Information recommender system attempts to present information that is likely to be useful for the user. Some information recommender systems recommend persons which users are likely to follow in Twitter. Showing recommendation reason is an important role of the systems. However, current recommender systems give only simple or quantitative reasons for the recommendation. In this paper, we aim at giving precise and non-quantitative reasons which are also easy to understand. We make use of formulas in first-order predicate logic for explaining the reason. In order to build such formulas, we use Inductive Logic Programming. We succeeded to extract several useful formulas from micro-blog.