Abstract
Gaze is a crucial method to communicate through non verbal signals. Due to our social nature, users usually expect to interact with synthetic characters just like they interact with other people. In our work, we are concerned about the authenticity and believability of synthetic characters during this reciprocal sequence of actions. Multiparty conversations are many times underestimated in synthetic character's field. Our work focuses on gaze as a multi-user communication tool. We present a solution, based on psychology theories of attention where multiple users interact with a synthetic character without breaking the suspension of belief. The work presented in this dissertation is located in this area of relevance and addresses the specific question of creating autonomous believable behaviour to support the engineering of believable synthetic characters. This dissertation specifically covers an eye gaze system based in anticipatory mechanisms, towards different users. The anticipatory behaviour proposed depends on each user's internal model, on the task progress, and it is inspired by psychology theories of visual attention, such as the spotlight model, covert and overt attention, and bottom-up and top-down processing. We believe that one synthetic character that is able to distinguish users and reacts differently towards them is rather believable that one that does not. In this dissertation, we show how supporting multiple users can be used to control the behaviour of a synthetic character interacting with different human users, and autonomously produce believable and understandable behaviour.