This study proposes an agent-based approach on an evacuation problem to examine individual protective responses caused by social interactions. The closest town from Kori nuclear power plant in the Republic of Korea is targeted in the illustrative problem. Its street network and residents are represented in a mathematical model of matrices, binomial functions, and if-then rules. A two-level factorial design of 6 factors related to social interactions on street and social networks has been adopted with 2 responses for the numbers of ever-contacted and evacuated residents. The results indicate that the type of available interactions has a strong impact on the information recognition for evacuation, notwithstanding many domestic studies in the evacuation simulation have not incorporated the social interactions prior to the leaves. Thus, this study is expected to support not only the implementation of the social interaction for individual responses under disasters, but the interpretation of its potentials on evacuation participation of residents.