FLAME: Fuzzy Logic Adaptive Model of Emotions
My master's research focused on developing believable characters. Previous research showed that simulating the emotional process is necessary for building believable characters (OZ project, see Resources' page). My research aimed to investigate the role of the learning process on believability. I designed a model called FLAME (Fuzzy Logic Adaptive Model of Emotions) (El-Nasr et al. 2000, see publications' page ). Emotional states have no definite boundaries. Therefore, one possible method for representing these states is to use fuzzy linguistic variables and fuzzy sets. The model used Fuzzy Logic inference rules and OCC (Ortony's work on Emotions, see Resources' page) to derive an emotional state based on the character's goals and expectations. I used reinforcement learning, conditional learning, among other learning algorithms, to derive expectations given a situation. Emotional states were translated to behaviors based on a number of rules derived from psychology research.
Using FLAME I was able investigate the impact of emotions on character believability. Additionally, I investigated the impact of learning on believability. The model was tested through a simulation called PETEE (a PET with Evolving Emotional Intelligence), where users rated the believability of the character, among other character qualities. The results collected from these experiments demonstrated that learning has a major impact on the character's perceived believability.
This work won 'best student paper' award at the Autonomous Agents Conference '99.
PETEEI (A PET with Evolving Emotional Intelligence)
PETEEI is a simulation of a pet using the FLAME model. The pet learns from its interaction with its user. It eventually aquires emotions such as shame and pride depending on its interaction with the user.
Additionally, PETEEI was designed to recognize and cope with the various moods and emotional reponses of its owner.
An Emotionally Expressive Baby
In this project, we have integrated the emotional agent model simulated in PETEEI with visualization models for mapping emotions into facial expressions to create an interactive simulation of an agent (a baby agent) that produces appropriate facial expressions in a dynamic environment.
