Evaluating Explanation Function in Railway Crew Rescheduling System by Think-Aloud Test

Abstract

A previously developed interactive system with an explanation function generates crew schedules automatically by using a rule base in an 'if-then-because' format. Delivering 'because' information to the user is helpful for decision making because it visualizes the computing process. We evaluated the effectiveness of the explanation function by think-aloud testing in which the verbal comments of each participant were recorded while the participant was solving a pre-specified problem of crew rescheduling in our system. A cognitive model for using the system was obtained in a user state transition form by protocol analysis of the comments. The results revealed that users spent 28% of their time for interpreting 'because' information and this information played an important role in the decision making. Interactive processing with the help of the explanation function is practical for real-time railway crew rescheduling.