Daniel E Rose on search interfaces
Returning to the chi2003 papers, I was interested in the observations of Daniel E. Rose from Altavista in his paper "Reconciling Information-seeking Behavior with Search User Interfaces for the Web" [Microsoft Word doc - 226kb].
"Search interfaces reflect the attributes of the tools we use for searching, rather than what we know about how people look for information ... In fact, nearly every web search enigne offers users the identical search experience, regardless of the task they are trying to accomplish"
Rose's paper looks at the way Altavista's Prisma feature allows users to add exact phrase matches to their search strings in order to refine their results.
This is similar to features introduced in the recent Yahoo! search re-launch, where you can see an expandable list of suggested categories underneath the search box, for example initially four refinements for the query 'bbc', then an additional six, and finally a full and overwhelming 95. And also to the Teoma 'refine' feature which expands to a more manageable fifteen, and Vivissimo's clustering feature.
It was interesting to see Rose use screen-shots from an example search utilising the Prisma feature, and comment that because the iterative refinement approach modelled reduces the results to 8,880:
"The search is now sufficiently narrow that no further refinements are suggested. The first search result, which exactly matches the user's topic, is shown in the figure."
However I can't help but notice that whilst demonstrating the strength of the refinement system, the example shows one of the major flaws in the presentation of search results to the user - that even the best search technology cannot always provide an abstract or description that makes sense to the end user. Prisma may have helped the user construct a complicated search syntax using relevant exact phrases, but the description still doesn't convey to me what I would find if I clicked on the link.
Elsewhere in the paper Daniel Rose also observes that:
"There is a discrepancy between the rich and varied search behaviour of users and the simplistic, one-size-fits-all approach of web search engines."
and proposes three very promising principles that should guide future search interface design and development:
- Different interfaces (or at least different forms of interaction) should be available to match different search goals.
- The interface should facilitate the selection of appropriate contexts for the search.
- The interface should support the iterative nature of the search task. In particular, it should invite refinement and exploration.