Curtis-Trudel, Andre E and Roe, Niall and Voudouris, Konstantinos (2025) How deep learning can justify pursuit, and why it matters. [Preprint]

Text DLJP Oct 2025.pdf Download (366kB)

Abstract

We are in the midst of a deep learning revolution in science – or so many scientific and philosophical commentators would suggest. This article addresses a relatively underexplored aspect of this putative revolution, concerning how deep learning models (DLMs) impact the decision to rationally pursue a scientific idea. First, we develop an economic model of pursuitworthiness and use this model to analyze two ways that DLMs can justify pursuit: (i) by increasing the expected epistemic value of pursuit, and (ii) by decreasing the expected practical cost of pursuit. Then, we put this analysis to work. We argue (i) that it clarifies the sense in which DLMs may be said to be revolutionizing science, by radically impacting the economics of scientific activity, and (ii) that it brings into sharper focus certain scientific risks – what we call ‘illusions of pursuitworthiness’ – incurred by the shift toward DLM-driven science.

---

Export/Citation:

EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL

Social Networking:

Share |

---

Item Type:	Preprint
Creators:	Creators Email ORCID Curtis-Trudel, Andre E ae.trudel@gmail.com 0000-0002-2008-6134 Roe, Niall niallroe@gmail.com Voudouris, Konstantinos k.voudouris@helmholtz-munich.de
Keywords:	deep learning, economics of science, C.S. Peirce, pursuitworthiness
Subjects:	Specific Sciences > Artificial Intelligence General Issues > Technology
Depositing User:	Andre E Curtis-Trudel
Date Deposited:	21 Oct 2025 12:32
Last Modified:	21 Oct 2025 12:32
Item ID:	26983
Subjects:	Specific Sciences > Artificial Intelligence General Issues > Technology
Date:	October 2025
URI:	https://philsci-archive.pitt.edu/id/eprint/26983

Monthly Views for the past 3 years

Monthly Downloads for the past 3 years

Plum Analytics

Actions (login required)

View Item