October 2, 2006
Tags on a Card Sort
Alok Jain, over at iPrincipia started a topic where he considered extending Web 2.0 techniques to information architecture (I/A). One point in particular that he mentioned got me thinking: how might tags be used in the context of card sorting?
For those who might not know, card sorting is a way to develop an I/A by asking users to group information in ways that make sense to them. Typically, index cards are created for each nugget to be categorized. Stakeholders are then asked to group them together. So Audi, Mini, and VW might be grouped under the category “Car”. Tagging, similarly, allows users to attach keywords to content items. So what’s the difference?
The difference, I feel, is in the relative flexibility that the two models provide. Card sorting tends to be more structured and rigid. Performing a card sort can force users to arbitrary groupings–the card either goes into this group or into that one. Often, items to be organized are multivariate and they can reasonably associated with multiple categories. This is incredibly common when working with a cross-functional group whose members have their own particular vocabulary. Think snow to us and aniu, kaneq, and pirta to the Eskimo. Choosing one “bucket” over another can cause some anxiety–and rightly so. Allowing only exclusive choices, we implicitly bound users’ creativity. In the process, we start to lose the thoughts that provide the context and qualify the choice. To compensate we take notes, but that data is harder to extract for analysis.
Along the same lines, the rigidity often leads to reproducing the current I/A or organizational structure. There seems to be a barrier when asking users to perform a large sort. Groupings are rarely absolute and it can be frustrating for the participant to rethink how information should be exclusively organized. Often, this leads to a risk-averse approach where they fall back upon the prevailing structure because they feel it is “correct.” This can lead to less-than-breakthrough results, especially with wary, uncertain participants.
For its issues, card sorting remains a useful tool. But we might consider complementing it with tagging. Tagging provides an implicit, almost conversational way of capturing more context behind the sort. When a user sorts the items into primary groups we would allow them to tag the item with additional keywords. This “phrase”, if you will, captures more context surrounding the sort and provides granularity, contrast, definition, and other semantic facets.
As we iterate, we could produce a “tag pool” to give users a starting vocabulary–with instruction that they can add to the pool at any time. Kinda like magnetic poetry. As tagging data is captured, this can drive further analysis to see what items are more commonly described in similar ways. Obviously, this would drive the raw taxonomy but it would also provide greater evidence and material for fleshing out thesauri and data dictionaries…