Federal Funding: AI Research for Secure Software Development

- EN - DE
Keeping track of complex software development is not easy - a new research proje
Keeping track of complex software development is not easy - a new research project at the University of Würzburg aims to solve this problem. (Image: Scholtes/JMU)
A research project at the University of Würzburg (JMU) has received 137,000 euros in federal funding. The project will develop an AI early warning system to prevent errors in the development of new software.

The research project is being carried out by a team led by Professor Ingo Scholtes, chair of Machine Learning for Complex Networks at the CAIDAS Centre for Artificial Intelligence and Data Science. "Our goal is to create an AI-based early warning system for software development," he explains. "It should detect problems in the organisational structure, provide early indications for emerging issues and help managers to take countermeasures." That makes it a real game changer - especially for complex software development, which often involves many people over long periods of time. "Managers can use the platform to quickly see who is working on what tasks, where key roles in the team are located, and what potential sources of error exist." The project is scheduled to begin in April 2024.

Previous Research Serves as a Basis

The development of the new AI platform is based on Scholtes’ previous studies on the social organisation of software teams. In these studies, for example, the computer scientist was able to show that the so-called Ringelmann effect also occurs in software projects. This is a phenomenon known from social psychology - it describes how the individual contribution of a team member tends to decrease as the size of the team increases. Put more simply, in large teams people tend to put in less effort because the responsibility for the outcome is spread across more shoulders. Scholtes’ team was also able to show that the structure of the collaboration network (i.e. who works with whom) influences how strongly this effect manifests itself in a team. The new project will use these findings to identify problematic collaboration structures and provide information on how to improve them.

The Würzburg researchers have recently started to create a database for the project. They are using git2net, a software developed at the chair, which automatically analyses data from the online collaboration platform github.com and generates temporally resolved collaboration networks between the members of a software team. These form the basis for further analysis, visualisations and AI applications.

The original title of the research project is "Data-Driven Platform for Proactive Risk Management in Software Development Projects".