![](bg7.png)
of MAS, robotics, and reinforcement learning, the im-
plementation of social learning strategies have drawn
significant attention (see, for instance, (Noble and
Franks, 2003)).
5 DISCUSSION AND FUTURE
WORK
In this paper we have sketched an interaction-based
approach to the adaption of individual decision sup-
port models in MAS. It is desirable when software
agents in complex, dynamic environments need to up-
date, adapt, or improve their knowledge base for de-
cision making. Sometimes, this improvement process
can be based on machine learning from observational
data, alone. But when available data is insufficient
in quantity or quality, when data is too expensive, or
when the machine learning process turns out to be too
complex, alternative approaches are needed. There
are two basic components in our approach: 1) a set of
specific knowledge transfer roles which extends a set
of basic knowledge management roles, and 2) a col-
lection of interaction protocols for knowledge trans-
fer.
Ongoing work on the proposed multiagent frame-
work comprises a prototype implementation of the
KM roles and interaction patterns introduced in Sec-
tion 3. We focus first on the adaption of rule-based
classification models, relying, for advice integration,
on the ABML approach by Možina et al. In the pro-
cess, we will also elaborate the meta-control, used
by an advisee to guide its interactive adaption pro-
cess over multiple knowledge transfer episodes, as a
flavor of local search in a model space. The proto-
type is implemented based on the JADE agent devel-
opment environment. Evaluation will be performed
in the PlaSMA multiagent-based simulation environ-
ment (Warden et al., 2010). In the future, we also
seek to enable a more far-reaching interoperability be-
tween heterogeneous agents in the context of inter-
active model adaption. This includes support for di-
versity in employed models (e.g., rule-based for the
advisee and ANN-based for the advisor(s)). It also in-
cludes support for heterogeneity in the training data
(with respect to attributes), discretization of values in
individual learning, and the naming of attributes and
concept classes. These extensions specifically call for
the provision of additional KM roles, enabling, for in-
stance, semantic mediation.
REFERENCES
Costantini, S. and Tocchio, A. (2005). Learning by Knowl-
edge Exchange in Logical Agents. In WOA, pages 1–
8. Pitagora Editrice Bologna.
Fischer, K., Schillo, M., and Siekmann, J. (2003). Holonic
Multiagent Systems: A Foundation for the Organisa-
tion of Multiagent Systems. In Holonic and Multi-
Agent Systems for Manufacturing, pages 1083–1084.
Springer.
Gehrke, J., Herzog, O., Langer, H., Malaka, R., Porzel, R.,
and Warden, T. (2010). An Agent-Based Approach
to Autonomous Logistic Processes. Künstliche Intel-
ligenz, 24(2):137–141.
Jakob, M., Toži
ˇ
cka, J., and P
ˇ
echou
ˇ
cek, M. (2008). Collab-
orative Learning with Logic-Based Models. In Adap-
tive Agents and Multi-Agent Systems III. Adaptation
and Multi-Agent Learning, volume 4865 of LNCS,
pages 102–116. Springer.
Jennex, M. (2009). Re-Visiting the Knowledge Pyramid. In
Proceedings of the 42nd Hawaii Intl. Conference on
System Sciences, 2009., pages 1–7. IEEE.
Langer, H., Gehrke, J. D., Hammer, J., Lorenz, M., Timm,
I., and Herzog, O. (2006). A Framework for Dis-
tributed Knowledge Management in Autonomous Lo-
gistic Processes. Intl. Journal of Knowledge-Based &
Intelligent Engineering Systems, 10(4):277–290.
Možina, M., Zabkar, J., and Bratko, I. (2007). Argu-
ment Based Machine Learning. Artificial Intelligence,
171(10-15):922–937.
Napierala, K. and Stefanowski, J. (2010). Argument Based
Generalization of MODLEM Rule Induction Algo-
rithm. In Rough Sets and Current Trends in Comput-
ing, volume 6086 of LNCS, pages 138–147. Springer.
Noble, J. and Franks, D. W. (2003). Social Learning in
a Multi-Agent System. Computing and Informatics,
22(6):1001–1015.
Ontañón, S. and Plaza, E. (2010). Concept Convergence in
Empirical Domains. In Proceedings of the 13th Intl.
Conference on Discovery Science, pages 281 – 295.
Pan, S. and Yang, Q. (2009). A Survey on Transfer Learn-
ing. IEEE Transactions on Knowledge and Data En-
gineering, pages 1345–1359.
Settles, B. (1995). Active Learning Literature Survey. Sci-
ence, 10(3):237–304.
Van Elst, L., Dignum, V., and Abecker, A., editors (2004).
Agent-Mediated Knowledge Management: Interna-
tional Symposium (AMKM ’03), Standford, CA, USA.
Springer. Revised and Invited Papers.
Warden, T., Porzel, R., Gehrke, J., Herzog, O., Langer, H.,
and Malaka, R. (2010). Towards Ontology-based Mul-
tiagent Simulations: The PlaSMA Approach. In Proc.
of the 24th European Conference on Modelling and
Simulation, pages 50–56.
TOWARDS MULTIAGENT INTERACTIVE ADAPTION OF INDIVIDUAL MODELS FOR DECISION SUPPORT
559