Navigating the Bermuda Triangle of
Knowledge Infusion, Explainability, and Scientific Discovery

Internationally renowned experts in various fields will give keynote lectures on three pillars of machine learning for science: knowledge infusion, explainability, and knowledge extraction, featuring specific theoretical considerations as well as example applications. Inspired by the keynotes, we will discuss in plenary sessions and break-out groups questions like:

• How can we systematically integrate scientific principles into ML models?
• How do we ensure interpretability and reliability rooted in domain knowledge, also as a general means to make ML compatible with societal needs?
• Can AI generate new scientific insights rather than ‘just fitting’ data onto preconceived models? In which ways could this be thinkably possible?
• How can understanding actually be gained from an ML model, and how about quantifying it?
• What are clarifying principles of information processing and transfer, found in natural complex or neuronal systems?
• Which role will foundation models play in reconciling information extraction with generalization capabilities of ML, particularly in the context of natural sciences?

The unique format of the workshop shall carve out “blue skies” opportunities in an interdisciplinary context that wouldn’t have emerged in isolated fields. After a week of intensive debate, our goal is to more safely navigate the ‘Bermuda Triangle’ of scientific ML. We look forward to welcoming a diverse mix of leading experts and eager-to-share-and-learn participants to Stuttgart!

We envision this workshop as an interactive event with ~50 participants, including ~10 keynote speakers who will share their unique perspectives, raise urgent issues, and start the discussion of our overarching workshop questions. Participants will have the opportunity to present their research and thoughts on the workshop topics in dedicated poster sessions. They will engage in synthesis & brainstorming, panel discussions, and a walk & talk session in the nearby woods, to foster newly built connections and distill main findings from the workshop into directions forward.

We will have a diverse set of highly recognized researchers with different backgrounds, from various international institutions and of all career stages – check back soon for confirmed speakers! 

• Dr. Anneli Guthke, research group leader for Statistical Model-Data Integration (SC SimTech/University of Stuttgart)
• Dr. Miriam Klopotek, research group leader for Many-Body Simulations and Machine Learning (SC SimTech/University of Stuttgart)
• Dr. Amin Totounferoush, Postdoc at the Institute for Artificial Intelligence (University of Stuttgart)
• PD Dr. Eric Raidl, PI of the Ethics & Philosophy of ML Lab at the Machine Learning Cluster of Excellence (University of Tübingen) 

In the spirit of an interactive workshop, capacity is limited and hence we will perform a selection of participants based on motivation, background and fit with the workshop scope. We request a moderate registration fee. More information will follow soon.