Goal Conflict Concerns
Proceedings of the Tenth International Joint Conference on Artificial Intelligence. Reasoning. Pages 1025-1030.
Abstract
A major problem for a knowledge based planner is the determination of potential plan failures due to goal conflict. Given a potential plan, it is computationally intractable to compare every effect of that plan against all planner goals. A commonsense planner named KIP, Knowledge Intensive Planner, is described. KIP determines user plans for the UNIX Consultant System. KIP uses goal conflict concerns to focus attention on those effects and goals which are likely to cause goal conflict. A concern refers to those aspects of a plan which should be considered because they are likely sources of plan failure. Violated Default Concerns allow the planner to use default knowledge effectively in cases where the planning situation has not been specified completely. Unintended Effect Concerns allow KIP to plan effectively when plans are used in novel situations. User goals are often only inferred in response to a threat from a potential plan. A new concept termed interests, is introduced. Interests are general states that KIP assumes are important to the user. While interests refer to general states of the world, goals refer to a concrete state of the world.
@inproceedings{ijcai1987p211,
title = {Goal Conflict Concerns},
author = {Luria, Marc},
booktitle = {Proceedings of the Tenth International Joint Conference on
Artificial Intelligence, {IJCAI-87}},
publisher = {International Joint Conferences on Artificial Intelligence},
editor = {John McDermott},
pages = {1025--1030},
year = {1987},
note = {Reasoning},
url = {https://ijcai.org/proceedings/1987/211},
}
TY - CPAPER TI - Goal Conflict Concerns AU - Luria, Marc T2 - Proceedings of the Tenth International Joint Conference on Artificial Intelligence, IJCAI-87 PB - International Joint Conferences on Artificial Intelligence PY - 1987 SP - 1025 EP - 1030 UR - https://ijcai.org/proceedings/1987/211 ER -Copied
Machine-readable versions of this paper, for tools and AI agents: