We discuss the generation of humorous acts by embodied agents. How can a humorous act be constructed from the discourse and when should it be displayed? Rather than introducing algorithms for humorous act production we discuss the issues that are involved. From our observations it becomes clear that current research on affective computing, research on generating and interpreting facial expressions and research on embodied (and intelligent) agents can and should be combined with humor research. Results can help to design new and interesting applications in human-computer interaction using embodied agents.