Context-aware computing aims to enable the on-line applications and services to be executed in a timely and automated manner. Many of such applications and services involve group-level interactions. For more smoothing communication within a community, specific group-level issues such as group dynamics must be considered. To do so, obtaining group-level contexts such as the role, conflict resolution and norms, are key ingredients to improve group performance. Since group context is not the same as a simple summation of individual context, as group is not just a simple set of individuals, awaring individual context is not sufficient for group-level communication support. However, context-aware computing research still has stressed more on individual context. This leads us to the motivation of searching for group context aware method. Hence, the of this paper is to propose a novel methodology which automatically recognizes group context. Especially, we focus on group conhesiveness in this paper just because group cohosiveness is one of the important variables to control the performance of group interaction. To verify the applicability of the proposed method, an empirical test has been conducted to compare the performance of the proposed methodology with that of conventional methods.