Database Reference
In-Depth Information
transitions from one level of abstraction of problem representation to another. In general, the
study provides an important example of a research approach that makes it possible to evaluate
data modeling at a detailed level as a problem solving process. Building on an important line
of research, Batra and Antony (2001) investigated the effectiveness of a consulting system
that is designed to reduce data modeling errors and found out that individuals with a low
initial knowledge level benefi ted from the consulting system.
Effects of specifi c characteristics of conceptual modeling grammars.
Several studies have
been conducted building on the foundation of Wand and Weber's theoretical work (Wand,
Monarchi, Parsons & Woo, 1995; Weber, 1997) on the use of ontology as a conceptual basis
for constructing and evaluating conceptual modeling grammars. These theory-testing studies
have focused on specifi c characteristics of the grammars and their impact on user performance
in specifi c types of tasks. Weber (1996) utilized a strong theoretical foundation in cognitive
psychology and philosophy to evaluate whether or not humans tend to see entities and at-
tributes as distinct constructs, and his conclusion based on a memory recall experiment is
that these, indeed, are separate elements. In another study building on the same theoretical
foundation discussed already above in the context of domain familiarity, Burton-Jones and
Weber (1999) confi rmed their theory-based predictions that using relationships with attributes
would have a negative impact on problem-solving performance in unfamiliar domains, but
they also found out that this result did not hold in familiar domains. Finally, Bodart, Patel,
Sim, and Weber (2001) studied the use of optional properties (attributes and relationships)
in conceptual modeling. Their three-experiment study found that models with optional
properties serve well when the purpose is to help users gain a surface-level understanding,
but that optional properties should not be used if the users need a deep-level understanding.
Building on a different theoretical foundation, Siau, Wand, and Benbasat (1997) investigated
the effects of the use of structural constraints (such as explicit cardinality constraints); their
results reveal that modeling experts pay much more attention to the structural constraints
than to the surface semantics conveyed in textual descriptions.
Having reviewed the results of prior usability research on conceptual data modeling,
we continue by evaluating the implications of these results and suggesting several new
avenues for future research.
POTENTIAL FOR FUTURE RESEARCH
Given the maturity of data modeling in practice and the results summarized above, it
would be easy to conclude that further human factors research related to conceptual data
modeling may not add substantially to the existing body of knowledge. In the next section
we hope, however, to demonstrate that because it has focused on a relatively narrow part
of conceptual data modeling, prior research has left several potentially important questions
still unanswered.
Most of the empirical studies reported above that have investigated conceptual data
modeling from the human factors perspective are based on the same relatively simple model:
in a controlled laboratory study, subjects with only modest experience complete one or
several modeling tasks in which they create a graphical representation of an organizational
situation based on a narrative using one or several conceptual data modeling formalisms.
The results are typically evaluated by grading the models using a solution created by the
researcher as a baseline; results achieved with different formalisms are then compared to
