Measuring the Utility of Gaze Detection for Task Modeling: A Preliminary Study

Search engines present readers with a list of documents ranked by predicted relevance to a keyword query. Salient sections of documents that are selected for examination can be highlighted using similar techniques. An ability to restructure information presentation based on an analyst’s initial interactions with an information space might improve search outcomes. This study takes the first steps towards designing an analysis environment in which structural overlays evolve in response to an analyst’s actions. Two sources of implicit feedback are explored: topical similarity to material included in a very brief written report, and eye behavior patterns. Results indicate that using eye-tracking can be as effective as lexical overlap, but more work is needed.