Philosophy of Science in Light of AI

About

The scope, practice, methods, and theories of the sciences have changed in manifold ways in recent years because of the influx of new computational technologies. Algorithms now select experiments; theories are developed from massive datasets; insight has been gained on previously unanswerable questions; and much more. On one level, these changes are nothing new: science has always changed as the available methods and technologies have changed. On another level, though, the possibilities provided by AI--broadly understood to include machine learning, computer vision, robotics, and planning systems--have altered huge swaths of science in a relatively short time.

Despite these shifts, philosophy of science has arguably lagged behind in analyzing and incorporating these new methods and technologies. There is a significant need to advance philosophy of science in light of the widespread incorporation of AI into scientific research. We aim to assemble a collection of papers for a top-tier journal (preliminary interest from Synthese) that address the impacts of AI on philosophy of science, broadly understood. That is, we seek papers that advance, revise, or reshape “first-order” questions in philosophy of science--broadly inclusive of work on theories, methods (including formal epistemological analyses of scientific methods), practices, and other aspects of science--based on the ways that science itself has changed due to AI. (We are less interested in work that applies ideas from philosophy of science to better understand AI itself.)

A non-exhaustive list of indicative questions includes:

• Does the use of AI methods provide new insight into the possibility of a value-free ideal for science?

• How (if at all) do we need to change our understanding of epistemic risk in light of widespread use of AI systems for data analysis? Ought we use AI-derived results any differently in decision-theoretic reasoning about what to believe?

• What is the methodological and epistemological status of science conducted principally by machine (i.e., so-called automated science)?

• Do the products of AI & robotic systems provide novel insights into traditional focal concepts of philosophy of science, such as explanation, causation, or theories?

• Are “robo-scientists” relevantly different or similar with respect to the roles of measurement and instrumentation in the construction of scientific knowledge?

• How might the use of algorithms and AI systems in resource allocation (e.g., decisions about who receives grant funding) impact the structures and practices of various human scientific communities?

• How should the performance properties of formal methods (including AI systems) be incorporated into human scientific reasoning? For example, what role should asymptotic performance guarantees play in our scientific epistemology? Are AI methods different from traditional statistics in any meaningful ways?

Extended Deadline: Submit your abstract by January 15, 2020 using the google form below.

Organizers

Atoosa Kasirzadeh, University of Toronto; Australian National University

David Danks, Carnegie Mellon University

Sarita Rosenstock, Australian National University

Brian Hedden, University of Sydney