Combining Causal and Diagnostic Assessment in a Probabilistic Model of User Affect
By Heather Maclaren
Abstract:
We present our most recent work in the development of a probabilistic model of
user affect, which is designed to allow an intelligent agent to recognise
multiple user emotions within an uncontrolled environment. Our model deals with
the high level of uncertainty involved in this task by combining information on
both the causes and effects of emotional reactions within a Dynamic Decision
Network. In previous work we designed the causal part of the model by relying on
empirical data integrated with relevant psychological theories of emotion and
personality. The focus of this talk will be on our work devoted to understanding
if and how some of the student’s emotional assessment could be more easily
provided by the part of the model that diagnoses emotional states from their
observable effects.
This is joint work with Cristina Conati.