Del.icio.us inside discussion at Mindcamp

Alex Barnett (the only other European at Mindcamp besides me as far as I know) organized with two other guys (sorry but I did not catch their names but I think they were students doing a project on this) a session to discuss why corporations are not tagging their internal content (hence the del.icio.us inside title of the session meaning inside the firewall) as an alternative to the traditional hierarchically organized Intranets for finding corporate content. I argued that this is already done to some extent since the addition of keywording as metadata is something that all enterprise content management systems lets you do and is basically tagging. I got the impression that my suggestion that tagging was not that revolutionary as Web2.0 people tend to think was not very popular. One guy that happened to be standing next to me supported the idea in the context of tagging images in a photo library which I brought up as an example of a corporate application of tagging that is currently widely used (he worked for a stock photography company and he seemed to be very well aware of the issues as I am since one of our company products is a very broadly used digital asset management system).
I think that what makes del.cio.us revolutionary is the consumer aspect and how by having so many people tagging the same content (basically, the same web sites), the most commonly used tags emerge as the best way to describe that content and become their defacto metadata. The main difference is that in corporations, only few people are typically allowed to tag content and add metadata to it so that "wisdom of the crowds" effect never appears. Could ECM companies change that? Yes, very easily, just make the metadata editable so that every time you are finding a document or posting a document into the corporate repository you can add as many keywords to it (or tags as it is now prefered) and that effectively becomes del.icio.us. Would corporations want that? Most likely no, it is not clear that enough people will actually do that for the tags to become more useful than the officially sanctioned ones. Someone suggested that companies should have some sort of reward or micropayments system to incentive employees to do so. That guy obviously does not run a company, I highly question the extra benefit of finding things more efficiently against the extra cost. You want to reward employees to do their job and, for the majority of employees, that is not improving the metadata of corporate documents so you do not want them to spend time doing that. For corporations that finding content across the organization is very critical (like a consulting company), there are already solutions there like Autonomy's software that basically tries to automatically extract concepts from documents across an enterprise so that you do not have to rely on user created metadata to find relevant content in a corporation.

Some people argued that del.cio.us was not going to be used in corporations because people do not understand it. Again, I highly question that, metadata has been around for tons of years and it is a pretty selfexplanatory concept (metadata, data about the data, get it?), I will agree that people do not understand del.icio.us the first time simply because their Web UI sucks.

Dave Winer was there with a "you kids do not understand anything, I blogged for 9 years and RSS already solves this problem" attitude that did not really contribute much to the discussion. I do not see how RSS helps adding the proper metadata into corporate content.

One funny remark came up from an Amazon employee that said the real problem that Amazon had categorizing their content was that "Amazon has a lot of books". No wonder, weren't they called "the worlds largest bookstore" at some point?

CD

Technorati Tags: mindcamp2.0

Comments

[Michael Braly]

Hey Carlos -- nice summary of the discussion. The "two other guys" are Geoff Froh and Michael Braly.

[Steve Eisner]

Hi Carlos -

I was sitting next to you in this talk and I made a lot of the same observations you did (though I think you are underestimating the value of RSS in the enterprise, not as a way to create metadata but as an end goal which could obviate metadata. More later...) Anyway, my impression from your comments was that you obviously "get" this stuff and I'd be interested in hearing more about what you had hoped to hear & say at this session!

I'm going to be posting my own thoughts on "delicious inside" as soon as I can get them typed up. I'm hoping we(the whole group, whoever is interested) can continue the conversation without the 45-minute time limit! I'll ping you to let you know then...

BTW a quick side note: You made a statement during the session comparing CMS and tagged-link environments. At the time we were talking about the "ACLs" problem where people may want to reduce the subset of people who can view a particular item. Although you're right that basically we've always had ways to create repositories of (links/articles/files/whatever) there is definitely something new about tags with regards to permissions. If you think of tags as "virtual folders", which I believe may be the easiest way to explain them to non-taggers, then the problem becomes more apparent. People tend to want to associate permissions not with individual files, but with the folders that contain them: "this folder is my private stuff, this folder is my project stuff". When you can have content associated with multiple folders it becomes very hard to give the user a good UI for this. It's a problem we've struggled with through several versions of our product, which is basically a CMS with tagging ;)

Steve

[Geoff Froh]

Hi Carlos,

Good comments on the session. And good comments *during* the session as well.

I also wish we had more time for the discussion. It was my first time at an unconference, and while the lack of structure was great I sometimes found the conversations losing out to the clock. Ah well...

I entirely agree with your assertion that free keywording is not a new practice in metadata. A seasoned LIS hand pointed out to me recently that even back in the dark ages Dialog used to allow their indexers to apply non-standard terms in a special field when they were categorizing content. They would then collect and analyze the shared keyword pool on a regular basis as a kind of "user warrant" for changes to their massive controlled vocabulary.

I believe the core difference between the practice of keywording and the kind of tagging that happens in del.icio.us is the concept of pivoting. As a user I can give something a label that is meaningful to me *and* I can see that you are using that same label. I can then pivot my viewing context to see what other things you might have tagged and what other kinds of tags you are using. As I examine this information, I begin to develop an idea of our relative affinity. In environments like del.icio.us, there are three relationship networks superimposed on top of each other -- the urls, the tags and the people. This can produce some powerful network phenomena.

As for incenting the users to tag, I am also in total agreement with you. Paying people has already been shown to produce horrible metadata; there's no reason to think it would work any differently for tags.

The key is making it something users already do. That is why I really believe something like del.icio.us might work quite well within an enterprise. People understand bookmarking. Most people bookmark, and (in our field research) they have difficulty managing them. Users don't have to understand the tripartite network affinity models or Lakoff base-level categories to get bookmarking with del.icio.us.

Of course, not everyone is going to adopt a new tool as readily. The big question is the one you asked -- what is the minimum threshold for interesting network effects to occur? What's the tipping point?

I don't think there's anyone who can say definitively at this point, and in fact, it probably depends profoundly on the context -- the type/size of the organization, the number of resources to tag, the complexity of the semantic domain, etc.

That is why a del.icio.us style system absolutely should not be the sole method supporting findability in an information environment. It is another component, along with full-text search and a well-designed taxonomy.

Anyhow, I'd be interested in continuing the discussion with you and anyone else in a mindcamp redux over some beers...

Geoff