Mind model. The agent mind represents intelligent agent components, and the architecture consisting of them. Once the agents are in position, they would be able to interact with the believable environment, and therefore strengthen their own believabil- ity. Here we explain how concept model informs agents with traditional agent architecture design, and reserves the space for emotion cognition. Instances of concepts are cues that trigger the activities of the minds of agents. Compare to the concepts, which (relatively) objectively represent the reality, mental activities are subjective because each agent can have different interpretations about the concepts. Mental activities include two major types: sentiments and intents. A concept can map to multiple sentiments, while multi- ple sentiments can also map to one concept. For example, Figure 1 shows that the presence of both darkness and silence in general trigger the emotion of fear. The mapping between concepts and sentiments can be put inside the concept model because certain commonness among human beings can be assumed. How- ever, darkness and silence instill different degrees of fear into different agents, and the weights of darkness and silence vary from agents to agents. Thus, the personalities of agents must be taken into consideration when calculating the triggering of sentiments. Moreover, some agents may fear a particular thing that others do not necessarily fear, such as a rat, a gecko, or something more personal. Such personal emotional triggers can be placed inside an agent to override the default mapping in the concept model. The triggering of intents involves two additional components of agent mind: goals and knowledge. Agents first have goals, and then look for knowledge that can be exploited to attain the goals. Knowledge is basically the connections be- tween concepts [15] isolated concepts cannot help generating the intents since agents rely on the causal effects of the environment to reach the goals. For the purpose of this paper, a goal is defined as a desired effect on a particular target instance. Omitting the target instance means the goal is to apply the desired effect on an arbitrary individual. The following algorithm shows how causal rules in the concept model can be used in the process of triggering new intents. Causal relations that are remembered in the memory of the agent can also be used in a way similar to this algorithm.
