Garau, Maia; (2003) The impact of avatar fidelity on social interaction in virtual environments. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text The_impact_of_avatar_fidelity_.pdf Download (11MB) | Preview

Abstract

The research presented in this thesis concerns the contribution of different levels of virtual human (or 'avatar') fidelity to social interaction in virtual environments (VEs). VEs present new possibilities for mediated communication by placing people in a shared 3D context. However, there are technical constraints in creating photorealistic and behaviourally realistic avatars capable of mimicking a person's actions or intentions in real time. At the same time, previous research findings indicate that virtual humans can elicit social responses even with minimal cues, suggesting that full realism may not be essential for effective social interaction. This research explores the lower boundaries of fidelity by investigating how different levels of responsiveness, photorealism and behavioural realism affect people's experience of interacting with virtual humans. The research presented comprises three between-group experiments involving over 200 participants. The experiments focus on distinct but interrelated questions. In the first experiment, conducted in an immersive Cave-like system, participants explored a library containing a group of seated virtual readers. The aim was to investigate the degree to which the virtual humans were responded to as social entities as their responsiveness increased across the conditions. Results indicated that responsiveness significantly affected a range of responses, including the degree of sentience attributed to the agents. The findings signalled the need to further define the different social responses affected by virtual humans. Participants' heart rate and electrodermal activity were also recorded to explore the possibility of employing objective measures to study responses to virtual humans. The remaining two experiments focused on behavioural realism and photorealism in the context of mediated communication between pairs of human participants. In the second experiment, participants interacted via a video-tunnel that provided them with a head-and-shoulders view of an avatar representing their conversation partner. Analysis of questionnaire responses indicated that the avatar whose gaze behaviour was inferred from conversational turntaking significantly outperformed the visually identical, but behaviourally less realistic, random-gaze avatar. An in-depth qualitative analysis of the interview responses yielded a theoretical model of the possible effects of avatar appearance and behaviour on perceptions of the communication experience. For example, one outcome of the analysis with respect to gaze was the fact that excessive upward eye movement tended to undermine trust in the avatar. In the third and final experiment, participants were either in a Cave-like system or a head-mounted display. Each partner was represented to the other by a full-body and life-size avatar. The first goal of the experiment was to discover how the gaze models from the previous experiment would perform in a more demanding immersive setting; the second was to determine the impact of photorealism. Results indicated that the higher-realism avatar was significantly improved by the inferred-gaze model, whereas the lower-realism avatar was negatively affected. This suggested a strong interaction effect between appearance and behaviour. The findings from both experiments have implications for inexpensive ways of implementing avatar eye gaze for improved interaction in shared VEs. The research presented in this thesis identifies a range of responses that can be affected by different levels of avatar fidelity. The overall conclusion is that even minimal behavioural cues can enhance the perceived quality of social interaction in VEs. However, the caveat is that there is a need for consistency between the fidelity of the avatar's behaviour and its appearance.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item