Human-human interaction studies have shown that we prioritize psychological resources (attention, memory, etc.) to significant others (e.g., family member, friends) and to a stranger assigned to co-work with us. We examine whether the human-robot interaction shares a similar nature. We adopted a “shape-identity matching task” in which participants first learned to associate 3 shapes with 3 names, and then judged whether the shape-name association in each trial was matched or mismatched. One of the 3 names belong to a social robot with a humanoid face (ASUS Zenbo), who was introduced as a partner to co-work with our participants. The other two names belong to the participant’s best friend and to a stranger. In half of the trials, the trials were printed in green, and in the other half of the trials, the names were printed in red. Each participant was instructed to respond to the trials in their assigned color (red or green), and the other half would be completed by their robot partner (co-work on the same task). We found that a robot partner was endowed with a “partner-advantage” similar to a human partner: the trials associated with the robot’s name were responded faster and more accurately than the trials associated with a stranger’s name. This advantage toward a robot is developed quickly and requires minimum interaction between human and robot. Interestingly, the effect is further boosted when the robot’s expected role (i.e., social companion or functional robot) matches the robot’s presence during the behavioral task (i.e., presence or absence). This was never observed among human partners. This unique feature in human-robot interaction implies that we may evaluate a robot partner heavily on the alignment between the expectation and its delivered service.