Search engines are indispensable for locating information in WWW, but encounter great difficulties in handling exploratory information seeking, where precise keywords are hard to be formulated. A viable solution is to improve efficiency and quality of exploratory search by utilizing the wisdom of crowds (i.e., taking advantage of collective knowledge and efforts from a mass of searchers who share common or relevant search interests/goals). In this paper, we present an epistemology-based social search framework for supporting exploratory information seeking, which makes the best of both search engines' immense power of information collection and pre-processing and human users' knowledge of information filtering and post-processing. To validate the feasibility and effectiveness of the framework, we have designed and implemented a prototype system with the guidance of the framework. Our experimental results show that an epistemology-based social search system outperforms a conventional search engine for most exploratory information seeking tasks.