Abstract
This research aims to investigate existing approaches to development and design
guidelines of affective interfaces in relation to Embodied Conversational Agents (ECAs)
from the perspective of technology, psychology and design. It also asks whether the
affective influences of ECAs differ across the various devices used to display these
characters. This research conducted two surveys to investigate the affective interactions
between users and agents. The first hardware interface experiment explores the emotional
influences of ECAs and user-agent relationships when ECAs are applied to computers
and portable devices. The second character interface experiment studies the affective
influences of ECAs on users in the learning tasks. The thesis asks whether there are
significant experiential differences between ECAs when they are represented by different
hardware interfaces and character classifications, such as emotional influences, character
preference, user engagement and user-agent relationships. This research followed design
guidelines that enabled the exploration of existing research and my own original
experiments into the practical use of multi-agent in the context of a language learning
website. I argue that the user is always at the centre of any ECA design, and therefore
needs to be at the centre of any design process. The thesis discusses some concepts for
developing agent interfaces with more positive affective influences for the completion of
learning tasks.
The main contributions of this research are summarised as follows: (1) the hardware
interfaces of ECAs have a distinct relationship to the affective responses of users,
including emotional influences and user-agent relationships, (2) character classifications
can be related to human affective factors, such as character preference, user engagement
and user-agent relationships, and (3) there are significant relations that need to be
accounted for among character preference, engagement and user-agent relationships.