About
Active since 2023, the DFG-funded Emmy Noether Group UDNN: Scientific Understanding and Deep Neural Networks seeks to analyze the philosophical implications of the ever-growing use of Artificial Intelligence (AI) methods in science. UDNN will focus on the use of Deep Neural Networks (DNNs) in fields such as particle physics and protein biology, and investigate its impact on scientific understanding. We will also explore the notions of „explanation“ employed in the field of eXplainable AI (XAI) and relate them to philosophical concepts of explanation and understanding.
Our work is guided by the following objectives:
(O1) analyse the goals, types, and success conditions of explanations in XAI, against the backdrop of the philosophical debate on scientific explanation and understanding
(O2) analyse the prospects, implications, and limitations of understanding without explanation from scientific uses of DNNs such as unificatory or objectual understanding
(O3) explore the possibility of novel types of scientific understanding and explanation present in XAI and in science using DNNs that are thus far not recognized in the philosophical debate
(O4) explore the epistemic implications of a possible lack of understanding in the face of predictive successes by DNNs in science
We approach these objectives by investigating the possibility of gaining scientific understanding along the following three paths:
Furthermore, following (O4), we will investigate the impact of potential limitations to understanding in DNN-heavy research on science’s epistemology, given that understanding is often recognized as science’s overarching epistemic goal.
