The Pace of Scientific Progress
A Quest for Better Metrics
Executive Summary
This article challenges conventional approaches to measuring scientific progress, arguing that common metrics like patent counts, publication volume, and funding levels provide incomplete and often misleading indicators. It proposes conceptualizing science as a growing network of questions and answers, suggesting that the pace of scientific progress is best understood as the rate at which we meaningfully expand this network with emergent questions and answers. This framework can be applied not just to scientific disciplines but to organizational learning across sectors, including venture capital investment strategies.
How do we measure the pace of scientific progress? It's easy to resort to simplistic approaches that appear objective but ultimately prove unhelpful – focusing on parameters that can be quantified with a level of (often misleading) precision.
Patent Counts: Patents have become a very noisy measure, especially since they've acquired a purely defensive role in much of the world today – both in trade and in domestic disputes.
Publication Volume: The number of published papers in a field has become even more problematic now with the growing number of papers produced in part or wholly by LLMs in different ways.
Research Funding: Funding often can fuel progress, so it's not entirely wrong as a metric, but it's less exact than we would wish for.
Economic Indicators: Even noisier measures include productivity growth, GDP growth, or general economic metrics that might assess overall progress in some narrow sense. These metrics typically don't capture scientific progress well, though scientific progress does contribute to overall economic growth.[1][2]
Why is measuring scientific progress so difficult?
The first task we have to solve to understand this problem is defining what scientific progress actually is. It seems obvious when a vaccine for a major pandemic is developed or one of Hilbert's 23 problems is solved. But upon closer examination, we realize that these empirical measures lack uniformity and precision – making it difficult to pinpoint exactly what constitutes scientific progress.
We see this clearly when comparing different disciplines:
Is physics progressing faster or slower than chemistry?
What about biology versus mathematics?
Economics compared to literary theory?
Not to mention that while it's easier to define precise problems to solve in physics and mathematics, even top biologists wouldn't likely agree on a single list from a theoretical perspective. It's very hard to create comparable measurements across scientific fields – we end up with apples and oranges at best.
A New Framework: Science as a Puzzle
One way to think about scientific progress is that it's about completing a puzzle, expanding our understanding of the world and how it works. What we want to capture in any measure is the pace with which we are:
Adding new pieces to the right places
Finding entirely new pieces
As we slowly, painstakingly try to complete the puzzle, we are making progress in science. The pace at which the puzzle grows – new pieces are found and existing pieces placed in a way that slowly reveals the larger motif – is what we should be measuring.
"But this is hard. What, exactly, is a piece? And when is it placed in the right place?"
The Network of Questions and Answers
We can learn quite a lot from the work of Tsao and Narayanamurti.[3] In their exploration of technoscientific revolutions, they describe science as a network of questions and answers – a powerful metaphor for understanding progress.
What we really want to understand is the pace at which we are adding new questions and providing new answers – and ideally, we want both to be progressing rapidly.
Consider the alternatives:
A world that slowly answers existing questions – where more questions are answered than new questions asked – is probably slipping into "normal science" (in the Kuhnian sense)
A world where we're only adding new questions without answering them is one in which empirical work is neglected – equally unhelpful
Measuring What Matters
What we should be measuring, then, is the relative speed of emergent questions & answers added to this network. That is doable, if challenging.
Interestingly, we should also measure emergent questions & answers in our own organizations – since we are all in some way engaged in a scientific project focused on understanding the world through our own organizational lens.
Applied Example: Venture Capital
Take the field of venture capital, for example:
A VC that invests solely in areas it understands would limit scope and rule out new discovery, quickly becoming obsolete
A VC that only invests in new trends without proper diligence fails to learn because they're simply following the herd
The ideal is a nimble VC able to project and hypothesize as they go, valuing failures as much as wins
Learning Loop Model
A successful VC works to understand why their bets are succeeding or failing as the external environment evolves. Is it the founding team, the sector itself, a market failure, timing, or something else?
From a scientific perspective, the pace at which the VC learns is the true measure of progress – not simply returns – because then greater returns become replicable. A one-hit wonder investment isn't as impressive as a series of strong investments based on hypotheses that are constantly refined through an empirical feedback loop.
The pace of scientific progress is the pace of learning, which is why measuring it properly is so crucial. For more ideas on how this can be done, please see Nicklas’ book on asking questions, titled Frågvisare.
About the Authors
Dorothy Chou is the Director of the Public Engagement Lab at Google DeepMind, where she helps enable meaningful public discussions through translating complex AI concepts. Dorothy is passionate about using technology as a force for positive change, both through her policy work and as an angel investor supporting underrepresented founders. With interests spanning bioethics and technology governance, she enjoys building bridges between technical innovation and social responsibility, working toward a future we can look forward to.
Nicklas Berild Lundblad is the Director of Public Policy at Google DeepMind, where he explores powerful questions at the intersection of technology, policy, and society. He thrives on connecting diverse stakeholders around shared visions for AI's future, describing his work as "a mix of foresight, insight and listening." An enthusiastic ambassador for thoughtful AI development, Nicklas enjoys facilitating conversations that bridge technical innovation with social impact, finding deep satisfaction in building collaborative networks that shape positive technological futures.
Terra Terwilliger is the Director of Strategic Initiatives at Google DeepMind, where she brings her Georgia roots and down-to-earth perspective to complex AI topics. As a strategic thought partner to the COO, she finds purpose in building a shared imagination about AI-enabled futures. Terra is passionate about harnessing technology's potential to improve lives, working with diverse teams to ensure AI benefits humanity in meaningful ways.
The views expressed in this article represent the authors' personal perspectives and not necessarily those of their affiliated organizations.
© 2025 Google DeepMind
Sources
[1] Research by Ayers established that scientific and technological progress actually generates economic growth. The evidence indicates that human welfare increases primarily due to advances in science and technology, with economic growth being a secondary factor.
Ayres, R. U. (1996). Technology, progress and economic growth. European Management Journal, 14(6), 562-575. https://doi.org/10.1016/S0263-2373(96)00053-9
[2] The IMF has highlighted that innovation (particularly through R&D) is a driver of long-term productivity growth. The analysis shows that both basic and applied scientific research contributes significantly to economic development, with basic research having wide-reaching effects across multiple sectors and countries.
Choi, S., Furceri, D., & Yoon, C. (2021, October 6). Why basic science matters for economic growth. IMF Blog. https://www.imf.org/en/Blogs/Articles/2021/10/06/blog-ch3-weo-why-basic-science-matters-for-economic-growth
[3] See Narayanamurti, V. and Tsao, J.Y., The Genesis of Technoscientific Revolutions. Harvard University Press.
Director, Government Innovation @ Tony Blair Institute | Governing in the age of AI
2moThis is great. A relevant idea is Joel Mokyr’s concept of the cost of access to knowledge (especially but not only at the frontier), and his suggestion that the Industrial Revolution was made possible by a great reduction in those costs through the Enlightenment. Your “puzzle piece” idea fits really neatly with this – if more people can get to the frontier of knowledge, and add more rapidly to it, then we are on the cusp of another major reduction in the cost of access to knowledge.