UC Berkeley Professor Marti Hearst has just completed Search User Interfaces, an academic book on the topic. Cambridge University Press will be releasing print copies in September but the full text is available online now for free.
The terms of service for the online version of the book does not permit posting any of its contents so, even though short excerpts from the book would probably be acceptable fair use, I’ll respect the terms of service and I won’t be quoting from the book in this article.
I don’t consider this series of articles to be a formal review of the book but more of a sampling of ideas I found interesting and instructive.
This is a book about search system usability, specifically about how to design and evaluate search user interfaces. The book also considers different models of the information seeking process, how users specify queries, how search engines present results, and other topics. I was delighted to see current discussions of faceted search, personalization, and visualization. The book’s final chapter discusses promising trends in search interfaces.
The book is geared to two audiences, academic people who want to research or teach search usability and developers of such systems. However, the book is approachable enough that someone like me who is in neither category finds it quite valuable. Although the book does not give any attention to federated search (the term does not appear at all in it) and references metasearch only four times, I imagine that most of the teachings of good search engine usability design apply to federated search systems as well. I do also acknowledge that federated search has its own unique usability challenges: displaying incremental results and organizing/navigating results from different sources are two that come to mind.
Chapter 1 explores the design of search user interfaces. Hearst discusses the importance of keeping the interface simple. She notes that 1997 search results from Infoseek look almost identical to those from Google in 2007. Presumably her point is that the search industry has been slow to adopt changes that improve usability. She further notes that, even with a simple search interface like Google, many users – in particular novice users – have many problems.
Hearst refers to a set of usability factors that she believes should influence user interface design: learnability, efficiency, memorability, errors, and satisfaction. She also discusses seven guidelines for search interfaces that support the usability factors. Providing shortcuts for advanced users, for example, supports efficiency. Showing some results immediately and highlighting search terms also support efficiency. Chapter 1 is packed with very concrete suggestions for improving the user experience.
Chapter 2 tackles the question of how to assess a search user interface. In the discussion, Hearst refers to three main aspects of usability, as defined by ISO: effectiveness, efficiency, and satisfaction. The goal is to measure search interfaces against these criteria, as objectively as possible. Hearst discusses a simple yet surprisingly effective approach to testing an early design which she refers to as low fidelity, or lo-fi, testing. Prototype screens of the application are made on paper. Potential users of the system interact with the various pages of the prototype and provide input on what works for them and what doesn’t. Hearts makes the interesting point that evaluators are typically less hesitant to criticize a paper prototype than one that looks like much work went into implementing it. Lo-fi designs lead to hi-fi implementations and further rounds of evaluation. Hearts discusses other approaches to assessing usability: employing usability experts to assess a design, and watching users “in the field” work with an application.
The chapter continues with detailed discussions of how to perform formal testing and controlled experiments. These studies are difficult to perform and require quite a commitment of time and resources. In performing a formal test, questions come up such as what to control for, how to obtain the right participants, how to measure their responses and preferences, and how to track their behavior over time. The rest of the chapter discusses how to mine search logs for helpful information, and presents cautions to consider when evaluating search interfaces.
Continued in Part II.
Tags: federated search