Learning and Adaptation for Crisis Response

By their very nature, crises place overwhelming demands on the human planners who must cope with them. This research effort aims to develop an intelligent advisory system that will assist humans in responding to complex crises, focusing especially on the task of planning and scheduling responses to chemical spills and fires.

The basic approach involves retrieving an appropriate plan from a preexisting plan library, adapting this plan to the current situation, and tracking it during implementation, adapting it further as the need arises. The advisory nature of the system supports unobtrusive collection of users' decisions, which in turn provides training data for learning.

Experimental studies have shown that using previous plans reduces user response time while maintaining response quality and that, over time, the advisory system can adapt its behavior to different users. A longer-term goal of this subproject is to use learning methods to acquire coordination strategies among different planners, and thus to reduce the number of conflicts that arise during the crisis-planning process.

This work was funded by the Office of Naval Research through Grant N00014-96-I-1221.

Contributors to the Project

  • Dr. Melinda Gervasio

  • Dr. Wayne Iba

  • Professor Pat Langley

  • Stephanie Sage

  • Related Papers

    Gervasio, M. T., Iba, W., & Langley, P. (1999). Learning user evaluation functions for adaptive scheduling assistance. Proceedings of the Sixteenth International Conference on Machine Learning (pp. 152-161). Bled, Slovenia: Morgan Kaufmann.

    Langley, P., & Fehling, M. (1998). The experimental study of adaptive user interfaces (Technical Report 98-3). Institute for the Study of Learning and Expertise, Palo Alto, CA.

    Iba, W., Gervasio, M., Langley, P., & Sage, S. (1998). Evaluating computational assistance for crisis response. Proceedings of the Twentieth Annual Conference of the Cognitive Science Society. Madison, WI: Lawrence Erlbaum.

    Gervasio, M., Iba, W., & Langley, P. (1998). Learning to predict user operations for adaptive scheduling. Proceedings of the Fifeenth National Conference on Artificial Intelligence (pp. 721-726). Madison, WI: AAAI Press.

    Gervasio, M., Iba, W., & Langley, P. (in press). Case-based seeding for an interactive crisis response assistant. Proceedings of the AAAI-98 Workshop on Case-Based Reasoning Integrations. Madison, WI.

    Gervasio, M., Iba, W., Langley, P., & Sage, S. (1998). Interactive adaptation for crisis response. Proceedings of the AIPS-98 Workshop on Interactive and Collaborative Planning. Pittsburgh, PA.

    Iba, W. and Gervasio, M. T. (1997). Crisis response planning: A task analysis. Unpublished manuscript, Institute for the Study of Learning and Expertise, Palo Alto, CA.

    For more information, please send email to

    gervasio@isle.org .