Søren From Soelberg M.A.
Dept. 2: History and Social Sciences
Institute for Philosophy, Department of Philosophy of Technology
work +49 6151 16-57579
- Philosophy of (explainable) AI
- Philosophy of science
- Philosophy of mind
Towards a philosophy of explainable artificial intelligence for its use in critical infrastructure systems (working title)
Artificial Intelligence (AI) has returned with a vengeance. Countries around the world are scrambling to take advantage of a technology that is estimated boost the economy and change fundamentally both industrial production and the way city infrastructure functions. Rather than being programmed by hand like the AI systems of the last century, contemporary AI consists of deep and elaborate artificial neural networks that are trained on very large datasets to automatically find optimal solutions for complex problems. While the resulting AI models match or even surpass human capability, the inherent complexity of these “black boxes” make them difficult to understand, which is a major stumbling block to the broader application of AI in sectors such as critical infrastructure, healthcare and defence.
This dissertation is a contribution from philosophy to the emerging research field of “Explainable Artificial Intelligence” (XAI), which is tasked with “opening” the black box by developing new technical methods and (re-)establishing theories, definitions and desiderata for AI explainability. Whereas XAI is an interdisciplinary enterprise that involves computer science, the social sciences and the humanities, philosophy has, but a few notable exceptions, yet refrained from engaging with XAI. There is much work to do, especially regarding clarifying explainability as a concept and explicating what constitutes appropriate application of AI in critical sectors.