The Resilience of Information Architecture

At the IA Summit, Grant Campbell and Karl Fast presented "From Pace Layering to Resilience Theory: The Complex Implications of Tagging for Information Architecture." I highly recommend both the paper and the slides, and if you're really into resilience, check out Ecology and Society.

Of course, that doesn't mean I totally agree with Grant and Karl. First, I don't agree that information architects have ignored tags. To the contrary, IAs are infatuated with collaborative tagging. We simply haven't found many opportunities (yet) to integrate tags into our practice.

Second, the authors commit the same sin as Clay Shirky and David Weinberger by ignoring context. Tagging has flourished in the free-wheeling blogosphere, but has had little traction in the realm of corporate and government web sites where authority is an equal partner to findability. It remains unclear when and whether tags belong in the information ecologies in which IAs typically practice.

That said, I've begun a Web 2.0 consulting project in which I'll get to integrate tags and taxonomies, so I'm on the hunt for inspiration. Amazon's experiment with tags seems to have failed. Has collaborative tagging ever been successfully integrated into the web site of an established institution? How about examples of sites (even small ones) that elegantly combine tags and taxonomies? Thanks!

The presentation that Karen Loasby of the BBC gave at the 2005 IA Summit comes to mind. But that was for an intranet and I can't quite remember if the tags the journalists applied were only subsumed selectively by the CV, or if they were also used on their own for navigation. (And I can't seem to open her ppt.)
http://www.iasummit.org/2005/conferencedescrip.htm#66

Well I'm not sure I totally agree with it either. That's why the presentation has a fourth item at the end, about resilience theory, which asks "but what if we're wrong?" Also the talk argues that a major value of tagging is how it can help us reconceptualize IA; It's not just another item in the toolkit ((the talk argues this better than the paper).

Your comments are good, though I have responses to both (of course!)

We are not arguing that IAs have ignored tags (though it would be easy to get this impression from the paper). Rather, we are pointing out two basic responses to tagging: tagging is progress and tagging is regress, each of which has both moderate and extreme variations.

IAs were initially on the regress side, for the most part. But the success of delicious and flickr has made it impossible to wholly dismiss collaborative tagging. So it's not surprising that many IAs have embraced tagging, albeit with some pragamatic skepticism. I think it's important to appreciate this skepticism, for it balances the open enthusiasm held by those who have long held more optimistic and occassionally utopian opinions of tagging and folksonomies and so forth.

Enter pace layering, which has been used (notably by you, in Ambient Findability) to incorporate tagging into the conceptual framework for IA. But the way this has been done implicitly argues that the work IAs do, the kinds of rich structures we create, is still deeper, richer, more important, and of lasting value. Tags are great, this says, but they're just not as good as CVs in the long run, and never will be.

We are not arguing that IAs have ignored tags, but that they have recognized the value of tagging in a safe, comfortable, and somewhat patronizing way--a way that doesn't upset their worldview. This view may be *entirely* correct. Or not. Either way, tagging deserves a closer look. Right now we have lots of opinions and some anecdotal evidence, but not much deep understanding. So we may be overlooking important things in much the same way that when information retrieval was first applied to the Web it overlooked the assumption that user-supplied metadata would be trustworthy, or that hyperlinks had little value for relevance ranking.

The paper goes on to explore complexity and resilience theory, from which we argue that tagging has complex implications which a simple pace layering model does not reveal.

As to ignoring context, we most definitely do ignore it. But is this a "sin?" While recognizing and understanding context is often important there are many cases where it is not.

Ignoring context is vital to developing a scientific understanding of many phenomena. Take baseball. If we want to understand the physics of baseball--how long the ball will stay in the air, where it will land, and so forth--we *must* ignore context. We should not worry about the rules of the game, the price of hotdogs, or the third baseman's salary. These are immaterial to throwing a baseball. Moreover, only by ignoring context will we be able to extend our newfound understanding to other realms, like throwing a football or a cabbage.

In a sense, the baseball problem does involve context. For example, we need to understand gravity, which is different on earth than it is on the moon. We might also need to understand wind, if there is a lot of it. And if the ball is particularly large and we are able to throw it at a high speed, we may need to know about drag and coefficients of friction. These factors constitute what we might call the physical context.

This is not what most people mean when they ask about "context". They usually mean "what about the social, political, and cultural contexts?" They use the term as shorthand. It's critical to know the relevant context(s) for a particular problem. If we are studying baseball as a physical process, we must address the physical context. If studying baseball as a cultural phenomenon, we must obviously address the cultural and social contexts. Get the context wrong and you won't get very far.

