Predicting stopping behaviour: a preliminary analysis

The analysis of search transaction logs often characterizes a search session but rarely looks at the end point. When do users stop, and what cues are present suggesting that stopping is eminent? In this preliminary analysis of the logs of 288 search sessions conducted in a laboratory setting, we identified the activity performed by participants as well as search transitions that were invoked over the course of a search session. The 4331 search transitions (15 per task on average) contained a total of 9295 actions. We isolated the final transition in each search session for detailed analysis. As hypothesized some behaviours are predictable, and suggestive of stopping behavior, with the potential for modeling.