(How) Can Robots Make Commitments? A Pragmatic Approach