University of Cincinnati Carnegie Mellon University

I am a philosopher of science working on the role of mathematical and computational methods in science, in particular, machine learning (ML) based methods. A number of questions animate my current research: What are the range of harms that can emerge from the deployment of ML in socially-sensitive contexts, and how can we most effectively prevent or remediate said harms? What is the role of ML in science, and what can this teach us about the nature of science? How do scientific applications of ML go wrong, and what do such misapplications reveal? How does mathematics apply to the world, and what is the epistemic value of such applications? How can a better understanding of the role of mathematics in science help us to improve scientific practice?

Mel Andrews

Making AI Beneficial

The methods of ML are rapidly taking over consequential real-world decision-making. The use of these tools in socially-sensitive contexts can, and often does, result in social harms, prompting the emergence of the field of ML ethics. There is, however, a gaping need for philosophical clarity within this new scholarship. Some of this must come from normative theorists, social and political philosophers, and, in particular, philosophers literate in legal doctrine. Equally important, and often passed over, is the need for philosophers conversant in the technical details of the ML-based tools utilised in consequential social decision-making; philosophers who are capable of clarifying the epistemic status of the methods of ML and differentiating between those issues which are irreducibly normative in nature and those which concern the functionality of these tools as means of securing knowledge about the world---facets often conflated in these discussions. Disputes currently being waged in the sphere of ML/AI ethics over predicting and apportioning risks and opportunities under conditions of uncertainty are not new. Many of them unwittingly recapitulate debates that took place within the quantitative social sciences in the mid-century, debates that occurred in the 19th century over the foundational assumptions governing social welfare and insurance, and debates dating back to the very origins of statistics and probability theory---domains which, since their incipience, have fundamentally concerned precisely the socially-oriented deliberations which currently animate ML ethics. The field of ML ethics therefore stands to benefit in critical ways from the perspective of history and philosophy of science and mathematics.

Machine Learning in Science

My research projects in this vein include 1. evaluating the supposed atheoreticity of machine learning methods when compared to the scientific use of classical mathematical or statistical tools, 2. investigating the ability of unsupervised learning (UL) methods, especially clustering techniques, to carve at joints of nature, 3. characterising DL methods as an applied mathematics singularly equipped to countenance complex phenomena, and 4. critiquing and offering alternatives to proposed epistemic norms for ML/DL (including explainability and interpretability). The first of these projects has resulted in a paper, The Devil in the Data: Machine Learning & the Theory-Free Ideal, currently archived and under review. The remaining projects are avenues of ongoing research.

The Use of Mathematics in Science

Traditional philosophical accounts of models and applied mathematics treat worldly phenomena and mathematical representations thereof as though they were, fundamentally, the same sort of thing---or sufficiently alike as to render subjecting them to straightforward comparison a sensible activity. From the confines of such a framework, philosophy renders itself incapable of differentiating between aspects of the reality science aims to approximate and features of the conceptual tools scientists wield towards these epistemic ends. The most prevalent and insidious errors made in applying and interpreting mathematics, errors of reification, flow from the inability to appropriately distinguish representational content from representational medium. And yet, from the vantage point of most going philosophical accounts of the math-territory relation, no such distinction can be drawn. Armed with traditional accounts of modelling, philosophers of science are therefore unable to weigh-in on the misuse of mathematical and computational tools. In contrast, the cognitive prosthesis account of applied mathematics I offer renders such epistemic errors visible and elucidates modelling best-practices. I develop and apply this account to many methodological disputes concerning the use of formalism in science in ongoing work. A chapter exploring the implications of my account for both first-order debates in philosophy of science and methodological considerations for philosophers engaging mathematical practice is solicited and in-preparation for a book entitled Philosophy of Science: A User’s Guide, edited by Sophie Veigl and Adrian Currie (MIT Press). An application of this view to a dispute in neuroscientific practice was put forward in a paper published in Biology & Philosophy: The Math is Not The Territory: Navigating the Free Energy Principle.



2022, Spring Term – Co-instructor
Philosophical Foundations of Machine Intelligence
Carnegie Mellon University

2021, Fall Term – Primary Instructor
Introduction to Cognitive Studies
University of Cincinnati

2021, Summer Term – Primary Instructor
Introduction to Cognitive Studies
University of Cincinnati

2021, Spring Term – Primary Instructor
Medical Ethics, focus on algorithmic injustice
& machine learning in medicine
University of Cincinnati

2020, Fall Term – Teaching Assistant
Contemporary Moral Issues, focus on Bioethics
University of Cincinnati

2020, Spring Term – Teaching Assistant
Introduction to Cognitive Studies
University of Cincinnati

2019, Fall Term – Teaching Assistant
Introduction to Philosophy
University of Cincinnati

2019, Summer Term – Instructor
EVOS Seminar Series
Focus on Natural History and the Tree of Life
Binghamton University

2017, Spring Term – Teaching Assistant
Child Development with David Henry Feldman
Tufts University

2018, Winter Term – Co-instructor
EVOS Seminar Series, focus on Science Communication
Binghamton University

2017, Summer Term – Co-instructor
EVOS Seminar Series,
focus on Biological Individuality & Identity
Binghamton University,



Mel Andrews
Department of Philosophy
McMicken Hall,
2700 Campus Way,
Cincinnati, Ohio 45221