vrml writes In medicine, it is well-known that sugar pills sometimes produce the same effects as real drugs (Placebo Effect). But could that happen with computers too? The first scientific study of the Placebo Effect in computing, just published by the International Journal of Human-Computer Studies , gives an affirmative answer. The experiment considered affective computing, that is those fancy applications that claim to know user's emotions by detecting physiological parameters with sensors. Researchers took two well-known affective computing systems and used them to control in real-time the state of an avatar that looked more and more nervous as users' stress level increased, and more and more relaxed as it decreased. But they also considered a third system in which, unbeknown to users, the sensors were disconnected from the computer and the avatar state was controlled by a random stream of physiological data instead of the real user's data. Results show that participants believed that the sham application was able to display their stress level. Even worse, only one of the two (costly) affective computing systems produced better results than the placebo. This suggests that evaluations of such novel computer applications should include also a placebo condition, as it is routinely done in medicine but not yet in computer science.