A variety of works on building and adapting life-like agents for human-computer communication have been conducted. An important aspect we should consider when designing them is how they are considered by users. Through the appropriate mechanisms we should permit the users' suspension of disbelief for agents and give users the conviction that they can proceed with the communication efficiently and friendly by intervention of agents. We present some preliminary work, a generalized and extended framework for the generation of emotions in anthropomorphic agents. This model has been designed as part of an emotion-based response system which can be used in various applications. A basic idea of the model is placed on the cognitive appraisal model, but it is extended to include other emotion eliciting factors, which are not considered directly in the theory but known to be closely related to the generation of emotions. By representing the structure of the cognitive theory as a simple weighted network, the method provides a flexible way of modeling different aspects of personalities and temperaments.
展开▼
