1. 1/17 | www.janclaes.info
Jan Claes and Gilles Vandecaveye
The impact of confusion on syntax errors
in simple sequence flow models in BPMN
2. 2/17 | www.janclaes.info
Context – introductory statements
Conceptual models are useful
Conceptual models contain errors
Syntax errors are the easiest to study objectively
Syntax errors are the least relevant to study
To study errors, study the creation process
Process-data is available for ‘process models’
3. 3/17 | www.janclaes.info
Problem statement
About syntax errors
in simple sequence flow models
Different levels exist: clear error versus confusion
What is the relation between both?
What can we learn from this?
4. 4/17 | www.janclaes.info
How many syntax errors?
Send PO to
factory
Send PO to
warehouse
Update
contact info
Update
financial info
Close
file
Gateway combines
join and split
semantics
ERROR
Clearly not allowed
by the specification
IRRESOLUTION
No consensus whether it
is allowed/what it means
CONFUSION
Allowed, but
not recommended
6. 6/17 | www.janclaes.info
Error or not?
Send PO to
warehouse
Update
contact info
Update
financial info
Close
file
(wrong symbol)(missing symbol)
DEFINITE ERRORUNCERTAIN ERROR
7. 7/17 | www.janclaes.info
Data collection
Tool = Cheetah Experimental Platform
Task = build sequence flow model
from structured textual description
Language = subset BPMN constructs
start event + end event
XOR gateway + AND gateway (split + join)
Activity + edge (including label)
Participants = 126 master students
Business Engineering
at Ghent University in 2015
13. 13/17 | www.janclaes.info
Conditions for causality
Theoretical argumentation
Confusion influences cognitive load
Cognitive load influences effectiveness
Observed correlation
Not always significant though (inaccurate measures)
Poor external validity though
Temporal precedence
Poor external validity though
? Control for third variables
Maximum attention for equal conditions between
subjects, but no guarantees
15. 15/17 | www.janclaes.info
Conclusions
No solid proof, but input for further research
H1. Using multiple end events causes the modeler to
forget to close all paths in the model
H2. Modelers are not always aware of when using
implicit gateways is allowed and when not
H3. Having multiple start and end events causes the
modeler to forget adding the postponed paths for which
a gateway was already created
16. 16/17 | www.janclaes.info
Conclusions
No solid proof, but support for existing guidelines
Use 1 start and 1 end event (7PMG)
Use no more than 2 start and end events (10PMG)
Model as structured as possible (7PMG, 10PMG, ASP)
Use design patterns to avoid mismatch (10PMG)
Use explicit representation (CSP)
Limit the difference in the number of input/output flows
between splits and joins (QI)
Use of textual annotation (CSP)
Naming guidance (CSP)
17. 17/17 | www.janclaes.info
Do you have any questions?
Do you have feedback?
Thanks for you attention!
Jan Claes
jan.claes@ugent.be
www.janclaes.info
Editor's Notes
Seven Process Modeling Guidelines (7PMG) [14]
Ten Process Modeling Guidelines (10PMG) [15]
Concrete Syntax Patterns (CSP) [17]
Abstract Syntax Patterns (ASP) [18]
Quality Indicators (QI) [16]