Conceptual Modeling of Events for Active Information Systems - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 268

on the ontological definition of an event as an identified causal occurrence that produces state changes
according to rules and data that describe the event. Conceptualizing an event in this way is consistent
with the definition of an “entity” in conceptual modeling. It provides a parsimonious representation for
active information systems not available in a state-tracking conceptualization.
A number of research challenges remain. These challenges involve the development and evaluation
of constructs, models, methods, and instantiations by which this ontological definition of event can be
effectively represented in the conceptual modeling of active information systems. First, constructs are
needed to conceptually represent (1) the rules that govern state transitions inherent in an event and (2)
the data used by such rules to determine state transitions when the event occurs (Ramesh and Browne
1999). Such rules are likely to be most appropriately represented at the event-type level. Therefore, a
mechanism for categorizing events is also necessary.
Second, such constructs demonstrate effectiveness in enabling the conceptual representation (model-
ing) of information systems and also lead to effective designs and implementations (methods). Current
conceptual modeling formalisms (grammars—such as the Entity-Relationship model—have no such
constructs (Chen 1976; Silberschatz and Korth 1996). They must be developed and evaluated (Hevner
et al. 2004). The parsimony and understandability of models built using these constructs must be as-
sessed, and methods to guide their construction and evaluation must be developed (March and Smith
1995; Gemino and Wand 2005).
Such work will enable the formalization of an ontology of artificial systems (Simon 1996; Sowa 1999;
Geerts and McCarthy 2002; Allen and March 2007a). Such an ontology is needed if we are to represent
and build information systems that provide traceability to lessons learned through the instantiation of
new or modified rules that allow past events to be reanalyzed and different state-histories to be gener-
ated, including proposed future states. The concept of Active Conceptual Modeling (Chen and Wong
2005) is a step in this direction. This effort focuses on enhancing our fundamental understanding of
how to model continual learning from past experiences and how to capture knowledge (i.e., causal
event rules) from state transitions (Robinson and Hawpe 1986). The conceptualization of an informa-
tion system as an event-processing mechanism provides the basis for ontological constructs to gain
this understanding. It provides a theoretical framework for the representation of episodic and semantic
memory (Tulving 1983; Pillemer 1998) within the existing Entity-Relationship formalism (grammar).
Future research should investigate how this conceptualization can be used to develop intelligent learn-
ing-based applications in areas such as homeland security, global situation monitoring, intelligence,
surveillance, and reconnaissance.
Such applications require the capability to analyze and reanalyze interrelated events in order to
form conclusions that ascribe attributes to extant objects in the real-world domain. These applications
require the capability to predict future states based on prior states and posited events. They must be
capable of reporting not only state history as it was believed at a prior point in time, but also as it is
currently believed based on events that occurred but were not known to have occurred at that point in
time. Such capability requires not only the differentiation of the valid time (when the event occurred)
from transaction time (when it was recorded in the systems) (Snodgrass 2000), but it may require the
definition of “determination time” (when it was determined that the event occurred).
The Department of Homeland Security may, for example, be concerned about interpreting events to
ascribe a “likelihood of terrorist activity” (an artificial attribute) to individuals, groups, and organiza-
tions based on events they have initiated. However, the initiation of an event may not be known until a
point in time much later than the time at which it occurred. Rules for ascribing the likelihood of terrorist