PhD Student

Toward a new philosophical perspective on computational sciences: exploratory approaches to astrophysics and astrochemistry  

In the domains, where scientists work at the extreme limit of what can be achieved from a theoretical and experimental point of view, it is of the utmost importance to be able to validate their results, i.e., to compare them to observations, such as to assess and quantify their accuracy. Very often, however, these domains are also those where observations tend to be either sparse or uncertain. This is the case, e.g., when astrophysicists appeal to high-resolution numerical simulations to understand the evolution of galaxies, or when astrochemists want to know whether the interpretation of the spectroscopic observations allowing to detect and characterize interstellar molecules is done on the basis of reliable theoretical data. These epistemically opaque contexts require to identify and assess the impact of the main sources of uncertainty possibly distorting the model’s outputs, to make sure that an agreement or disagreement with observations can be properly interpreted and exploited to further guide the model’s development. The problem is all the more complex that developing models in high uncertainty contexts is often done iteratively, on the basis of successive approximations that permit to enrich the model and to determine its most fundamental ingredients. In this situation, a disagreement between predictions and observations is not necessarily a symptom of the model’s failure, but rather a guide to improve its empirical adequacy. These scientific domains reveal a real lack of epistemological tools that would allow to describe and analyze this science that we would like to refer as exploratory, whose aim is not to compare highly accurate predictions to well-understood observations, but rather to maximally exploit sparse empirical data to constrain the model’s development. The goal of this doctoral project is to start building a new philosophical framework able to characterize the progress and success made by exploratory approaches, notably through the following research topics:

• assess the reliability of theoretical and experimental data that are neither reproducible in the traditional sense, nor directly comparable to observations; 
• Formulate evaluation suitability criteria for models developed iteratively;
• Analyze the possible contribution of Machine-Learning based techniques to these exploratory approaches.

The thesis must  contribute to the development of new methods of confirmation, exploring the logical space existing between traditional methods of confirmation, verification or validation, implying abundant or non-problematicobservations, and non-empirical confirmation methods that seek to fully dispense with them. Such new methods are meant to capture and best support the practice of scientists working in these fields. This thesis is part of an interdisciplinary project, reuniting philosophy of science and astrochemistry. The selected candidate will therefore benefit from co-supervision in these two areas and be trained in astrochemical methods.

Pré-requis

• Training in philosophy of science AND in physics or chemistry is highly recommanded. 
• Programming skills (Python, Fortran) could be recommanded depending in the research project that the candidate will defend.