Personalized search methods try to tailor the search results to the user's needs and preferences. A popular source of information for personalization is social networks of users. In this paper, we aim at proposing a social-network-based personalized information retrieval (PIR) method that: i) is holistic (not just considering some local neighborhood of the searcher); ii) is efficient in term of computational/storage cost; iii) its personalization component has a probabilistic basis. To the best of our knowledge, no such method exists. We propose an interestedness measure and several variations of it for quantifying the interestedness of each user in another user based on a social network of users. The general idea is to try to mimic a searcher's behavior in the real world to estimate the interestedness. The measures are then exploited to personalize the retrieval results. We evaluate the resulting PIR methods and compare them in terms of retrieval effectiveness and computational/storage cost. We also compare them with some baseline methods. In summary, our analyses suggest that retrieval based on at least one of the measures performs well in terms of both retrieval effectiveness and computational/storage cost.INDEX TERMS Information retrieval, personalization, personalized pagerank, social networks.