Towards Diagram Understanding - Studying the impact of layout quality on UML diagram comprehension using eye tracking

Nick Baltsen, Henrik Christoffersen

AbstractOptimizing the quality of diagrams can help to improve the message that the diagram conveys. This is relevant in projects where visualising information is necessary and inter-personal communication is high. In this thesis we studied how layout quality impacts the understanding of diagrams.
To do this, we exposed participants to a large variety of UML diagrams distributed between three different UML diagram types in a controlled experiment. During the experiment the participants had to perform different comprehension tasks, while performance, mental workload and behavior was captured using an eye tracker. The diagrams used in the experiment were categorized based on size (small/large) and layout quality (good/bad). We studied the results at a low level, by studying line crossings and line bends, but also at a high level, by studying performance, mental workload and exploration behavior.
Based on the results, we found that good layout quality improves performance and decreases mental workload. We also found that modelers perform better with less mental workload on large diagrams. Interestingly, we also found that the mental effort associated with understanding line crossings are higher than UML diagram elements. Lastly, we found that in diagrams with good layout, modelers tend to start diagram exploration in the top-left corner and follow the intended exploration pattern. However, in diagrams with bad layout modelers tend start have a more unstructured approach to exploration start and diagram exploration.
Our findings imply that diagram designers should carefully consider the layout when designing diagrams. We assert that line crossings, diagram size, modeler expertise and diagram type should be given individual considerations when designing diagrams.
TypeMaster's thesis [Academic thesis]
Year2014
PublisherTechnical University of Denmark, Department of Applied Mathematics and Computer Science
AddressRichard Petersens Plads, Building 324, DK-2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk
SeriesDTU Compute M.Sc.-2014
Note
Publication linkhttp://www.compute.dtu.dk/English.aspx
BibTeX data [bibtex]
IMM Group(s)Computer Science & Engineering