Paper of LiNCS researchers accepted to JASIST

Web search engines are the gateway for users to access health information. This study explored whether a search interface based on the Bing API and enabled by Scatter/Gather, a well-known document clustering technique, can improve health information search. Forty participants without medical background were randomly assigned to two interfaces, a baseline interface that resembles typical Web search engines and a Scatter/Gather interface. Both groups performed two lookup and two exploratory health-related tasks. We found that, except that the baseline group was more likely to rephrase queries and less likely to access general-purpose sites than the Scatter/Gather group when completing exploratory tasks, the two groups did not differ in behaviors and task performance. Participants in the Scatter/Gather group largely overlooked the features (keywords, clusters, and the re-cluster function) designed to facilitate the exploration of semantic relationships between information objects, a potentially useful approach to facilitate users in the rather unfamiliar domain of health. The results suggest a strong effect of users’ mental model of search on their use of search interfaces and a high cognitive cost associated with using the Scatter/Gather features. It follows that that novel features of a search interface should not only be compatible with users’ mental models, but also provide sufficient affordance to inform users of how they can be used. Compared to the interface, tasks showed more significant impacts on search behavior. More efforts should be dedicated to identify salient features of health-related information needs.


With collaborators from UT Austin: 

Zhang YBroussard RKe WGong X.  2013.  The Evaluation of a Scatter/Gather Interface for Supporting Distinct Health Information Search TasksAccepted to the Journal of American Society for Information Science and Technology. :1-25. Google Scholar BibTex RIS