An Instance of What? Determining Cues, Attributes, and Goals in a Virtual Task Environment

摘要

The construction of instance-based cognitive models is often advocated as a way to reduce the complexity of modeling. Through an in-depth case study of modeling driving and adversarial behavior in a virtual environment, we show that the representation of the attributes and cues that make up instances can be enormously complex. Eye tracking and Cognitive Task Analysis are employed as methods to identify the key attributes, and their utility is demonstrated through automated goal recognition and driving performance in a virtual domain.