engleski

Metanarratives and Believable Behavior of Autonomous Agents

The aim of this conceptual paper is to propose separating social attitude engines from emotion engines of autonomous agents, as one way of increasing their social intelligence and consequently the believability of their behavior. The aim of the proposed separation is to clearly distinguish between the social and psychological aspects of agent behavior. In the existing emotion engines, the two aspects are blended to a degree which frequently prevents modelling of the elements of complex social interactions found in contemporary society. Our view is that the development of the proposed separate socio-political modules of social attitude engines could enable introduction of political and ideological elements into agent behavior. One way of introducing these elements into the agents social attitude engines is via their narrative knowledge. In order to accomplish this, Jean-Francois Lyotards notion of metanarratives has been used in this paper, as well as Fredric Jamesons reinterpretation of that notion. Three globally recognizable ideal types (neo-liberal, fundamentalist, and alternative) have been supplied with narratives translated into a conceptual model applicable in modelling of conversational agents. In the future, the presented socio-political attitude model should be expanded by means of addition of attitudes from various other areas of social life, in order to develop complex social attitude engines.

socially intelligent autonomous agents, conversational agents, believable behavior, metanarratives

Skup je održan 16.07.2002.g., Bologna, Italija.