Flow, as described by psychologists, refers to a state of total involvement in a task. Scientists and practitioners consider the role of flow to be essential because the phenomenon is directly related to desired outcomes, such as increased well-being and task performance. Originally, flow could only be measured by surveys. However, surveys interrupt individuals' flow. Furthermore, it is not possible to measure flow in real-time via surveys. Recent research shows that flow can be classified using machine learning based on physiological data such as heart rate variability. Due to the highly individual nature of physiological data, privacy protection and anonymization of such data is a particularly important issue. A promising approach to further improve and ensure privacy could be the concept of differential privacy, which can be integrated at different levels of the flow classification process.
In the context of this thesis, the suitability of differential privacy approaches on different levels will be investigated. The most promising approach will be prototypically implemented and tested. For this purpose, a data set from a flow field experiment will be provided. The highly individual character of physiological data and the limited size of the data set compared to the complexity of physiological patterns and features pose particular challenges for this research.
We expect extensive knowledge of procedures within established machine learning methods, such as Random Forest, Support Vector Machines, and Neural Networks. In addition, experience in pre-processing and analysis of physiological data for the thesis is required. Ideally, legal knowledge of handling personal data in general and health data in particular is present.
If you are interested in this topic and want to apply for this thesis, please contact Nico Loewe with a short motivation statement, your CV, and a current transcript of records. Feel free to reach out beforehand if you have any questions.