Substantial improvements must be made in the usability of AI planning technologies in order for them to achieve widespread adoption. In particular, planning systems must be designed to better serve the needs of users, who generally want to play a central and ongoing role in the plan development process. In this paper, we argue that improved usability requires a new representational layer that captures metatheoretic properties of a planning domain. A domain metatheorywould provide an abstract characterization of planning elements that highlights key semantic differences among them. This paper presents a candidate model for a domain metatheory, as well as an instantiation of that model for a travel-planning domain. The paper also describes three user-centric planning capabilities that the model enables, namely, user directability of planning, generation of qualitatively different plans ,a ndplan summarization.
[1]
Karen L. Myers.
User Guide for the Advisable Planner
,
1998
.
[2]
R. Michael Young,et al.
Cooperative Plan Identification: Constructing Concise and Effective Plan Descriptions
,
1999,
AAAI/IAAI.
[3]
Karen L. Myers,et al.
Generating qualitatively Different Plans througt Metatheoretic Biases
,
1999,
AAAI/IAAI.
[4]
John Levine,et al.
Generation of Multiple Qualitatively Different Plan Options
,
1998,
AIPS.
[5]
Karen L. Myers.
Planning with Conflicting Advice
,
2000,
AIPS.
[6]
Greg Linden,et al.
Interactive Assessment of User Preference Models: The Automated Travel Assistant
,
1997
.
[7]
David A. Shlapak,et al.
Perspectives on Theater Air Campaign Planning
,
1995
.
[8]
Karen L. Myers.
Strategic Advice for Hierarchical Planners
,
1996,
KR.