Training for crisis decision making – An approach based on plan adaptation

Abstract

The human ability to take the right decisions is very important in real world critical situations. An interesting problem always worth being investigated concerns how to teach decision making skills to humans. The real nature of taking decisions is extremely difficult to describe in detail and, as a consequence, training it according to fixed protocols is also challenging. This is because it comes out as a combination of natural talent, competence from previous experience, ability to quick reasoning, leadership, resilience to stress, and so on. We have addressed this problem while building a new learning environment to train crisis decision makers.The environment, called Pandora, is grounded on Artificial Intelligence planning techniques known as “timeline-based”. This technology is used to create and manipulate segments of lesson’s content over time. Planning a lesson corresponds to logically organize events over time that are then rendered in front of trainees during the lesson’s actual enactment. This paper shows how the machinery of continuous plan adaptation is functional to create variety and novelty in the lessons thus engaging the trainees during the teaching interaction. In particular, it shows the different uses of plan adaptation to take into account the basic reactivity of the trainees, the background deductions from user modeling, and the mixed-initiative interactions guided by the trainer.