Our paper treats tagging like baseball, more or less. We look to complexity theory and resilience theory as tools for understanding tagging, independent of social and cultural contexts, be it the free-wheeling blogosphere or the realm of corporate and government Web sites. Obviously this can only take us so far, for tagging is intensely social, which is probably what distinguishes it from classical free-form indexing. But there are times when you need to deliberately ignore some complexity to understand the fundamental elements.

So yes, we ignore context, but out of necessity because of the questions we are asking.

Thanks for such a thoughtful response Karl!

There's also been an interesting discussion in response to my two qustions over on the iai-members list:

http://lists.iainstitute.org/private.cgi/iai-members-iainstitute.org/2006-April/000773.html
(members only)

Since lots of smart stuff has already been said, I'll simply offer up a few final reflections:

* I understand the value of ignoring context, but in this case it's problematic, because user-generated tagging is so unproven outside a few very popular examples.

* It will be interesting a year from now to see how the intranet experiments with tagging have worked out.

* I like Dan Klyn's points about the value and feasibility of admin- and merchant-assigned tags in an e-commerce environment.

* Both Rashmi and Karl point to the social nature of tags. It will be interesting to see if social search that draws upon social networks and tags improves relevance and/or user satisfaction.

* Adding resilience theory to the pace-layering model will enrich our thinking about IA and UX in general (with or without tags).

Was just thinking...why is it that people tend to point to the success of del.icio.us and make an association to tagging? The intent of del.icio.us was for bookmarking. As these intents are different, but related, how can you qualtify the relative contribution of tagging to this success?

I'm not a good person to answer that question, since I rarely use de.licio.us, but it does seem that the tags are central to navigation and overall success.

I think that Paula has pointed out a very nice question.

My experience as del.icio.us user is totally bound to the need of availability of my bookmarks on every physical device I use. If I see a good article about a css technique, I just post it to del.icio.us tagging it the way I feel it is. However, this action usually follows the reading of a mere couple of paragraphs at the beginning of the article. What happens is that even if I don't know the purpose of the article I will give it a tag. Let's give an example: my article it's a step by step tutorial about styling a list with css. After I have scrolled the first two or three steps of my article, I receive a phone call and I have to postpone the reading of the article. At this point I post it to del.icio.us but I have to give it a tag. Which one?

"CSS" might be too wide in meaning; at first sight I should say "CSS styled lists", but what if, by the end of the article, the author gives a nice appendix with all the hacks for Internet Explorer? I didn't know by the time I posted the article to del.icio.us and the "IE CSS hacks" tag should be added to my post. But will I be that honest to expand the tag list AFTER I used that article?

Therefore, as long as tagging is a popular technique and social software has declared it a sort of "MUST HAVE" feature of a site, when it comes to IA, we have to consider how much we trust information sources. I believe that one of the reason of del.icio.us success is that is a sort of directory of what's new on the web. Uhm, I think that I've already read this somewhere else... isn't it the purpose of any generalist portals? Only, contents are provided from the top (the user) to the base (the collection).

Opposite to the old school portal, where the taxonomy of contents was decided by the workgroup, the base owner, the TAGSonomy, or folksonomy comes from the user. But this leads us back to the nature of the tag itself: I might provide a generic tag that yes, describes my content, but doesn't qualify it in a more specific way. By the way, I believe that this is the kind of tag that a "bookmarker" like I am would attach to a content.

Or I could post a wrong tag, a tag that doesn't describes one of the hottest feature of the article.

I tend to believe that the popularity of tags is due to the quality of the content itself: even if I simply file a content under the generic "CSS" tag, this content will enrich the general user expectation that the web is useful, and the tag will be considered useful.

Del.icio.us as personal portable link storage is just one of its beauties.

It's success, I think, is because it addresses the problem of credible sources. There is so much information out on the web, how do we determine what is good quality and what is not?

If I'm looking for "navigation design" for instance, I can type these keywords into del.icio.us, and immediately get a good idea of quality sources because of the number of people who have bookmarked a given link in their del.icio.us profile. Hey presto, instant peer review.

This also feeds into the idea of emergent navigation (a la flickr).

Peter, if you're after a site that integrates tags and taxonomies, check out http://www.actnow.com.au and also the IA's discussion of it here:
http://www.disambiguity.com/2006/05/actnow-launches-20-for-good-not-evil/

It's not pure tagging, but its close.

ps forgot to mention that, of course, the bookmarks one finds when searching for "navigation design" have also been tagged by others with "navigation" or "navigation-design". Del.icio.us supports semantics rather than more librarian-like classification. You could argue that this is a weakness, but it is also its strength...

Thanks for the pointer to the Act Now site which is very interesting. Of course, they're not really tags, but controlled vocabulary terms masquerading as tags in a tag cloud...which is wonderfully devious.

Flickr has succeeded with their tagging system, don't you think?
Here is an interesting link to articles about it:
http://www.flickr.com/press.gne