DFG Research Training Group 2167

The Gaze Sharing project to support social navigation and guidance is associated with the DFG Research Training Group User Centered Social Media at the University of Duisburg-Essen. It examines innovative human-machine interfaces that also make use of the recording, evaluation and reflection of eye movements (eye tracking).

As part of the project, the development and testing of algorithms that are used to analyze eye movement data is essential. The necessary findings and data are obtained in realistic scenarios within empirical studies. The domain of "learning to program" (computing education research) in the sense of source code comprehension serves as the starting point. The observation and analysis of students solving comprehension tasks related to Java source code is a realistic scenario not only for universities and is highly relevant for the current teaching situation in computer science and computer science-related courses.

The studies carried out to date have shown that learners can encounter a variety of comprehension hurdles when reading source code in order to build up an understanding of the structure and content of the source code excerpts shown. These problematic points for understanding can be provoked, for example, by so-called "code smells", where certain constructs in the source text do not correspond to a standard used for the respective language. Another study has shown that the structure and the way in which the source code is presented (visualization) can lead to comprehension barriers.
In addition, a further study has shown that different help systems are conducive to better understanding and are used by the test subjects in practical use, as shown, for example, by interviews and performance data (see figure).

The data from the studies conducted show that test subjects can develop a variety of barriers to understanding, that source text is sometimes perceived very individually and that support systems are used very differently. The data and findings obtained can be used in studies and research projects that go beyond the dissertation project in order to create and research systems that support program comprehension.

The research work is the subject of Fabian Deitelhoff's dissertation entitled "Developing Eye Tracking Methods for Detecting Source Code Comprehension Strategies". Fabian Deitelhoff successfully defended his cooperative dissertation with the University of Duisburg-Essen in December 2020.