Fasching, F. (2024). Towards trustworthy artificial intelligence in railway operations : a qualitative case study on adopting human-centric and ethically responsible AI integration for resource planning [Diploma Thesis, Technische Universität Wien]. reposiTUm. https://doi.org/10.34726/hss.2024.118360
With Artificial Intelligence (AI) shifting heavily into the focus of large corporations andresearchers, both the development of systems incorporating techniques and methods using various forms of AI, as well as the research in this field has sped up tremendously the last few years. This development and usage of novel technologies due to an urgency created by external influences often means that human-centrality and ethical alignment is neglected for a reduced time-to-market. Surveys have found that the automation and other technological advancements in AI are a source of anxiety among people. Due to the high impact and transformative force of AI, debates and research focusing on the values and principles of AI are of high importance for a sustainable implementation of AI technologies that allow for trust in the socio-technical environments in which they are embedded. There has been a lot of research on defining ethical guidelines and establishing a general framework for a human-centric, ethically correct approach for the implementation of AI systems. Nevertheless, very little to no research has been done on the nexus between guidance documents and concrete implementations in companies, focusing on the impacts on affected employees and what measures companies can take to ease the transition process from known technologies and processes to novel AI technologies, providing a bottom-up, change management-focused approach that identifies human-focused problems that arise during such an implementation. In this paper, we bridge the gap between guidance documents, focusing on the Ethics Guidelines for Trustworthy AI and a concrete implementation of AI technologies at a large railway company, affecting several hundred employees. Using qualitative research - in concrete purposive sampling, semi-structured interviews and a document analysis - we identify relevant stakeholders and analyse the impact of the introduction of AI technologies across affected subcompanies on fears, expectations, experiences and ethical concerns. Based on these results evaluated through a structured content analysis, we define concrete short-, medium- , and long-term change management and ethically responsible measures with KPIs to enable a human-centric and trustworthy integration of AI. Following communicative validation, triangulation, we validate the results according to the quality criteria by Mayring.