Modelling emotions in an on-line educational game

Emotional expression and recognition are important in social interactions between people. This is particularly evident in communications between a teacher and a pupil when facial expressions signal levels of understanding and enjoyment will change the teacher's explanation to a pupil, and effective e-learning systems must adapt in the same way if they are provide the social interactions that are necessary for effective pedagogy. Affective computing can equip computer systems with the ability to process social signals and respond accordingly. However, social signals are inherently ambiguous and confusion will result if the approach to processing them is too mechanistic. This paper presents and analyses empirical evidence for this ambiguity, and proposes a possible solution. The techniques are applicable in a wide variety of applications where continuous measures of performance are being assessed.