Barrett, Neil (2007) User preference and embodied conversational agent groups. Masters thesis, Memorial University of Newfoundland.
- Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
The interface used to interact with the everyday computer is evolving, possibly towards embodied conversational agents (ECA/agent). ECAs are virtual characters with a body and persona that typically communicate with users via natural language speech. ECAs are successful in many ways and multi-ECA environments exist; yet there is a lack of results, especially pragmatic results, pertaining to a user's experience in multi-ECA environments. -- An experiment was realized, using a general framework built for simulating ECA groups, to understand user response to such groups. In particular, this experiment studies the difference in user response to a second additional agent in a single agent "group". -- WOZECA, a framework for simulating ECA groups, was built to accommodate ECA characteristics researched to date and facilitate ECA experiments. The framework presents several ECA group interface characteristics using a behind-the-scenes operator and configurable video clips. This flexibility permits multiple different experiments such as the one presented in this thesis. -- The experiment manipulates the number of embodied conversational agents (levels: 1 and 2 ECAs) and agent behaviour (levels: neutral, positive) in a two factor, two level design. The experiment manipulated ECA behaviour. The literature suggests users will respond socially and naturally to the interface. This prompted the use of the following constructs to quantify user response: group cohesiveness, the user's emotional state (valence & arousal) and the group's impact on user self-efficacy. -- The results suggest that a second additional agent influences user response. Specifically, ECA behaviour has more influence on group cohesion and valence than does the number of agents. Furthermore, group cohesion depends on the user's personal style, age and the ECA group's behaviour. WOZECA successfully simulated ECAs during the experiment and is capable of a variety of future experiments.
|Item Type:||Thesis (Masters)|
|Additional Information:||Includes bibliographical references (leaves 109-116).|
|Department(s):||Science, Faculty of > Computer Science|
|Library of Congress Subject Heading:||Human-computer interaction; Intelligent agents (Computer software); Interactive computer systems; Natural language processing (Computer science); User interfaces (Computer systems)|
Actions (login required)