Computing Emotion

In a market place where efficient cognition and critical thinking skills are prized values across the gamut of professions one may wonder why governments and economists are concerned about the well-being of the population for anything more than humanitarian reasons. The main reason, from a psychological perspective, is that cognition is deeply affected by emotion: from attention, to evaluatation and preference, to decision-making, goal generation, planning, motivation and learning. There have been several theories of emotion from William James who pointed to the role of bodily changes in the production of emotion to other approaches focusing on the role of appraisal in the creation of emotion. As usual the closest approximation to the truth seems to be an intermediate point where both the body and the brain together create the experience of emotion. The structure of the nervous systems into it's parasympathetic and sympathetic aspects would support this contention and a main point of actual synergy and of analysis is the heart. So it turns out that the millions of crooning love songs were right and the heart can tell us quite a lot about emotion, as can other measures of physiology such as skin conductance, temperature and also body movements. So the question arises as to if it is possible that a computer can analyse your physiological response to a stimulus and tell you what you're feeling. Why would we want to do this? Well then an intelligent computer that understood a users emotional state could, in combination with other data, learn user preferences, and respond accordingly. Imagined uses could be the perception of driver fatigue and advising a driver accordingly, or to manipulate the drivers mood through adjustment of the music in the car (if angry then classical music- or would this make you worse!?). In the work place, similarly frustation could be monitored and adequate times for 'time-outs' could be advised. On the positive side, biofeedback could help workers achieve so-called 'flow' states. Nasoz et al. take the first step towards such 'futuristic' devices in calculating alrogithms based on physiological data (heart rate, galvanic skin response, body temperature) to predict six emotional states (sadness, anger, surprise, fear, frustration, amusement) with an accuracy of 60-70%. Considering that people are often quite limited in their own capacity to identify what emotion they are experiencing (emotion differentiation) this is quite an encouraging result though it is achieved in a tight experimental paradigm. It also opens up the question as to who knows better what you're feeling, you or the computer!

read more...