Information Technology Reference
In-Depth Information
Cognitive Complexity in Business Process
Modeling
Kathrin Figl
1
and Ralf Laue
2
1
Vienna University of Economics and Business Administration, Austria
kathrin.figl@wu.ac.at
2
Computer Science Faculty, University of Leipzig, Germany
laue@ebus.informatik.uni-leipzig.de
Abstract.
Although (business) process models are frequently used to
promote human understanding of processes, practice shows that under-
standing complex models soon reach cognitive limits. The aim of this
paper is to investigate the cognitive diculty of understanding differ-
ent relations between model elements. To allow for empirical assessment
of this research question we systematically constructed model sets and
comprehension questions. The results of an empirical study with 199
students tend to suggest that comprehension questions on order and
concurrency are easier to answer than on repetition and exclusiveness.
Additionally, results lend support to the hypothesis that interactivity of
model elements influences cognitive diculty. While our findings shed
light on human comprehension of process models, they also contribute
to the question on how to assure understandability of models in practice.
Keywords:
Business Process Models, Understandability, Cognitive
Complexity.
1
Introduction
Business process models (BPM) serve as a basis for communication between
domain experts, business process analysts and software developers. To fulfill
this purpose, such models have to be easy to understand and easy to maintain.
Comprehension of process models is relevant for all tasks in which users interact
with models, as for example in business process redesign or implementation of
process-aware systems.
Many researchers have recently turned to investigate comprehensibility of pro-
cess models and investigated various influence factors as modularity [1], domain
knowledge [2] and notational aspects [3]. In addition, various complexity metrics
have been proposed for BPM in the past years (see [4,5] for the discussion of rel-
evant concepts and [6] for a comprehensive survey on related work). It has been
shown that some of these metrics are significantly correlated with the number
of control-flow errors in a BPM [5] and with the understandability of a BPM,
measured in terms of correctly answered questions about the model [7,8]. [2] and
[7] discuss how global measures (like the number of split nodes in a BPM) affect
the understandability of a BPM.
