First, I have to get this off my chest, IBM's user experience group is going to be experimenting with folksonomy in parts of IBM's intranet. It's very exciting because they are going to be pursuing folksonomy as user data, an approach that just seems to make so much sense for a corporate environment.
This news flash comes from a fascinating talk I attended at the IA Summit today on the use of taxonomy in IBM's intranet. The scale of this intranet is remarkable. It serves 315,000 IBM employees worldwide in multiple languages with personalization by business role and interest among other facets. Even more remarkably, they have used a controlled taxonomy, one version of which contains 3700 nodes, to organize the information in this intranet.
To understand why IBM might consider investigating folksonomy, you need to understand the role of taxonomy in their intranet and how that role is affected by the scale of their operation. My conversation with the IBM folks after the talk was the one of the most fascinating and revealing I have ever had in this regard. I was impressed by the group and their dauntless pursuit of solutions, and I thought it might be worthwhile to spend the rest of this post giving them some free advice on how they might use folksonomy. It will be interesting to hear what they actually do.
The nutshell version of how IBM uses taxonomy
In a nutshell, IBM uses taxonomy to represent the knowledge contained in the intranet. In some sense, you might think of it as being the official vocabulary for talking about things at IBM. Users learn the vocabulary, and then the vocabulary allows them to access relevant knowledge on the intranet. Now, if you consider that the average person's active vocabulary (i.e., words that they are likely to use in daily conversation) typically consists of only a few thousand words, you can see that learning a 3700 node taxonomy could be like acquiring a whole new active vocabulary, a daunting task for users.
That's one area where folksonomy might eventually help. If people classify items on the intranet with their own free tags in some sort of internal bookmarking system, it would be possible to map between the user's personal vocabulary and the official vocabulary, allowing the user to eventually search with their own vocabulary.
What's in it for users?
How would IBM encourage users to do free taging? In other words, what would be in it for them? Well, for one, the motivation might just be intrinsic. People have been bookmarking for years and trying to organize those bookmarks for about the same amount of time. Most bookmark organizing systems are failures because they are browser-specific. If IBM could offer a non browser-specific bookmarking system, somehow contained in the intranet itself, then this might be enough of a win to get people to use it. Based on what some of the IBMers described of their intranet participation, people do seem to be willing to put in this level of effort because the intranet is already such a valuable resource to them.
The key point is that the bookmarking system be easy and offer high value. One example system IBM might consider in building their system is Firefox plus some of its delicious extensions like foxylicious. Or, they could consider the incredible Mac application cocoalicious which allows easy posting to delicious, full searching of bookmarked items, and faceted tag browsing.
Additionally, as I discussed here, making the bookmarks visible and getting managers to actively promote using them may also help get good participation, even absent financial incentives. Managers who have funding responsibility for this experiment will be motivated to encourage use, so getting their active help may not be too hard either.
What's in it for taxonomists?
What's most interesting in the IBM example is that the folksonomy experiment is originating from the people who control the vocabulary. Reading between the lines of their presentation with some confirmation from our after the talk discussion, maintaining the taxonomy in light of changing business conditions is a daunting task. Anyone inside of a large corporation knows that new business concepts, strategies, product categories, and products are constantly popping up. Business divisions and product lines are also eliminated and merged.
Can taxonomists use data from people in the field in some systematic way to help update their centralized view? Quite possibly, particularly in the area of new innovation, but I think this is generally a more problematic area. Corporate communications is not just pulling up user views, it is also telling users about things they may not already have a notion of. In this case, they have to be pushing out a way of viewing the world. In my view, this is one of the major reasons for having a controlled vocabulary that has been centrally decided. One solution might be to reconsider the internal communication function as one of creating links between the different divisions and effectively propagating new conceptualizations from the divisions where they originate to other divisions in the organization. Of course, all of this needs to occur with management's oversight. Management would suddenly not be in the business of creating vocabularies, perhaps a relief. But, they would need to decide which emergent ones to encourage, a new challenge, a task that is more like managing the flow of change rather than defining the definitive vocabulary of change.
Relating this to the larger folksonomy debate
Clay Shirky recently publicized an essay on folksonomy by Matthew Locke in which Locke asserted that folksonomy would only be useful when precise communication did not matter. My cut is that Locke's statement simply indicates that he was neglecting problems at the IBM scale. It is vital that people in IBM communicate, and they all use the intranet. Therefore, it is vital that they be able to use the vocabulary. But, the vocabulary is, in cases, too vast to be easily learned, particularly if one is forced to wrangle with all 3700 nodes.
Second, things are too rapidly changing for a centralized system to keep pace.
To sum up, unless, as in the medical profession, people have years to spend acquiring a controlled vocabulary, some systematic means must be provided for mapping between their vocabulary and the standard. Folksonomy seems to provide one promising data collection method for doing so. Second, in changing domains, some means must be provided for promulgating change across units that does not bog down in centralized, bureaucratic choke points. Again, folksonomy may provide a means toward one such mechanism. Even with management continuing to play a role of deciding what to promote or not, they are in essence guiding change that is already occurring. This should ease maintenance but perhaps continue to raise governance issues.
[Update: The IBM team has graciously made the slides of their talk available here.]
What a fantastic example of the difficulties with maintaining a controlled vocabulary. Thanks for pointing me over to it. You made two very interesting points: one is that we need to ask what's in it for users, and the other is the distinction between industries in which it is absolutely necessary to learn a CV vs. an example of where we don't want that to happen.
Do you know if any of the IBM folks have written up their experience? How about slides from the IA summit?
Josh, thanks for the note. I am following up with the IBM folks about the slides and will post updates here. It could take a while as they are a large company and will likely need approvals.
I should also note that their efforts with folksonomy are just getting under way, and that part will take some time to report.
A nice structured point about an advanced and corporate meaningfull use of folksonomies.
I couldn't follow the IA Summit. Thank you for letting us know this fascinating initiative from IBM and please go on talking about updates in the next days.
I've quoted you at my blog http://www.infospaces.it/wordpress.
Bye
The Locke reference is interesting here, about folksonomies only being useful when precise communication doesn't matter. Think it depends upon who the consumer is of the tags. If its the author only, where you tag to enable yourself to find stuff, then what works for you works for you and precision is in the eye of the beholder. But when you tag for others to consume then the importance of precision grows, whether its via folksonomy or a formally defined tax.
Wow, clearing out spam, I erased my own comment. At any rate, we are currently running experiments in social tagging where we have discovered that people are using tags to label conversations. It's not entirely clear that people are trying to achieve some sort of semantic precision or just mark the conversation for future reference.