Ross Mayfield's Weblog

Tagging in the Enterprise

One of the most popular sessions at PC Forum was the Roundtable on User-generated Metadata. Moderated by David Weinberger and Esther Dyson, it engaged the former audience in a conversation.  Good thing too, as many of the experts were in the room.  You can see from my raw notes that it covered the topic widely. 

But the gem was from two comments by JP Rangaswami, CIO of Dresdner Kleinwort Wasserstein.  Not just because he is one of my favorite customers, but its a rare insight into tagging in the enterprise:

Using Socialtext in my bank... Two dimensions that work: people tag things for themselves and whatever I am doing, can I do it in a way that makes collaboration easier. All it is doing is making things easier. The thing that got me on tagging was when I went to Ross and he said it was the simplest thing that could work. I will do things with tags to try to help people remember.  And I don't want to be committed into a structure.

Going to use tags to solve a problem in my organization. People label things differently in different cultures in the same organization. Today English might be the language, but there are perhaps 300 dialects and the labels are different. Tagging lets structure cross-reference, where patterns emerge. Important for me in a commercial context. To let people in different contexts collaborate, keeping it simple and gaining a high adoption rate. Not pushing or pulling, its a community, which is why I like social software.

I helped provide an intro to tagging alongside Caterina Fake from Flickr and Dave Sifry from Technorati -- making points on relative cost, potential scale and social incentives that drive adoption.  But after JP's comments, I knew to shut up and let the customer do the talking.

The other enterprise social software highlight from the show was John Seely Brown and John Hagel finally talking about their forthcoming book:

The notion of productive friction has major implications for IT across multi-tiered process networks. Hagel said the combination of service-oriented architecture (loosely coupled), virtualization and social software (a shared collaborative workspaces like Wikis) is key to developing a work culture that can support productive friction and facilitate conflict resolution, allowing the stakeholders to browse the context to figure out how to "unstuck" an exception (problem).

See Dan Farber's post and the session page on the wiki.

Socialtext Did Not Invent Tagging

Good article in the Guardian about Tagging.  Except this statement:

But, extending an idea developed by SocialText for its wikis, del.icio.us also encourages users to "tag" bookmarks...

This implies that Socialtext invented tagging, which is, uh, categorically false.

Socialtext Categories share similar properties to tags, and was implemented two years ago, but takes a different approach.  Its a flat namespace, anyone can add/delete/modify a category with a simple interface, very wiki-like.  We also added rel tags to Categories to support Technorati's tag search.  You can play with it here.

On the other hand, we did invent metadata ;-)

Searching Wiki Feeds

Tim Oren picks up the RSS deficit in wiki land, via a Google translation of a German post and Dave Johnson post where Scott Rafer comments:

“Much of the work to be done is on the wiki side, unfortunately. Feedster, et al, would be thrilled to make wiki changes as easy to search as everything else, but (…) the Wiki vendors need to make RSS output a standard option”

Much of this thread was started by Jeremy Zawodny's valid complaints about RSS feeds that are barely-human-readable Recent Changes statistics.  He picks on the Channel9 feed, but its a common feature for wikis.

Socialtext was one of the first to provide RSS feeds for Recent Changes (partially because Steve Gillmor was bugging me for them).  We chose RSS 2.0 full text feeds as the first implementation in recognition of how news aggregators were adding track changes, which complements the diff of History when logged into the Workspace.  You can find the same approach with Kwiki, Purple Wiki JSP Wiki and other open source wikis by now.

The problem is in high volume wikis, getting a copy of every changed page is too burdensome, a problem noted by Jeff Nolan (btw, go read his 10 questions to ask a VC).  This part of the reason we offered tightly integrated group weblogs within Socialtext.  Any wiki page can be added to a weblog which has its own RSS feed.  One of our users created a convention called a Track Blog, where instead of flagging or bookmarking things of interest, they add it to their own blog (like a Watchlist) which pings them when there is an update.

The Pull Model of attention management puts the user back in control of what consumes their time.  Email notifications at the interval of their choosing, RSS the subscribe to, and more imporantly, unsubscribe from on their own accord.  To state it once again, RSS is pull, not push.  The model only works when a user can leverage:

• Transparency -- when everything is on a need-to-know and C.Y.A. basis, occupational spam proliferates and social discovery suffers.  When people work openly you can browse the periphery of your attention when its less scarce.
• Amplification -- when other people find something of interest they can edit it or link to it to bring back to top of group mind.  In other words, when you miss something in a first scan, there is a greater chance people will bring it to your attention. First order merits of attention are usually personal, covered by email and IM. Second order merits of attention are more difficult to judge at first pass and are best offloaded to a group.
• Search -- when you have confidence in your ability to recall the past, you can focus on the critical path of the present.

Which brings me back to Scott's comment.  I believe we helped start a general trend for RSS in wikis and this conversation may help raise the bar again.  Even though the vast majority of Socialtext wikis are private (providing private syndication), our handful of public spaces will ping cooperatively (we ping Technorati today). 

Meanwhile, Jimmy Wales and others are working on Wikia, a wiki search engine, and Wikipedia produces a nice diff feed.  Adapting to MediaWiki covers 1/4 of public wikis.  There are well over 100 open source wikis, a wonderful diversity to respect, and search engines would do well to adapt to them over time just as they have with less standard blog implementations.

Tim's basic point was Wikis do not supply contentful RSS feeds.  I'd suggest that blog search engines have had the ethic of just ping us and feed us, we'll do the rest -- which should apply not only to blogs, but wikis and whatever else we dream up.

As almost a side-note, I should mention that the wiki world isn't wild about nofollow for at least one simple reason. On a blog you have an author and the audience (commentators?).  Within a wiki, everyone is an author.  We are still evaluating where we will use nofollow, I personally see it as  great industry cooperation creating a tool to use.

Emergent Intelligence

This post is an excerpt from a longer one on M2M about Technorati Tags, to highlight how thinking is overrated...

There are strong similarities to how wikis and tagging works.  Tagging lowers transaction costs for contributions and fixing mistakes.  This increases participation and the probability of the right data actually existing in the first place.  It also enables a dedicated community to self-govern (and note that as in the case of Wikipedia, the enthusiasm hasn't worn off)

A single tag can be applied in error, and be fixed locally, but that matters less when viewed in the aggregate.  Larger patterns arise that are statistically significant.

The other day I was listening to an interview with Malcom Gladwell about his book Blink, which posits that snap decisions are better than carefully considered judgements.  Especially when made by experts who have developed a muscle memory of the brain.  One of the callers pointed out (at 9:00/30:15) we are better than making snap decisions work better at discrimination (does it belong in the good category or the bad category) between things than characterization (determining the nature of things).  Fine, I thought, that's tagging.

Gladwell's theories seemed to run counter to those of another popular book these days, The Wisdom of Crowds by James Surowiecki, which holds that group decisions are better than those of individual experts.  But not only are these two views complimentary, Surowiecki and Gladwell are having an open conversation about it this week.

So just think about the emergent intelligence mechanism we are creating with a neural network overlaid on the net.  Considered blog posts gain authority through link attention.  Consensual wiki pages gain authority over time.  Links and snapshots bridge across places, physical and virtual. Tags are applied in the blink of an eye and patterns emerge from the crowd.