|
Organization responsible: TNO People involved: Project description: The goal of this project is to develop, demonstrate and evaluate an intelligent agent for training support of a tunnel operator. This companion agent challenges the learner to improve his performance and stimulates him to analyze the actual performance by means of introspection.
A companion agent is a computer-simulated character, which has human characteristics and plays a non-authoritative role in a simulator-based learning environment (e.g. a simulated co-learner or advisor). Companion agents are typically useful in modern, constructive learning situations where learners can take control of their own learning process (Stehouwer, Stricker and van Gemeren, 2006). The companion agent challenges the learner to improve his training performance and stimulates him to analyze the actual performance by means of introspection. The agent operates on the same authority level as the operator, and is therefore less threatening than a traditional, authoritative instructor.
The development of companion agents is a challenging job in itself. The chosen interaction metaphor needs to meet the learner characteristics (preferred learning style, foreknowledge, the person’s character, etc.) and the type of tasks to be performed by the learner (e.g. situation assessment, decision making, system operation, and communication tasks). When designing and implementing a companion agent in accordance with a chosen interaction metaphor, complex design choices have to be made. These choices regard the agent’s verbal communication (message content and meaning, intonation, tone colour, word craft, etc.) as well as the agent’s visual communication (facial expression, body posture, clothing, etc.).
In our research, we focus on the design and implementation aspects regarding the verbal communication between the learner and a companion agent. We intend to develop a conversational companion agent that assists a tunnel operator who is training an incident management task. In particular, we focus on the agent’s verbal communication during the incident assessment tasks of the tunnel operator. When assessing an incident, the operator has to interpret his most recent observations and classify the actual incident situation as one of the standard incident types (varying from a small car incident (with bodywork damage only), to a large-scale fire incident with casualties). These observations can either be observations in camera pictures from the cameras mounted inside the tunnel, or observations from witnesses inside the tunnel.
The main research question of this project is: What are the necessary agent functionalities (e.g. perceiving, information collection, communicating) for a conversational companion agent that accompanies a tunnel operator during the training of situation assessment tasks? Publications:
|
