Franziska Raudonat is conducting research in the project “ExamAI – AI Testing & Auditing”. Together with her colleagues, she is investigating how AI can be used in the area of HR. She is also addressing issues such as fairness, liability and transparency. In addition, Franziska Raudonat is studying for a Master’s in Business Informatics, specialising in data science, at Saarland University and the Technical University of Kaiserslautern and is currently writing her master’s dissertation.
1. Ms Raudonat, why did you decide to carry out research in the project “ExamAI – AI-Testing & Auditing”?
The idea came to me while I was studying business informatics with a focus on data science – among other things, through a lecture given by Prof. Katharina Zweig, who heads up the research project. During my studies, I also discovered that AI systems can do a great deal from a purely technical perspective. However, I asked myself how we can ensure that decisions made by algorithms are also fair, comprehensible and transparent. When Prof. Zweig then came to me with the research project, I realised that it overlapped very well with my research interests.
2. What is your role in the research project?
In the project “ExamAI – AI-Testing & Auditing”, I work in the area of “AI systems in personnel and talent management”. Here, the focus is on the use of AI in the area of human resources ...
3. ... an area in which people also have misgivings about the use of AI. Where do you stand on this?
I can absolutely understand that people have concerns. Especially, for example, if applicants for a job vacancy do not know what is decided by AI during the application process and how this decision is reached. That is why I also believe that our project has a responsibility to inform, to explain and point out that AI use can definitely offer benefits.
This is particularly true for the HR area where human decisions are known to be susceptible to bias. Here, AI algorithms provide an opportunity to break up traditional mindsets. But this is only possible if algorithms are designed to be transparent, fair as well as comprehensible – and if people are involved in this process to verify that this is the case.
However, these types of more complex applications are still a long way off. As far as we know, AI is currently only used in the HR area in Germany for scanning CVs, nothing more.
4. Another no less relevant topic is that women are under-represented in AI research. Why do you believe it makes sense to advocate for greater diversity in science – especially in the area of research into artificial intelligence?
It is particularly important for us in the field of research – specifically in AI research – to work more in diversified teams. Incidentally, this applies not just to the relationship between men and women but also beyond that. If algorithms are developed only by one group, then a large part of society is missing – and will also be ignored in the result of a decision taken by an algorithm.
To take just one example, the optical sensors on automatic soap dispensers were trained using “white” without considering the fact the not all people have “white” hands. As a result, these types of soap dispensers did not function for all groups in society. This is a simple example that triggered many discussions. It also demonstrates that it is better if diversified teams work on algorithms because they will probably take more factors into account during their development.
5. Although it is clear that artificial intelligence has already entered our everyday lives, much of it is still a vision of the future. Which AI innovations would you personally favour most?
I can actually think of quite a few, but one that I am particularly excited about is a form of AI that helps school-leavers choose apprenticeships, studies or jobs. The AI system could offer suggestions as a guide, for instance, based on school education, personality traits, skills and other factors.