Using Bayesian methods and simulator data to model lifeboat coxswain performance

Abstract

Lifeboat training is normally performed in controlled conditions to minimize the risk to trainees and equipment. Participants are given limited or no opportunity to practice skills in operational scenarios that represent offshore emergencies. For this reason, human performance in plausible emergencies is difficult to predict due to the limited data that is available. Simulation provides a means to collect novel data on human performance and learning in situations that are otherwise prohibitive due to risk. In this study, we use simulator data to shape knowledge of the problem space of lifeboat coxswain training and skill transfer. We use Bayesian inference to produce human performance probabilities (HPPs) to model the performance of lifeboat coxswains as they practice lifeboat tasks for the first time. Data collected in an experiment are used (1) to generate probability distributions to predict the amount of practice needed for new coxswains to achieve competence on lifeboat launching and maneuvering tasks, (2) to study how skills learned in training transfer to a new scenario, and (3) to make comparisons between task difficulty. The methodology can be applied to other problems to assess training effectiveness and improve instructional design. Models can be continuously strengthened with additional data to improve predictive accuracy. Probability distributions can be used to assess competence in new scenarios and to diagnose strengths and weaknesses using machine learning.