Category:
Problem discussion
This position paper relates the experience of
one such information system under development, which is in the analysis phase,
and defends the position that an agent modeling approach to these kind of
problems is suitable, even though the full power provided by agent systems
is not currently needed. The system currently under analysis can be viewed
as an application integration problem and will most likely be implemented
using EAI or B2B technologies [2]. The information system requirements are
already well known, and our work focuses on identifying the requirements of
the partners, i.e. other federal or local government agencies, businesses
and citizens and defining a common domain ontology.
The federal government agency wants to integrate a number of its existing applications into a single information system for its own purposes and make it useful to others by transforming it into an open system. An open system provides access to the information system to its external partners. It also creates links with the data of the information systems of its external partners (so it cannot be viewed as simply a client-server type of information system).
An open system is needed because:
One of the main objectives of an open information
system is to share data between its partners, and to avoid a chaotic situation
where everyone would have their own redundant, and inconsistent data that
could not be shared. The inability to share the data would hinder the ability
to define business level processes across partner boundaries, e.g. between
the federal and regional governments. The model that is proposed is centralized:
partners interact with the federal government agency which owns the bulk of
the shared data, but they do not interact directly (they can of course, but
not with this information system).
Most current object-oriented analysis methodologies
are aimed at modeling a single information system.A0 The above type of problem exhibits two characteristics
that indicate that an agent modeling approach to these kind of E-government
problems is suitable even though the full power provided by agent systems
is not needed at the moment:
Since the information system currently focuses
on sharing distributed data, existing integration technologies can be used
for implementation. However, as the information system is deployed and users
gain experience they may come to expect features more elaborate that agent
technology can more easily provide. The agent approach provides many interesting
modeling concepts for analyzing these kind of problems [3, 4, 5] such as goals,
roles, organizations, and ontology. An important issue is whether adopting
an agent modeling approach is too cumbersome for such a problem given the
tight time constraints of the study.
The main objective is to exchange data, not to
define inter-partner business processes. The basic issue is thus to define
a domain ontology that is agreed upon by all partners, i.e. each partner can
translate to and from the domain ontology to his own information system. One
must notice that the existing legacy data models cannot easily be changed
because of the existing data, and that the common ontology needs to consider
this.
Our approach is to
The approach thus uses the few agent concepts
such as goal, goal refinement, role and ontology that are needed to solve
the problem at hand. The ability to use a subset of an agent methodology is
important: it allows us to experiment new modeling techniques on new types
of problems without compromising the tight time constraints of the study.
Furthermore, it allows the model to evolve in the future as more sophisticated
features are required by users by using a more complete set of agent concepts
if needed.