Interactive Meta-Reasoning: Towards a CAD-Like Environment for Designing Game-Playing Agents - Computational Creativity Research: Towards Creative Machines - page 359

failure and corresponding revisions to the agent's model. Thus, model-based meta-
reasoning too enables experimentation, in this case with the agent's experimentation
with its own design.
The REMmodulewithinGAIA's architecture has the capability of meta-reasoning
over the TMK model of any agent designed in the TMKL2 language for adapting
the agent design to avoid previously encountered failures as well as achieve new
goals similar to the original goals of the agent. This results in GAIA supporting
interactive meta-reasoning: once the designer has constructed the TMK model of an
initial design of the agent, he may experiment with agent design by executing the
agent in the Freeciv world, evaluating its behaviors and revising the agent model.
Alternatively, the designer may ask REM to use its meta-reasoning and propose
revisions to the agent model. The designer may accept REM's recommendations,
reject them, or refine them so that the designer and REM cooperatively revise the
agent's model.
The four experiments in self-adaptation described here cover a small range of
retrospective and proactive agent adaptations. They demonstrate that (i) it is possible
in principle to design game-playing agents so that their teleology can be captured,
specified and inspected, (ii) the specification of the teleology of the agent's design
enables localization of modifications needed for the four instances of self-adaptation,
and (iii) this self-adaption in turn enables the agent to play interactive games, monitor
its behavior, adapt itself, play the game again, and so on. The next steps in our work
are to (a) empirically investigate many more adaptation scenarios, and (b) generalize
from adaptation scenarios to classes of adaptations.
Acknowledgments We thank Lee Martie for his contributions to the construction of GAIA. We are
grateful to the US National Science Foundation for its support for this research through a Science
of Design Grant (#0613744) entitled “Teleological Reasoning in Adaptive Software Design”. An
earlier version of this paper appears in Rugaber, Goel and Martie [ 65 ].
References
1. Aha, D., Molineaux, M., Ponsen, M.: Learning to win: case-based plan selection in a real-time
strategy game. In: Proceedings of ICCBR-2005 (2005)
2. Amabile, T.: Creativity in Context, 2nd edn. Westview Press, Boulder (1996)
3. Anderson, M., Oates, T.: A review of recent research in metareasoning and metalearning. AI
Mag. 28 (1), 12 (2007)
4. Anderson, M., Oates, T., Chong, W., Perlis, D.: The metacognitive loop I: enhancing reinforce-
ment learning with metacognitive monitoring and control for improved perturbation tolerance.
J. Exp. Theor. Artif. Intell. 18 (3), 387 (2006)
5. Benami, O., Jin, Y.: Creative stimulation in conceptual design. In: Proceedings of ASME
IDETC 14th Annual Conference on Design Theory and Methods, Montreal, Canada (2002)
6. Birnbaum, L., Collins, G., Freed, M., Krulwich, B.: Model-based diagnosis of planning failures.
In: Proceedings of 8th National Conference on Artificial Intelligence, p. 318 (1990)
7. Blum, A., Furst, M.: Artif. Intell. 90 (1-2), 281-300 (1997)
8. Boden, M.: The Creative Mind: Myths and Mechanisms, 2nd edn. Routledge, London (2004)
9. Bozdoc, M.: The History of CAD. http://mbinfo.mbdesign.net/CAD-History.htm (2003)