Folksonomy Folktales 2010

3 – lack of a central authority

Being out from under the thumb of those dictatorial librarians seems to be something that folksonomy advocates very devoutly wish for.  My guess is that it has more to do with ideology than any actual tagging and/or searching activities and is strongly associated with a technology culture that mistrusts academia and especially anything that smacks of the humanities. 

However, regardless of why folksonomy advocates hate the central authority of librarians so much, there are good reasons for not taking this claim at face value.  Let’s start with the idea that is often cited that folksonomies “allow a decentralized collaborative view to emerge”.   This sounds refreshing and liberating but in reality, this decentralized view tends to be surprisingly rigid and conservative.  If you track the top 50 or so tags at sites like Delicious or others, they have actually changed little over the last 3 years.  And in studies that I’ve done, this conservativeness holds true at lower levels also, once you get a minimum number of tags and tagged objects.

More importantly than the surprising conservativeness of popular sites like Delicious, is the idea that these popular sites do have a central authority, and it is the most oppressive and most dangerous type of central authority there is, the authority of the majority.  There is a reason that we don’t see true democracies in politics any more – they are the most dictatorial form of government there is.  Against the will of the people, there is no recourse, no way of insuring the rights of the minority. 

To move from politics to tagging, think of this example, the tag “design” has been in the top 25 most popular tags for the last three years.  Now the word design has a number of meanings and there are a number of different kinds of design, but an examination of the web sites tagged with design on Delicious reveals an almost exclusively small set of meanings having to do with computer or software design.  This is an indication of the nature of the Delicious community, which is dominated by computer geeks as opposed to word geeks or artistic geeks or scientific geeks.  Now suppose that you, like what looks like about .01% of Delicious users, want to tag your favorite interior design site with the word “design” because either the system doesn’t handle noun phrases or because you just find it easier to use single word tags.  And now suppose you want to use the tag “design” to find other sites that deal with interior design or other people that are interested in interior design.  Good luck!  If you click on the word “design” in your basic tag cloud, you will get approximately 1 million hits 99.9% of which have nothing to do with interior design.  Happy wading through page after page of computer design links.

Now, there are clearly ways to get around this, but they all entail having to compensate for the overwhelming impact of the majority (the collaboratively emergent majority).  This makes it harder to tag because no one but you is worried about the rights of the minorities.  And even if you devise a workaround, what are the odds that the other interior designers will think of the same workaround?  And as the number of self-selected computer geeks swamp Delicious with their version of what “design” or other terms mean, this effect will grow even more pronounced.

It seems to me that having a system in which there was a central group of authorities or librarians that you, as a minority can appeal to might work a bit better than letting the collaboratively emergent dictatorial majority unconsciously ride roughshod over the minorities.

Another type of authority that emerges on social sites is that of well-connected individuals and fashion leaders. I would like to see an in-depth  study comparing this type of authority with a designed central authority but my fear is reflected in the W.B. Yeats line – “The best lack all conviction, while the worst of full of passionate intensity.”  As long as there are mechanisms for detecting these passionate worst (from the simple ability of people to detect bullshit to editorial staff on Wikipedia) then the worst of this problem can be at least alleviated.  However, given the madness of crowds structure of social networking sites, I’m worried that this will get worse or at least settle down to a mediocre middle dominating fashion like they do on TV.

4 – Quality of tags is 90% of taxonomy

Now this one can probably just be dismissed as wishful thinking or mindless ideologically inspired nonsense. First, there have been to my knowledge no studies comparing the quality of tags on folksonomy sites with the quality of tags selected from a taxonomy that could be used for the basis of such a numerical claim.  Early studies of tagging inside the firewall suggest that 90% is wildly exaggerated.  I could with as much justification say that they are about 50% or 10% as good. 

5 – Better than nothing at all – often accompanied by argument of expense and difficulty of getting people to use hierarchy

When pressed, such as an early exchange between Clay Shirky and Lou Rosenfeld, it often seems that the last defense that folksonomy advocates fall back on when the issue of quality of tags is raised, is well, they are better than no tags at all. This is often accompanied by the argument that since taxonomies are so expensive and so difficult for people to use, at least folksonomies will get people tagging and even poor tags are better than no tags.  Aside from the overstatement of the expense and difficulty of using taxonomies to tag, there are grounds to doubt that any tags are better than none.  I remember a study that demonstrated than just asking people to tag as they wanted to actually made documents harder to find. 

It is important to keep in mind that tagging is not an end in itself. It is an activity that is meant to lead to using those tags to find information.  And if bad tags, make it harder to find information, then having more tags is not a good thing.  At the very least this claim needs a lot more study before we can claim that some tags are necessarily better than none. 

