Two things send our Uncanny Valley creepiness meter straight to red - realistic robots that flirt and anything that laughs. 

Realistic laughing is actually not trivial. We see it in real life, we can often tell when someone is fake laughing. On computers it is even more off-putting. Scholars at the Numediart Institute of the University of Mons are members of a project called ILHAIRE - The Science Of Laughter, which seeks a better understanding of laughter features in order to incorporate them into interactive systems. 

Stéphane Dupont, senior researcher at the circuit theory and signal processing laboratory of the Mons Polytechnic Faculty, is project coordinator of ILHAIRE, and talked to about how laughter makes virtual characters more realistic and changes how we interact with them.


Why should laughter be incorporated in virtual systems?

Laughter is very important in human interactions. It may occur up to three times per minute! In a conversation, laughter regulates the flow, brings some nuances, betrays a lack of confidence, or reveals feelings such as joy, amusement and irony. However, current voice-based interactive systems, such as those you may use to book a plane ticket, or avatars in serious games [games for learning purposes], for example, do not incorporate laughter. They are efficient in recognizing neutral speech or providing realistic vocals synthesis. But they are not expressive. Incorporating laughter into these systems might lead to applications such as making the perception of a message more positive or managing the frustration of users confronted to a problem during a reservation process.

How do scientists go about studying laughter?

They are many approaches. For example, we want to make a machine able to recognize human laughter and its characteristics. Future applications could feature a virtual companion, either on a screen or as a robot, who would help exploit the fact that laughter is contagious. They would thus be susceptible to enhance the enjoyment of a person watching a funny movie.

We have been working with a team of psychologists from the personality psychology and assessment section at the University of Zurich, in Switzerland, to assess the positive effects of these virtual agents on a person's mood. We also aim at synthesizing laughter using algorithms in a very controlled way and to incorporate these algorithms in human-computer interaction systems. Among other topics, we have studied are so-called gelotophobia—the fear of being laughed at—and the positive effects of laughter to relieve chronic pains.

What does your experiments consist of?

For example, we fit our candidates with microphones, movement sensors and 3D cameras, and we show them a funny movie. We collect the data from which we build models mimicking physical effects of laughter. We incorporate these models into avatar programs. Our avatars are then able to laugh while displaying realistic facial motion and gesture typical of real laughter. 

What are your achievements to date?

We have built ten databases and related models corresponding to different situations of laughter: watching a movie with a virtual companion, engaging in a three-person discussion, playing different games. These databases are freely accessible to the science community. Our laughter synthesis tools have found an application in virtual characters designed by Cantoche, a company based in France and the US, to meet the needs of companies in the fields of e-learning or customer assistance. Finally, we will pursue our collaboration with the Zurich team in order to learn more about the positive effects of laughter.