Personal tools

Strict vs. Fluid Taxonomies

From Social Patterns

Jump to: navigation, search

Part of leaving the design unfinished involves determing which elements to nail down and which to leave more free-form. Using Flickr, as I often will, as a sort of canonical thriving social application, it's easy to see examples of both design decisions at work. Some elements of Flickr's interface are rigidly defined. These include the object model, the site's master navigation, and the short list of predefined relation types.

The object model provides for people, people have collections of media objects, people can join groups and submit media objects to them, and so on. The site's master navigation has these items at its top level: (Home, You, Organize, Contacts, Groups, Explore). The short list of predefined relation types allows for a person to define another person as a contact (this relationship need not be reciprocated to take effect), and optionally to further classify the person as a friend, a family member, or both.


At Flickr, a contact can be classified as friend, family, or both (but that's it).

I'm not saying these design and information architecture decisions have never changed. Since its launch, Flickr has added a second media type (video), and has refactored its navigation menus without changing the basic philosophy. It has also changed from a free-form connection model that allowed users to define additional relationships to the narrow one it has today (because relatively few users took much advantage of this feature, so it offered limited value for the maintenance required).

This last change backed off from the more fluid taxonomy approach that, where appropriate, can enable users to invent concepts, labels, classifications and groups in an evolving way that meets their needs without requiring that you, as the designer, fully anticipate every conceivable scenario that your social application might foster and support.

In addition to these "rigid" taxonomy elements, Flickr does also give its users unlimited freedom along some carefully defined axes to invent whatever meaning they need. Examples of this include Flickr's well known free tagging feature that enables users to tag their own objects and gives users the option of permitting others to tag them as well.


There's no way the designer of a social application can anticipate every tag a user might want to apply. What controlled vocabulary, for instance, would ever include a tag called "thehairofchrisheilmann"?

Another free-form taxonomy element inherent in Flickr's design is the unlimited ability to create groups with any conceivable name or purpose. This feature involves a number of patterns we'll discuss presently, including the concept of a group, ridiculously easy group formation, discussions, joining, invitation, and the ability to add media objects to a group's "pool."

Flickr users also invented the concept of an award associated with a group. These often gaudy images are offered to users in the comments on a particularly relevant image or video and generally accompanied by an invitation to join the associated group (or at the very least to proudly display the award which, incidentally, then functions as a sort of advertisement for the related group). Many people consider these awards tacky and pushy, but they do represent an innovation invented by users and permitted (but not directly supported) by the Flickr UI.

In this way (with or without awards) groups can function as browsing "pivots" for the user, taking them from the image of a friend, to a related group, and then on to other images.


A caricature of Merlin Mann in the style of da Vinci's "Vetruvian Man" prompts an invitation to a group dedicated to just such parodies and variations.


The Vitruvian variations group showcases a series of images with a common theme.


Another image in the group then prompts this further invitation that appears to facetiously parody the whole "Hi, I'm an administrator for a group called..." social interaction.