I study how climate scientists study the Earth’s climate. Particularly, I am interested in how scientists use data, including computational modeling and machine learning techniques. How do they manipulate data? What are the methods? And how do they draw conclusions? Can we trust the results and methods? Occasionally, I offer suggestions to scientists as well.

I mainly do conceptual work, rather than empirical work (the kind of work that has to do with implementing algorithms and analyzing data). But this does not mean that I do not care about hands-on experiences. Quite the contrary, I value these. I make up for my lack of experience by learning from and working with the best scientists.

Current projects

• “A Trip in Plato’s Cave: epistemic uncertainties of explainable Artificial Intelligence (XAI)”: XAI is a family of methods used to probe AI models, especially artificial neural networks. Some hope that XAI can help increase trust, but my coauthors and I argue that this is misplaced hope because of their inherent uncertainties.

Past projects

• “Machines learn better with better data ontology,” in which I link the age-old problem(s) of induction and cutting-edge machine learning. I lead the readers through Goodman’s New Riddle of Induction (1955). Induction is inferring the future or the unobserved from the past or the observed. It’s an interesting read if you are curious about “What does induction have to do with machine learning?” It will also be useful if you are teaching philosophy to students who are interested in AI. Check it out here: paper.
• “If a tree grows no ring and no one is around” deals with a debate in dendroclimatology (using tree ring records to reconstruct past climate). Scientists use tree ring data to calibrate other sources of scientific data, e.g., radiocarbon dating, and to corroborate climate modeling. But what if there were missing tree rings? Does this mean all our tree ring data are messed up? How would scientists find out? Check it out here: paper, talk.