People's ascription of intentional agency to robots necessitates understanding how, when, and why people attribute robot behavior to underlying intentional states. While many studies explored mind attribution to robots including its determinants and consequences, little attention has been given to the attribution of second-order mental states, such as a robot's beliefs about people's intentions during interactions. In an online study (n = 155), participants watched a video of a humanoid robot tracking a ball hidden under one of two cups. 19% of participants predicted that the robot could correctly locate the ball after it was displaced twice by a person deliberately distracting the robot, indicating implicit attribution of second-order beliefs to the robot. These implicit attributions influenced participants' actions in a subsequent interactive game with a virtual counterpart of the robot but did not affect their explicit assessments of the robot's second-order reasoning. In contrast, observing the robot demonstrate second-order reasoning by correctly identifying the ball's location affected participants' explicit attributions but not their behavior in the interactive game. This reveals a complex interplay between implicit and explicit attribution processes in how people interpret robot behavior.