Abstract — Agents within a group can have different perceptions of their working environment and autonomously fulfil their goals. However, they can be aware of beliefs and goals of the group as well as other members so that they can adjust their behaviours accordingly. To model these agents, we explicitly include knowledge commonly shared by the group and that obtained from other agents. By avoiding actions which violate “mental attitudes ” shared by the majority of the group, agents demonstrate their social commitment to the group. Defeasible logic is chosen as our representation formalism for its computational efficiency, and for its ability to handle incomplete and conflicting information. Hence, our agents can enjoy the low computational cost while performing “reasoning about others”. Finally, we present the implementation of our multi-agent system. I.