Does a Better Science Need the State?

2144 California Street, N.W. (1951). NSF awarded its first grants from here.

At Macroscience, I’ve argued frequently for a vision of improving science that imagines a revitalized role for government. I believe we need courageous public institutions, armed with potent new policy tools, in order for us to have a well-functioning economy of science. I don’t believe that those institutions are a sufficient condition, but they’re certainly necessary.

But this view, about the importance of the state in the creation of better science, is of course not the only one. One alternative is that the reform of science is a project best taken up outside the bounds of a government-led paradigm. 

There is a lot to favor this position. State institutions are often rigid, loaded with historical baggage, and subject to political pressures that limit effective decision-making. In contrast, nimble, private funders freed from such constraints can rapidly launch and accelerate promising research efforts. Private wealth, philanthropic capital, and for-profit investment can also in many cases meet or even exceed the financial resources the government can bring to bear on a scientific problem.

Additionally, it may be the case that the big scientific progress these days happens in the private sector. Many foundational breakthroughs in promising areas like artificial intelligence and bioengineering are emerging from private enterprises and their labs, rather than through government sponsorship and support. Sure, we can tell tales about the many times pivotal innovations have had their deep roots in public funding and subsidy, particularly in fields like artificial intelligence. But, regardless of these origins, many of the major, present-day advancements surrounding these apparently linchpin technologies are now in private hands. 

All these factors suggest that if dynamism is anywhere, it is not in the government. We achieve better science faster by accelerating private efforts, not through the painstaking process of reshaping public institutions and bolstering state capacity. This view casts the government in the role of a secondary player. It helps marginally at best, and is an enemy of progress at its worst. 

I disagree. In fact, it is precisely the growing size and influence of private scientific enterprise that actually increases the need for public institutions to play a critical role. 

This partly has to do with the “animal spirits'' pathologies that can arise in any open market. As an intuition pump: do you believe that the immense interest in large language models, and the total cognitive energy they have absorbed, are an optimal distribution of focus across all possible research problems? My sense – even accepting that the influence of LLMs will be absolutely enormous – is still that there is overallocation. Markets can and do overinvest resources to “hot” areas, leaving potentially big gains in other research domains on the table. 

There’s a compelling approach taken in this dense but extremely fun 2021 paper by Hopenhayn and Squintani, formally modeling how competitive markets in R&D can systematically produce suboptimal overcrowding in hot areas. This suboptimality results in part from more researchers entering a hot area than is necessary to maximize the probabilities of a solution, and the switching costs imposed across committed researchers when a problem is solved in an overcrowded area.  

As the economy of science shifts to a scenario where more and more of the action takes place in an open market paradigm, it will be subject to herd mentality and monocultures in certain problems and methodologies. Add to this the classic underprovisioning of public goods in markets, and the government starts to have a crucial role in science that is growing in importance over time, rather than fading. 

It is worth underscoring the point that a crucial role for the government need not mean a highly interventionist role for the government. The era of ascendant private power in science and technology will limit the influence of centrally planned government efforts to drive progress forwards. It will also blunt the ability for governments to control or even be aware of key scientific developments, as I argued in a recent DARPA RFI with Caleb. 

Instead, issues like the problem of misallocation suggest a very different objective for the public institutions of science, one of counterbalancing. Governments might support Long Science when the market is producing Short Science, create incentive systems to combat intellectual monopolies, and offer support to diversify the organizational portfolio of science. The important role of government in this context is to secure competitiveness in the marketplace of ideas, nurturing the viability of alternative approaches, researchers, and institutional designs. 

Science is a dynamic system. The flow of knowledge, the communities of research, and the key fields and technologies change from era to era. This necessitates that we evolve the mental models and tools for governing science with it. One major point of inspiration in the present transition is political economy: as the fulcrum of scientific practice starts to look more and more like a market, there’s a wealth of thinking to be drawn from economic policy. A posture of “counterbalancing” starts to greatly resemble the way an institution like a central bank thinks about financial markets, and I think that’s no mistake. 

---

Thanks to Santi Ruiz and Caleb Watney, whose comments sharpened the argument in this piece significantly. Thanks too to Kyle Myers, who pointed me to the Hopenhayn and Squintani paper.