Detecting the Eureka Effect in Complex Search

In search tasks that show a high complexity, users with zero or little background knowledge usually need to go through a learning curve to accomplish the tasks. In the context of patent prior art finding, we introduce a novel notion of Eureka effect in complex search tasks that leverages the sudden change of user’s perceived relevance observable in the log data. Eureka effect refers to the common experience of sudden understanding a previously incomprehensible problem or concept. We employ non-parametric regression to model the learning curve that exists in learning-intensive search tasks and report our preliminary findings in observing the Eureka effect in patent prior art finding.