We present here an investigation into the use of social laughter in human-machine interaction. We employed an overhearer study, in which the interaction between a virtual real-estate agent and a client visiting an apartment was evaluated. Two conditions were considered, without and with laughter, the only difference between the two cases being the inclusion of laughter, at appropriate locations in the agent's speech, in the latter case. A large pool of participants listened to the recording corresponding to one of the two conditions, while watching a slideshow of the visited apartment. They were asked to judge the interaction of the agent, based on the following dimensions: professionalism, communication, pleasantness, formality and spontaneity, as well as expressing the likelihood of recommending the apartment and the agent to acquaintances. The results showed that the scores of the five dimensions decreased while the apartment and the agent ratings increased, in the laughter case, but only the formality difference reached significance. Nevertheless, a linear regression analysis showed that condition had an effect on the ratings, mainly through interactions with other factors. We discuss these findings and propose further directions of research to be followed.