Machine Learning and Signal Processing (MLSP) Seminars
2022-23 Semester II
Tuesday, 28th March 2023, 6:30 PM, at L-17, IIT Kanpur
- Title: Explainable AI between philosophy and computer science
- Speaker: Dr. Stefan Buijsman, Assistant Professor and Managing Director of the digital ethics centre, TU Delft
- Abstract: Machine learning is adopted across the world, and promises unrivalled accuracy. At the same time, we often cannot explain why these systems provide us with specific outputs, e.g. why a model rejects a loan application or classifies a certain picture as showing a wolf. In response, a range of fields has started to working on making these AI models explainable under the name XAI. I will approach these issues from a philosophical perspective, where the structure of explanations has been discussed for decades, and show how these theories can interact with the technical tools developed by computer scientists. There is a fruitful collaboration possible here, refining the work on explanation in both fields, and I hope to show this through highlighting work on causal explanations in both philosophy and computer science.
- Bio: Stefan Buijsman received his PhD in the philosophy of mathematics at age 20, from Stockholm University. Since then he has worked on the interaction between philosophy and other fields. First he focused on the interplay between cognitive science of mathematics and the philosophy of mathematics, around the question of how we learn math. Since joining TU Delft, where he is now an assistant professor and managing director of the digital ethics centre, he has worked on explainable AI and the interaction between philosophy and computer science. He has also authored several popular science books on mathematics and artificial intelligence.