Science is getting faster, but not that much faster

In previous blog posts, I’ve described my view that the pace of science is becoming much faster over time. Here’s some data to suggest a different story.

These two graphs are from Our World in Data, based on World Bank and UN data. The first graph shows how the number of scientific papers published each year has grown over time. The second graph shows the same number, but standardised by the human population size.

Articles published each year increased from 0.98 million in 1996 to 2.93 million in 2020. This gives a ratio of roughly ~3.0x.

Articles published each year per million people of human population increased from 169 in 1996 to 374 in 2020. This gives a ratio of roughly ~2.2x.

So, there are about 3 times as many papers being published each year compared to a few decades ago. Some of this trend appears to be due to the fact that there are simply more humans than there were a few decades ago, meaning that society simply has more scientists. This explains part of the trend (the slope of the second graph is smaller, but not by much). The majority of the trend seems to be that each scientist is publishing more papers.

But it’s only a 3x increase! This is less than I had previously been assuming. And “papers published” is an imperfect proxy for scientific knowledge; the increase in “publish or perish” pressure at many universities around the world today means that academics tend to be incentivised to publish in smaller chunks (though I also think there are legitimate reasons why academics might want to publish in small chunks, like reducing the duration of feedback loops). For sure, there have been actual productivity gains; in particular, digital systems make the process of doing research, writing papers, and publishing papers much more efficient, which means that each scientist can produce more work than an identical scientist working 30 years ago. So we might conclude that the actual amount of science happening is indeed greater than it was in 1996, but that the ratio is less than 3x.

So scientific development is faster than it was 30 years ago, but not that much faster (maybe roughly twice as fast), and this seems to be attributable partially to more humans and partially to more science-per-human.