Scheduling with Uncertain Resources: Elicitation of Additional Data

Abstract

Abstract-We consider the task of scheduling a conference based on incomplete data about available resource and scheduling constraints, and describe a procedure for automated elicitation of additional data. This procedure is part of an interactive system for scheduling under uncertainty, which identifies critical missing data, generates related questions to the human administrator, and uses answers to improve the schedule.

Learning of personalized security settings

“…For example, a travel assistant may elicit user preferences [13]. We have applied similar techniques to eliciting preferences in the context of scheduling [2,3,6,7]. Researchers have also studied user modeling in recommender systems [10,14].…”

Scheduling with uncertain resources: Learning to make reasonable assumptions

“…The system's elicitation mechanism analyzes the expected schedule improvements due to potential data requests, selects the most important requests, and presents them to the user [3]. If the user provides the requested data, the system uses them to improve the schedule.…”

“…The task of scheduling under uncertainty gives rise to several problems, including the representation of uncertain data, the automated construction of schedules based on these data, and the use of past experience and common sense to make reasonable assumptions about unspecified resources and constraints. To address these problems, we have developed a system for scheduling based on uncertain data [1][2][3][4][5][6], which has been part of the RADAR project (www.radar.cs.cmu.edu) at Carnegie Mellon University, aimed at building a software agent for assisting an office manager.…”

Scheduling with uncertain resources: Learning to ask the right questions

“…The examination of the previous elicitation techniques has shown that they are inapplicable to this problem, and we have developed a novel elicitation mechanism, based on evaluating the impact of missing data on the schedule quality [3]. We now present a learning procedure that improves the performance of this mechanism and helps to adapt it to new scheduling scenarios.…”

“…We have described this scheduling system in a series of four papers; specifically, we have explained the representation of uncertainty [2], search for near-optimal schedules [11], elicitation of additional data [3], and collaboration between the system and its user [10]. We now present a learning mechanism for improving elicitation strategies.…”