Among of the factors that a↵ect likeability, reciprocal response towards the other party is one of the multiple variables involved in social interaction. However, in HRI, likeability is constrained to robot behavior, since mass-produced robots will have identical physical embodiment. A reciprocal robot response is desirable in order to design robots as likeable agents for humans. In this paper, we discuss how perceived likeability in robots is a crucial multi-factorial phenomenon that has a strong influence on interactions based on reciprocal robot decisions. Our general research question is: What type of reciprocal robot behavior is perceived as likeable for humans when the robot's decisions a↵ect them? We designed a between/within 2x2x2 experiment in which the participant plays our novel Alternated Repeated Ultimatum Game (ARU G) for 20 rounds. The robot used in the experiment is an NAO robot using four di↵erent reciprocal strategies. Our results suggest that participants tend to reciprocate more towards the robot who starts the game and using the pure reciprocal strategy compared with other strategies. These results confirm that the Norm of the Reciprocity applies in HRI when participants play ARUG with social robots. However, the human reciprocal response also depends on the profits gained in the game and who starts the interaction. Similarly, the likeability score is a↵ected by robot