Of course, others have argued for similar conclusions before. But the argument has usually been made anecdotally. We bring a new type of evidence to the discussion.
-
-
Prikaži ovu nit
-
The trouble with anecdotal argument about the question "What’s the rate of progress in science" is that it's too big and vague a question, and answers are easily swayed by feeling. That makes it easy to dismiss answers that you don't like.
Prikaži ovu nit -
To avoid this problem, we surveyed scientists at leading institutions, asking them to do pairwise comparisons ranking Nobel prizewinning discoveries in their disciplines. Eg: discovery of neutron vs cosmic background radiation? Etc.
Prikaži ovu nit -
We gathered 4,483 such comparisons. From this portfolio of questions we can back out progress in science (according to this metric) over the decades.
Prikaži ovu nit -
-
Here's the results for chemistry, and for physiology or medicine, showing, perhaps, a slight improvement:pic.twitter.com/9SJKd4j4xt
Prikaži ovu nit -
What's being plotted: the probability a discovery made in that decade is ranked above discoveries made in other decades.
Prikaži ovu nit -
The kicker is: the amount we're investing in science has gone up enormously (think 10-100x) over the same time period, whether you look at $, number of scientists, or number of publicationspic.twitter.com/9gBu2FREw0
Prikaži ovu nit -
We're in an age of diminishing returns to scientific efforthttps://www.theatlantic.com/science/archive/2018/11/diminishing-returns-science/575665/ …
Prikaži ovu nit -
Lots of disclaimers: this is just one metric, there's plenty of shortcomings of the metric. We're very aware of that and discuss this in the essay.https://www.theatlantic.com/science/archive/2018/11/diminishing-returns-science/575665/ …
Prikaži ovu nit -
Nonetheless, the conclusion should be taken seriously, not dismissed lightly. This is in some sense a collective judgement from scientists themselves: science is getting vastly more expensive, and far from accelerating, progress is at best constant (by this metric).
Prikaži ovu nit -
There's lots of corroborating evidence: e.g., the rise in ages at which scientists make key discoveries. Here are the average ages of discovery for early versus recent Nobel prizewinning discoveries. (Jones and Weinberg: http://www.pnas.org/content/108/47/18910 … )pic.twitter.com/vCsj7UNjhU
Prikaži ovu nit -
The rise in size of scientific research teams also suggests it’s getting harder to make discoveries. (Fortunato et al: http://barabasi.com/f/939.pdf )pic.twitter.com/MPtSlh36IL
Prikaži ovu nit -
There's also the massive decline in the growth of economic productivity since the 1950s, as documented by people such as Robert Gordon https://www.amazon.com/Rise-Fall-American-Growth-Princeton-ebook/dp/B071W7JCKW/ref=sr_1_1?ie=UTF8&qid=1542218574&sr=8-1&keywords=robert+gordon+rise+and+fall+of+american+growth … and
@tylercowen https://www.amazon.com/Great-Stagnation-Low-Hanging-Eventually-eSpecial-ebook/dp/B004H0M8QS/ref=sr_1_7?ie=UTF8&qid=1542218607&sr=8-7&keywords=tyler+cowen …pic.twitter.com/1ECYsSIQwv
Prikaži ovu nit -
In a similar vein, there’s the recent work by Bloom, Jones, Van Reenen, and Webb, suggesting that ideas are getting harder to find: https://web.stanford.edu/~chadj/IdeaPF.pdf …pic.twitter.com/cbrao49xf0
Prikaži ovu nit -
This all suggests it's getting much, much harder to make progress.
Prikaži ovu nit -
One response is to say "Well, it's inevitable that things get harder", and to shrug and continue on your way. But if something is requiring ~100 times the investment as formerly, it's a good idea to seriously consider whether it's possible to do better!
Prikaži ovu nit -
As far as we know there's no serious, large-scale, organized institutional response to the challenge of diminishing returns in science. And given that science is a principal driver of our civilization's progress, a high-order bit for humanity, that's a problem.
Prikaži ovu nit -
So, what to do? That's a subject for another essay (or multiple lifetimes of building). But I can't resist a few thoughts.
Prikaži ovu nit -
One huge success of science is how good it is at displacing ideas. If an individual or group has a new, genuinely better idea about the world, it can rapidly grow and displace old ideas. Evolution! General relativity! Etc.
Prikaži ovu nit -
It's mirrored in the business world: one person can start a business, and with skill and luck that business may grow to outcompete billion-dollar incumbents.
Prikaži ovu nit -
But suppose an individual starts a grant agency or university in their proverbial garage. They simply can’t grow it to outcompete incumbents ("We're replacing the NIH!" “We’re replacing Harvard!”), even if their approach is vastly better.
Prikaži ovu nit -
That is: there is no strong growth model or notion of competitive displacement for scientific institutions. And this means stasis and homogeneity and monoculture, a lack of organizational change and learning. This is terrible for science.
Prikaži ovu nit -
Indeed, it creates a sense that science _must_ be done this way. We must have PIs, a group is composed in such-and-such a way, scientists have a particular career path, are of a particular age, have a certain type of mentoring, produce a certain kind of output, etc.
Prikaži ovu nit -
But we could change each (or every!) one of these in radical ways.
Prikaži ovu nit -
Furthermore, it produces apathy. Every scientist has ideas for how to do things differently at the institutional level. But without a growth model for the best ideas, it's easy to feel it's not worth it, that things are forever stuck.
Prikaži ovu nit -
If you start a better grant agency, it's not going to displace the NIH. But perhaps it should.
Prikaži ovu nit -
A few ideas I like (no implied endorsement by Patrick, or originality on my part). Very telegraphic & incomplete - lots of nuance missing, and obvious problems that need to be addressed.
Prikaži ovu nit -
Figure out how new fields are produced. At the moment there's a _lot_ of inhibitory forces that slow the rate of production of new fields. Can we programmatically 2x or 10x or 100x the rate of new field production?
Prikaži ovu nit -
Far more varied funding strategies: eg by golden ticket (where 1 reviewer can ok a project, https://www.nature.com/articles/d41586-018-02743-2 … ); by variance in reviewer scores, using high variance (loved by some, hated by others) as a positive signal; or randomized allocationhttps://mbio.asm.org/content/7/2/e00422-16 …
Prikaži ovu nit - Još 5 drugih odgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.