Emotional Synthetic Forces

Abstract

(Maximum 200 words):Report developed under a STTR contract number DASW01-99-C-0037 entitled Emotional Synthetic Forces. The objective of this research was to make the decision-making process of complex agents less predictable and more realistic, by incorporating emotional factors that affect humans. To this end, researchers adopted an approach that promotes the emergence of behavior as a result of complex interactions between factors affecting emotions, integrated in a connectionist-style model, and factors affecting decision making, represented in a symbolic model.This report explains the model sub-components, integration, and testing, which includes a description of the behaviors we used in the development of prototype, the design of experiments, a representative set of behavior patterns that emerged as a result of exercising the model over the design space, and recommendations for future work. Results of prototype suggest an effective means of increasing the variability of computer-generated forces (CGF) behavior in manner consistent with modern day emotional theories. SUBJECT TERMS FOREWORDComputer-Generated Forces (CGF) found in most current combat simulations act more lil<e drones or robots than lil<e liumans. Tliey are too predictable and lack individual human traits and distinct personality types. They almost always have perfect knowledge of the battlefield and do not experience human emotions such as anger, or fear. Unlike the foes that our Soldiers will face on the battlefield their decisions and actions are not influenced by their arousal levels or emotions. Thus current CGF do not provide a realistic opponent for training. The U.S. Army Research Institute for the Behavioral and Social Sciences (ARI) has sought to remedy this training deficiency by funding through the STTR process three different approaches to representing human personality and emotions.This report documents an approach that integrates a connectionist model of emotions from Chown (1993) into a robust Soar cognitive architecture (Rosenbloom, Laird, & Newell (1993). The emotions model includes pleasure/pain and clarity/confusion subsystems that differentiate between positive and negative emotional states, leading to varying arousal levels. Arousal then affects the decision process which is implemented within a symbolic cognitive architecture. Static and dynamic tests of this model suggest that the addition of the emotions model increases the variability of CGF performance, making the CGF less predictable -more like humans and less like robots.ARI's Simulator Systems Research Unit is interested in making training more realistic. One way to do this is by improving the human performance models used to control CGF behaviors. The work described is part of ARI research task 202a, Virtual Environments Research for Infantry Training and Simulation. This research has been briefed to members of the military modeling community including PEO-STRI, the RDE Command, and NAVAIR-TSD, and was presented as a paper at the 23"* Army Science Confe...

DEVS for Human Behavior Modeling in CGFs

“…In their 'Trace' model [11], cues from the environment are appraised as positive (pleasure) or negative (pain), with a certain degree of arousal associated to it, and also a degree of information clarity. Algorithms are used to derive pleasure, pain, clarity and confusion levels.…”