Memo 01.31.2020 5 minutes

If America wants scientific innovation, fix the R&D problem


The conditions for success simply no longer exist.

In 2018 Dr. Frances Arnold and two colleagues were awarded the Nobel Prize for their work in the chemistry of enzymes. Two years earlier her pioneering work in directed evolution, a novel way to synthesize enzymes for specific purposes, had won the Millennium Technology Prize. Following on the Nobel honors, Arnold co-authored a May 2019 paper on a specific enzyme evolution process, published in the prestigious journal Science, with Inha Cho and Zhi-Jun Jia.

This January she formally retracted the paper, a stunning reversal. And therein lie clues to the challenge of restoring innovation to U.S. research and development, a restoration that David Goldman eloquently called for in his response to Marco Rubio’s National Defense University address, which in turn called for a new industrial policy for economic and defense needs.

Worldwide, scientific research is beset with two related problems that have become systemic. The first is a crisis of replicability of published results. While some results, especially in the social sciences, suffer from poor or ambiguous experimental design and other possibly inadvertent defects, there is growing evidence for a deeper lack of integrity among a dismaying number in the research community. In the case of Arnold’s retracted paper, a review of the work showed that Cho’s lab notebook was missing “contemporaneous entries and raw data for key experiments.”

Arnold apologized for the need to retract, noting that she “was a bit busy when this was submitted” and that she “didn’t do my job well”. That job, as the senior researcher on the paper, was to oversee her lab and the work of junior researchers in it, especially when her name would appear as a co-author in research journal papers.

And that in turn calls attention to a deeper, structural problem in current R&D here in the United States. Complex and expensive research facilities, overwhelming emphasis on publication records for academic hiring and promotions, and a massive federal research grant apparatus have combined to incent such practices as citation networks—the practice in which researchers heavily cross-cite each others’ work and thereby artificially boost the perceived scientific consensus and the perceived expert status of given researchers as key authorities on various topics.

Hence the challenge for restoring American innovation.

Unshackling Strategic Innovation

Goldman’s analysis of our competitive disadvantages in the face of Chinese policies and his identification of innovation as our potential strategic advantage are deeply insightful. But what he does not address is how we can activate that advantage in today’s R&D context.

What has changed since the Apollo program and the Strategic Defense Initiative, two efforts Goldman cites as past successes? First, both of those programs addressed problems that were generally seen as acute national concerns: to reach space before the USSR could dominate it and to be able to defend against strategic nuclear attacks. To succeed, both programs required major, sometimes hard-to-predict breakthroughs—but for specific identifiable challenges. The result was sufficient focus to exclude generic career process research, while demanding innovation and insight sufficient to bring unanticipated benefits when commercialized for other purposes.

Second, both the Apollo and the SDI programs were executed outside of today’s academic research culture. Even when basic research occurred in academic settings, it was focused on challenges that demanded the very best from the best minds in their disciplines. And that also limited the self-serving tendencies of bureaucratic grant program managers whose incentives can be counterproductive to true innovation.

Third, both of the programs Goldman praises drew on researchers and engineers with rigorous education and training—professional status that was highly respected by the public and therefore attracted excellent minds willing to discipline themselves to tackle time critical, technically challenging work.

When Vannevar Bush convinced President Roosevelt to create the Office of Scientific Research And Development during WWII, the new federally funded agency had an urgent focus: find ways to enable the U.S. to effectively fight against the Wehrmacht. As the first OSRAD director, Bush reported directly to FDR. The results were both impressive and key to victory in the war: radar, significant airplane design improvements, and more.

When the war was over, Bush successfully lobbied for the creation of the National Science Foundation (NSF). And within the Defense world, the later threat posed by the Soviet launch of Sputnik spurred creation of the Defense Advanced Research Projects Agency (DARPA). Whereas NSF is at least in theory primarily focused on basic research, DARPA seeks solutions to pressing defense needs at the early applied research through proof of concept stage. Programs that succeed often then form the basis for full DOD system development by commercial contractors.

Successful innovation emerged from DOD R&D precisely because DOD often had very challenging needs that were nonetheless perceived as existentially important. Only later did commercialization of the results filter into day-to-day products.

Matching Researchers to the Moment

When I wrote one of the early Internet protocol software stacks, few of my company’s clients saw a pressing need for it. They were deploying either proprietary capabilities from companies like IBM or were attempting to implement cumbersome industry standards reached by slow consensus (the OSI protocols, for those interested in historical odd topics).

But DOD had a pressing problem: how to ensure the ability of senior commanders and their control systems to communicate in the event a strategic nuclear strike destroyed part of the U.S. mainland phone lines? The resulting TCP/IP approach to communications was nimble and able to work around major infrastructure disruptions.

DARPA funded multiple demonstration implementations of the technology on different computing platforms and showed that those could communicate seamlessly. At the same time graduate students in computer science and engineering were starting to use UNIX, an operating system that was more or less non-proprietary and that also executed on a variety of hardware platforms. When the combination filtered out into the commercial world after the success of the ARPANET linking several research labs at universities across the country, the basis for the public Internet was in place. Add in deregulation of telecommunications companies plus advances in microchip capabilities (also spurred by DOD), and the 1990s saw an explosion of innovation that changed our world.

If we are to successfully launch urgent R&D efforts toward strategically important innovation today, we must re-create the conditions for success. Those do not, as a rule, exist in our current academic and federal bureaucracies, which are not driven by urgent, focused existential challenges that come close to those that motivated the Apollo and SDI programs.

What is needed is twofold. First, several key research foci must be identified as having just that degree of difficult but urgent importance for us as a country. And second, we must create organizational and funding spaces where a small group of highly skilled and motivated researchers can address them. Those most likely will need to be at least semi-independent from existing academic and federal bureaucracies if they are to succeed.

I hope that the thoughtful minds among our political and scientific leaders will rise to the occasion. The alternative is dire.

to the newsletter