In “Tag Clouds are the New Mullets” (Daily Report, 19 April 2005), I claimed that the weighted tag clouds meme popularized by Flickr and Technorati was about to cross a permanent shame threshold because of overuse. My comment suggested that the only sin of tag clouds was popularity. But the problems with tag clouds run deeper.
In the 1960s, Steve Reich created a new kind of music by playing identical audio tapes on two players and letting them slowly drift out of synch. Ten years later, Brian Eno generated much ambient music by putting tape loops to a similar use. Eno, an accomplished pop music composer, said at the time that he had become more interested in process than product. Eno didn’t compose Discreet Music, the technology did (or so he claimed). This once avante-garde reliance on process works better as art than as architecture.
We who make websites must strike a fine balance between guiding our users and allowing them to lead us. We listen but we also synthesize and invent. We conduct user research but we interpret the results. We ask what users want but we decide what they are really telling us — and we, not they, determine how best to fulfill the needs they didn’t necessarily realize they were articulating.
Tag clouds remove the guidance and artistry from our side of the equation, offloading all the work to our users. What’s popular? What’s important? Users decide. This might be okay if the process did not create a false intellectual equivalence between high- and low-level topics, and if it did not skew toward popularity at the expense of findability.
The idea behind tag clouds is that users know best. Their actions determine how other users navigate. Their choices leave a trail. Typically, though not always, the “important” topics get big while those considered less important (which in this case only means less popular) get small. Once they get small enough, they disappear.
In Flickr and Technorati, users create their own tags (“design,” “cats,” “California”). When enough people have used the same tag, it begins to show up in the cloud. Once a lot of people have used it, it becomes a visually dominant element, encouraging others to click it — and subtly discouraging them from creating their own tags.
As tag clouds come to replace expert taxonomies in common practice, carefully constructed hierarchies vanish. In their place is a flattened world where every idea, at any level, is a topic as worthy as any other. Eight Mile is a topic at the same level as Detroit, which is a topic at the same level as Cities, which is a topic at the same level as United States, and so on.
Instead of a hierarchy based on user-centered classification systems, the tag cloud “hierarchy” is based on raw usage. If several citizens of Detroit view a collection of photos tagged Eight Mile, upload their own photos of that street, and tag their photos Eight Mile, then Eight Mile becomes an important — and visible — category. If no one visits what would ordinarily be a “master” topic page such as Cities or United States, then those master categories shrink in size until they are invisible.
The intellectual problem is that tag clouds create a data world where subtopics are detached from their parents; where the very notion of parent/child relations no longer exists. The counter-argument is, who cares? If everyone digs Eight Mile, let’s make Eight Mile easy to find. Instead of relying on humans to mine the data every three months and have long tedious arguments about how to update the navigation, let’s allow software to do it in real time, based on actual user behavior. Let the process create the music. There is merit to this view, especially on the community sites from which it sprang. (There is no merit to it on single-author sites, where one person creates all the content and all the tags. If you don’t have a clear purpose for your site, who does?)
The less brainy and more pressing problem is that with tag clouds, topics either gain immediate, widespread traction with the public, or they disappear from the cloud. Once they disappear, it is as if they no longer exist. Few users will ever find them. Network effects being exponential, what is immediately mildly popular quickly becomes artificially very popular, while what has yet to become popular never will be.
In an ordinary IA structure, if a photo site contains pictures of Istanbul’s Taksim area, a user can find those pictures by clicking through a taxonomy based on the way folks look for such stuff (Turkey: Istanbul: Taksim). What are the odds of finding Taksim in a tag cloud? Unless the site is devoted to Istanbul nightlife, it’s unlikely that any user will ever find those photos, because they will not be popular enough to show up in the cloud. If the site’s goal is to let only the most popular stuff float to the top, then tag clouds work like James Brown. But if its goal is to offer a better way of letting users find any content they desire, then tag clouds are as wrong as the Patriot Act.
The same problem plagues any web content mining service powered by popularity. Popularity sometimes promotes quality but it is often a finder of mindlessness: extreme leftist or rightist rants, passed-out co-ed photos, embarrassing videos of people who can’t dance trying to dance and people who can’t sing trying to sing.
Every blogger knows of a half dozen services like Blogdex or Daypop that list “hot” posts in the selective ring of small publications some of us inaccurately choose to call “the blogosphere.” A post becomes hot when two people with somewhat visible blogs link to it. Once it appears in Blogdex or the Daypop Top 40, a hundred more bloggers will link to it, either because it interests them or just to signify their membership in the tribe.
Thanks to the exponential nature of such linkage, our lucky post soon has 500 links. Some people link to it without even reading or looking at it, simply because a trustworthy blogger like Kottke linked to it first. Less fortunate articles and discussions wither and die, unnoticed.
Tag clouds harness all that mindless accidental randomness and make it the driving engine for navigating deep, ever-expanding content troves. Older ways, based on library science, undoubtedly suffer from the disadvantage of not being new. But they help people find what they need. And that is what navigation should do. (Discuss.)
Trick out your Trashcan. Decorate your Drives. Mix up Mail. CandyBar 2 lets you change the Mac OS X icons you usually can’t!
This site is designed with Web standards.
This is the entire N.W.A. Straight Outta Compton album edited down into just the “explicit” content. ... Support the arts and enjoy curse words.
More than 1,300 digital images depict elevation views and floor plans for middle and upper class apartment buildings from New York City's pre-World War I residential building boom.
Although his name is not commonly remembered today, Coolidge’s images seem permanently seared into the American public conscience. They are among our most important and endearing images of American popular culture and stand beside the icons of other illustrators such as Rockwell, Flagg and Wyeth.
Like mood rings and fanny packs, like mullets and the Macarena, the weighted tag clouds meme popularized by Flickr and Technorati is about to cross a permanent cultural shame threshold. Brilliant as the idea remains, faddishness is choking its air supply. Damned clouds are everywhere.
It’s not just blogs that are using weighted tag clouds. Businesses are shoveling them into interface makeovers, with predictably mixed success. Thus Lulu, a company that helps people publish their own books, CDs, and other products, offers a half-hearted tag cloud to help customers browse categories.
It is of course wrong to compare weighted tag clouds to mullets, mood rings, and similar instances of mindless pop-cultural detritus. Tag clouds are not dumb. Their smartness is why so many have rushed to use them. But ubiquity and repetition quickly turn sweets to ashes. Kopy Kats Kill Klouds?
Those with long memories will remember ABBA. The rest of us may just about recall the good work of the CSS Samurai when they launched the Acid Test back in 1997 and challenged makers of browsers world-over to improve their support for CSS 1. Well, dammit, we’re at it again. ... Acid2 is a brand new test designed to push the limits of HTML, CSS, and PNG support in browsers and authoring tools. ... Early feedback is coming in from the likes of Safari developer Dave Hyatt over at Apple. [Hyatt has already used Acid2 to isolate and fix two CSS bugs in Safari.]