the same item can be classified in different parts of a taxonomy.
In order to prevent an epidemic of confusion and doubt, learners must be taught to ask questions, such as: is an item more about one thing than about another? How significant is one point being made compared with another? The inclusion of two-way ‘see also’ pointers can be a very powerful tool, giving both classifiers and finders richer options to move across subjects or issues hierarchies, for example.
When people are using a taxonomy, as both classifiers and users, these kinds of questions should show taxonomy working requires both care and time. This series has already dealt with some of the questions about the role of automated classification software, but not even the most extreme vendor hype claims that software cares. It is vitally important that all taxonomy users understand as much as possible what the software is doing, otherwise they a liable to misunderstand the meaning of what has been classified.
The challenge of changes
Most taxonomies must change as the worlds they describe change; some branches wither and need to be pruned as topics disappear from the source base; other branches grow and subdivide. For this reason, taxonomies must be kept under constant review and procedures need to be set up to add, amend, archive and delete terms and branches. It is more like a process of evolution than of maintenance.
Be warned, however. Just because a branch of a taxonomy appears to be withering, it is no guarantee that what it refers to might not reappear with a vengeance later. Delete only with extreme caution.
Finally, the most effective way to grow useful new terms and branches is to continue to encourage the use of the ‘don’t knows’ category, which can also provide a useful focus for group discussion and community building.
More than 5,000 years ago, the Chinese I-Ching (The Book of Changes) described what to do at a time it describes as ‘after completion’ (Hexagram 63):
The transition from the old to the new time is already accomplished. In principle, everything stands systematised, and it is only in regard to details that success is still to be achieved. Everything proceeds as if of its own accord, and this can all too easily tempt us to relax and let things take their course without troubling over details.
When water in a kettle hangs over a fire, the two elements stand in relation and thus generate energy (cf, the production of steam). But the resulting tension demands caution. If the water boils over, the fire is extinguished and its energy is lost. If the heat is too great, the water evaporates into the air. These elements, here brought into relation and thus generating energy, are by nature hostile to each other. Only the most extreme caution can prevent damage. In life there are junctures when all forces are in balance and work in harmony, so that everything seems to be in the best of order. In such times only the sage recognises the moments that bode danger and knows how to banish it by means of timely precautions.
Please send comments and questions to firstname.lastname@example.org.