Towards a method of identifying the causes of poor user experience on websites

User Experience, in particular the affective state of the user, is an important consideration in Human Computer Interaction, hence integrating affective measurements with software user experience testing would be valuable. Current approaches to this problem either lack the level of detail required to identify the causes of poor user experience, or can do so only with considerable human expertise and input. We aim to examine the possibility of automatically identifying the specific elements of a software system which cause user experience problems, without human input, by combining psychophysiological measurements and detailed user interaction data. This paper describes ongoing work to collect a dataset suitable for exploring the problem, and briefly discusses some future directions in which the data may allow us to proceed.