I also find it striking that we often go from folksonomy being a revolution to “well some tags are better than nothing at all”. That doesn’t sound all that revolutionary to me – especially when you move outside the realm of general internet bookmarking sites.

From my perspective, I would like to end this benefits section with what I think could be one of the more significant benefits from folksonomies and the discussion about them, which is that just getting people excited about metadata is a huge plus.  The real trick will be to maintain that level of excitement while enhancing the quality and usefulness of user generated tags. 

Revolution?  Limits of Folksonomies

The problem I have with folksonomy advocates like Clay Shirky and others is that they wildly overstate pretty much everything about how revolutionary folksonomies are, how difficult and limiting taxonomies are, and worst of all, the comparison of taxonomies and folksonomies. 

For something that is touted as revolutionary, folksonomies have a number of major limitations which severely restrict their impact both in terms of the kinds of uses that can be made of them and the types of web sites and environments that they are useful on. 

We’ve already covered a number of limitations on the use of folksonomy in specific sections above, but let’s summarize them here.  One limit which is one that all information strategies have to deal with, is the issue of scale.  When you only have a few thousand or few million web sites in total and the number of links to any single term is in the hundreds or thousands then almost any information strategy will work.  The problem comes when you start getting a million links to popular tags and large scale web sites like Delicious are already reaching that level for more and more of their tags.  Back when the internet was first really getting started regular search engines worked pretty well, but once the number of pages got large enough, the quality of the results got worse and worse.  I expect the same thing is already happening with large general sites like Delicious. 

And there is another side of the scale issue, which is that on vertical portal sites, the amount of content tends to be small enough that user generated tags might not be as easily overwhelmed with millions of hits, but on the other hand, the target for searching for content on these sites tends to be much more specific, with the focus on finding one or a few particular documents, not a general web site.  This means that the scale issue is here also. 

Another limit on the applicability of folksonomies is that virtually all the success stories are from Internet sites, but when the approach is applied to intranets, the value proposition seems to go way down.  Actually, within the enterprise, user-generated tags are being explored and are being found to generate a lot of value – but only when adapted to an environment that tends to include the central authority of information professionals working with users.  (See the section, Knowledge Architecture for Enterprise 2.0).  Folksonomies with their emphasis on having no central authority are much less valuable in an enterprise environment.

There is one more scale limit on folksonomies, that can be seen in a study done that compared LibraryThing, a social tagging site for librarians, and Amazon.  What they found was that both offered social tagging but that even though Amazon had 1000’s of times the number of users, they generated far less tags.  The suggested reason was that people only tagged if they had about 200 or more items to keep track of.  Less than that and there was no reason to tag for personal benefit and without a personal benefit, very few people bothered to tag much of anything.

Finally, another limit to folksonomies has to do with the well documented poor quality of tags on social networking sites – issues like plurals, misspellings, synonyms, idiosyncratic and extremely personal only tags like toread, funny, box47, or on the table. These kinds of tags severely limit the social usefulness of tags. 

However, there is another quality issue which is more significant in my mind and that is there is really no mechanism for improving the quality of tags that people will do.  First of all, there is no evidence that seeing how other people tag leads someone to become a better tagger.  There are some suggested ways that seeing other people’s tags might make your tags better, but I would argue that you are just as likely to learn bad habits as good tags (again good in the sense of finding info).  And, in addition, there is no evidence that anyone actually goes back and retags.  This would seem to go against the whole notion of the ease of tagging that is credited with getting people to tag.  Indeed, many sites don’t allow retagging.

Lastly, it is very useful to remember that, as enthusiasts often forget, the majority of people writing about folksonomies tend to be extremists.  Not just in terms of their enthusiasm for folksonomies, but also in terms of their overall behavior.  A journalist or thought leader is someone who by nature or by job is happy to spend hours every day playing and exploring and surfing and connecting – because that is what they love and/or that is what they are being paid to do.  But most of the world is not like them – something that many, so many writers forget.  Yes, the new generation spends more time connecting than the last, but no one, outside of the enthusiasts, spends as much time connecting, and thinking about connecting, and thinking and talking about connecting and tagging as the enthusiasts. 

Of course, to be fair, I should mention that librarians have been just as guilty of assuming that everyone was like them and the result was a whole lot of advanced search applications designed by librarians that no one but a librarian could or would use.

What are folksonomies good for?

Does all of the above anti-folksonomy rant mean that I think that folksonomies are completely useless?  No, not at all.  For certain limited areas and applications they are a useful and even exciting but hardly revolutionary addition. 

Within the limits discussed above, folksonomies can be a very useful mechanism for their core strength – discovery of other people who have similar interests.  For discovering and exploring other people with similar interests and sharing the bookmarks that they have added and thereby greatly expanding your resources in those areas of interest is very cool and an exciting way to learn and connect to people.