This study investigated human-robot cooperation in the context of prisoner's dilemma games and examined the extent to which people's willingness to cooperate with a robot would vary according to the incentives provided by a game context. We manipulated the payoff matrices of human-robot prisoner's dilemma games and predicted that people would cooperate more often in the situation where cooperating with the robot was a relatively more rewarding option. Our results showed that, in the early rounds of the game, participants made significantly more cooperative decisions, when the game structure providing more incentives for cooperation. However, their subsequent game decisions were dominantly driven by Cozmo's previous game choices and the incentive structure was no longer a predictive factor to their decisions. The findings suggest that people have a strong reciprocal tendency to social robots in economic games and this tendency might even surpass the influence of the reward value of their decisions.