Successful interaction between information systems and information searchers could greatly improve the information finding performance. However, almost all information retrieval systems deliver a ranked list of retrieved documents that somehow match the query, ignoring the information searchers' real information needs behind the query. In this talk, I will present an approach that delivers the retrieved documents according to searchers' question answering task. Our experiment indicated that paying attention to the task and the searcher's interaction could provide substantial improvement in task performance.