Early warning of task failure using task processing logs

This study realizes a task management system that enhances people's work productivity by evaluating the task process situation accurately with respect to the deadline. Often, tasks are not completed before their deadline because of poor time management, i.e., miscalculation of the total time required to finish the tasks. This type of mistake occurs repeatedly, because of a cognitive bias called palnning fallacy, and users fail to notice the situation until it is too late. In this study, we design a system that can correct the task process behavior of users and improve their time management skills by providing information on the future success or failure of their tasks in the early phase of processing. To accomplish this, we analyzed lifelogging data related to processing of 288 tasks undertaken by from 153 people, and investigate two types of prediction methods. The first prediction method evaluates the reliability of the task processing plan, and the second evaluates the progress situation. Simulation results show that the proposed method can predict the future of tasks with an accuracy of 77.3% when the task processing plan is provided as input, and with an accuracy of 95.0% when the plan and subjective progress rate of a task after the third time working on the task is known. The results suggest that the outcome of tasks in progress can be predicted from user inputs such as the subjective progress rate and cumulative work time. Using these results, we construct a system that provides support for decision making on task process behavior.