Progress studies folks often, and rightly, think about how to increase the productivity of scientific research. There is some reason to believe that the rate of breakthrough scientific discoveries may have slowed significantly in recent decades, to the detriment of long-term economic growth and human flourishing generally. Big questions about causes, in particular whether we have “picked the low hanging fruit” so that big new discoveries are just much harder to make than they used to be, are heavily debated but seem very much unresolved. But regardless of cause, a sustained slowdown in science would be a very bad thing for humanity— and if we could speed up science, that would conversely be very good.
What if we took a first-principles, input-focused approach to this problem? Here are some banal things that we know:
Research progress is made by researchers applying their efforts to unsolved scientific questions.
The capacity of people to do groundbreaking research work varies widely, both between different people and over the course of the same person’s lifespan.
Young adults and high-IQ people tend to be more likely than other humans to do groundbreaking research, all else equal.
The last of these facts, in particular, explains why some people worry that the levelling off of human population growth and IQ improvement— the fall in fertility rates below replacement in most countries and the end of the so-called “Flynn effect”— will produce a much less innovative, and thus economically and culturally stagnant or even retrogressive, future society. Smart, mentally nimble, energetic people are a fundamental input to research just as energy is a fundamental input to economic activity, and in both cases shortages bring strangulation.
I propose a natural measurement unit for the supply of this fundamental input: the quality-adjusted researcher-year, or QARY. Analogous to QALYs taking into account the fact that not all years of life are equally valuable, QARYs intuitively adjust for variations in effectiveness of research work. And though it is difficult, perhaps impossible, to quantify QARYs reliably, thinking in terms of increasing their supply can help us explore the space of potentially worthwhile pro-scientific interventions. I’ll sketch out a quick taxonomy below, so as to motivate folks interested in moving the needle on the rate of progress to think more holistically about how we might do it. At the highest level we can break down three types of interventions:
“Q focused” interventions that attempt to make each researcher more effective over the course of their career.
“R focused” interventions aiming to increase the raw researcher supply.
“Y focused” interventions to increase the length of research careers.
Less neglected avenues
AI
I mention this first because it’s trendiest: if we can augment human researcher brainpower with AI, that’s effectively a large QARY increase, focusing on the Q part. This is (rightly) already being explored extensively, as a quick websearch for “ai assisted research” will show, and I don’t have anything novel to add to that.
Metascience/institutional reform
Here is where a lot of progress studies effort has been directed over the past decade or so: theorizing on how we might fund people more efficiently so as to reduce the time tax of grantwriting, or change funding incentives and constraints to better enable more disruptive research, or otherwise coordinate science-related processes and institutions more effectively. If you want a pretty exhaustive exploration of the possibilities here, try this survey by Nielsen and Qiu; if you want an example of someone trying hard to put this technique into practice, Ben Reinhardt is as good as any.
This is also a Q-focused intervention, though there’s a possibility of increasing the R number as well if better scientific institutions make entering research fields attractive to more people.
Pronatalism
More kids now → more young adults 20 years from now → more potential researchers → less scientific stagnation. That’s the argument, and that’s maybe the purest R-focused ploy there is. How to get there, where the tradeoffs and ethical concerns lie, etc is a huge topic currently subject to extensive and furious debate, and commenting on it in detail here would derail this post, so let’s just acknowledge it and move on.
More neglected avenues
Maybe some of these are less neglected than I think; I’d be happy to be shown where I’m wrong.
Polygenic embryo selection
Embryo selection for polygenic correlates of general intelligence is probably the most politically controversial of the Q-focused interventions. It is also extremely likely to happen eventually, due to high parental value placed on having smarter kids. Could there be additional headroom to select specifically for correlates of doing groundbreaking scientific research? Maybe, maybe not, but seems worth looking into.
Finding and nudging promising people earlier
What if we got better at identifying who the “right people” were to make disruptive scientific progress, and offering them incentives and support earlier? Are the entrance criteria (like GRE scores) and incentives for graduate programs, for example, as predictive, selective, and appropriately welcoming as they could be? There’s R-focused headroom here in investigating that. It’s adjacent to metascience-type institutional reform, but most of what I’ve read in that literature seems more focused on better coordination mechanisms for people who have already gotten into a research field.
Speedrunning education
Now we get to the Y-focused interventions. Why don’t researchers’ careers start earlier in life? Largely because of how much time it takes to get far enough through the educational system that the state of one’s background knowledge becomes sufficient to do effective research. This is one of the main motivators for the low-hanging fruit hypothesis: it’s intuitively plausible that as knowledge advances, it takes longer to get people “up to speed” on what’s already been discovered so they can work at the frontier of new knowledge, thus leaving them with less time to make new discoveries before they retire.
But the length of time required to get educated enough to do research isn’t only a function of the extent of existing human knowledge: it’s also a product of the time-efficiency of the educational system. AI-assisted or otherwise tech-assisted education might help improve that time-efficiency, and so might institutional reform. Both are much-discussed topics in general, but both could probably use some more focus on helping the sorts of students who could do good research get prepared to do so more quickly. Even something as modest as converting K-12 education to K-11, or knocking the standard bachelor’s degree down from four years to three, could make a pretty big difference.
Extending cognitive healthspan
On the other end of the spectrum, there seems to be a somewhat robust scientific consensus that fluid intelligence— a crucial prerequisite for the creative thinking that leads to scientific breakthroughs— declines in most people starting in middle age, with the decline accelerating after age 60. This is probably a major reason why so many research fields are stereotyped as young people’s domains. Yet the rate of cognitive decline varies greatly across the population, and there are wonderful anecdotal examples of great researchers who remained productive well into their 90s, like John Goodenough and Etienne-Emile Baulieu. Making many more researchers’ careers that long-lived seems like a promising Y-focused goal, especially in the face of general population aging.
How to do that is a tricky, perhaps underexplored question. I’ve tried doing a literature survey here, and as far as I can tell there isn’t much that’s promising specifically for retaining fluid intelligence and researcher productivity after age 40 or age 60. Even early-stage studies are focused mostly on avoiding more severe forms of cognitive decline such as dementia, and there does not seem to be a clear consensus that anything really works, beyond the “boring” things we already know are good for healthspan-maximization in general: exercise, social connection, high-quality diet, good sleep, and the like. But if I had to pick one of the above interventions to be the beneficiary of increased investment, this is the one I’d pick— albeit with low confidence that I’m right, and keen awareness that, as a 45-year-old man who has always greatly valued my own cognitive sharpness, selfish interest probably biases me.
What else?
Are there other categories of QARY-increasing interventions that I’ve missed here? Please let me know if so!