Behavior believability in virtual worlds: agents acting when they need to.

Area of Science:

• Artificial Intelligence
• Human-Computer Interaction
• Cognitive Science

Background:

• Believability in intelligent virtual agents (IVAs) is crucial for user engagement.
• Agent autonomy, driven by internal states and needs, is key to achieving believability.
• Existing research on agent motivation often lacks a comprehensive model for internal driving factors.

Purpose of the Study:

• To present an architecture and model for representing and managing internal motivations in IVAs.
• To explore how bio-psychological needs theories can inform agent motivation.
• To enable emergent, self-determined behavior in IVAs through motivation synthesis.

Main Methods:

• Developed a novel architecture for managing internal driving factors in IVAs.
• Utilized Maslow and Alderfer's bio-psychological needs theories to model human needs.
• Implemented a motivation synthesis approach to generate emergent behavior.

Main Results:

• The proposed model effectively represents and manages internal motivations in IVAs.
• The architecture supports the generation of autonomous behavior based on synthesized motivations.
• Basic physiological needs were identified as foundational for driving action and behavior.

Conclusions:

• The motivational approach enhances IVA believability by simulating internal needs and drives.
• This architecture provides a framework for creating more autonomous and human-like virtual agents.
• Focusing on basic needs can lead to more robust and emergent agent behaviors, even without external stimuli.