This paper discusses what kind of partial agency is implementable for objects to bring more suitable agencies toward human agent interaction. Human beings have an ability to inform fellows about our intention, internal states and requirements using verbal talk, gestures, attitudes, timings and other representations. The se representations can keep our belief that we are sufficient agents mutually. The robots and virtual agent also mimic these representations, reproduce as if they have an agency, and interact with them. However, their agencies are sometimes too excessive compared to its task. This mismatch leads high cognitive load toward users and brakes interaction consequently. This defect prevents to apply human agent interaction method toward application field. The authors consider that our agency is constructed by multiple features and dividable. If these features are selectable, we can choose more proper design for virtual agents, robots, machineries, daily ho me appliances according to their traits. The authors categorized these agencies in several group and discusses about what elements achieves these features. The paper also shows what method can extract these features fro m human being.