Science is a people and a place at a time #

Academia has lost some of its lustre lately, and people are speculating about What Comes After the Academy.

I come not to bury the academy, but to examine it – and maybe even to praise it. Why did we decide upon this particular institutional form to house our scientific apparatus? Are universities generating a good ROI on or research dollars?

Universities have been quite productive: most of the fundamental advances in science and social science made in the 20th century, for good or ill, can be traced to the campus.

There are a few peer institutions we can use as comparatives: the various national labs (BNL, LLNL,LANL, Fermilab, etc.) the Manhattan Project, Bell Labs, Xerox PARC, Lockheed’s Skunkworks, Google, Deepmind, and OpenAI¹. But most of the public sector or commercial contenders for the Engine of Scientific Progress title are shorter-lived than Universities: Bell Labs was the longest lived, at ~80 years, while PARC was only active for three decades. The Manhattan Project was a goal-driven sprint.

I’d argue that that the University – particularly the American Land Grant University – is one of the most robust institutions for driving scientific progress yet devised. I believe that its power derives from accommodating the preferences of a particular type of people, a type that disproportionately makes scientific contributions. The University is the natural “container” for these people.

A University is like a city for the life of the mind. As NYC specializes in finance, LA in movies, Nashville in music, and San Francisco in tech startups, a university is a unique community form creating agglomeration effects for the life of the mind, a Levittown for cognitive endeavors.

That’s a neat trick! Why does it work?

How do universities work? #

Successful innovation machines are always attached to a money generator: for the national labs, this is explicitly the government (although the government is a direct or indirect sponsor of many of these entities), while for the private sector entities are supported by an enterprise enjoying outsized financial success.

Having a monopoly or virtual monopoly (Bell Labs, Google) or having the USGOVT as your top client (Bell Labs, Skunkworks) is helpful, as is being deemed critical to national security, especially in times of existential threat (Bell Labs, the Manhattan Project, Skunkworks).

Money and monopoly are necessary but not sufficient to keep our innovation machines running. What makes them work is that these elite institutions become a Schelling Point for smart people.

Like early 20th century Vienna, Universities became known for being where interesting people congregate. Preferential attachment worked its magic as their reputations grew. Pretty soon they had a pretty sweet agglomeration effect going.

Anyone who has tried to recruit top researchers to the private sector knows that they can be more reluctant to leave than a simple analysis of compensation/workload would predict. Part of this is probably due to the irreversibility of the decision: academia is an up-or-out profession, like the military, and once you’re out, it’s difficult to get back in. But the qualities of the academic life itself also play a role.

Science is driven by certain configurations of certain people with certain characteristics and those people prefer academia to industry. Why is that?

What is it about academia that scholars find so appealing? #

The top draw has always been the presence of other scholars, their colleagues. These colleagues are often also their neighbors, and opportunities for formal and informal socialization are more frequent than in many workplaces, even among other knowledge workers.

Part of the academic compensation package – becoming a member of an ivory tower elite – is very hard to emulate outside of the academy. Universities foster and promulgate an elite tribal identity (crests! scarves! mascots!) that has psychological and social benefits. Status is costly, and academics enjoy it at a discount. The physical surroundings of the university are often quite pleasant, with attractive housing and ample community services.

Academics also enjoy long time horizons to work on ambitious projects, and are not bothered by mundane financial and bureaucratic details. The prospect of tenure, a permanent position and the freedom to define a research agenda, is the capstone incentive².

Academics are surrounded by ambitious young people, and use their energy to create innovation flywheels. The presence of young people also provides opportunities for mentorship, which is very rewarding to both junior and senior academics³.

Can we replicate these features elsewhere? #

This resembles one of the core questions of metascience: what are the factors driving scientific progress? And how can we build institutions to accommodate and enhance them?

The obvious models for an innovation machine outside of the university system are the elite industrial labs of the past and present: Nela Park, Bell Labs, Xerox Parc, Skunkworks, Deepmind, Google, OpenAI, etc ⁴.

Interestingly, not every tech giant with a money engine develops an academic-adjacent reputation: Google did, but Apple and Facebook largely did not, even though they are prestigious places to work, and Apple especially has a reputation for innovation. Here the outside view may be informative: unlike Google, Apple never really developed a reputation for encouraging its SWEs to spend significant time on personal projects of uncertain business value (“20% time”). The impression that you can do unfettered blue-sky research without considering the bottom line is important. Openness to exploration is an important component of an innovation machine, and at some point on the explore/exploit curve institutions transition out of this space, even if they remain elite.

It seems that an aspiring innovation machine must be willing to disrupt itself. It is not sufficient to merely become best in class, or the best version of itself, it must be willing to usher in something entirely new (LLMs disrupting search, or the nuclear era disrupting conventional superiority). An innovation machine must treat OKRs as a convenient myth.

The innovation machine blueprint includes an elite, aspirational identity that attract the best and most ambitious. It may be physically secluded and include a residential component fostering community and self-assured elite identity. It must be well-funded, with high-risk ventures on different timelines regularly throwing off impressive results – even if these results are not the one initially envisioned by org leadership. And – if we take the example of Bell Labs and the midecentury US university as examples – this leadership should be of a piece with the research personnel, essentially ascended researchers with a strong research identity and orientation, still actively engaged in science and scientific mentorship⁵.

Academia was a Schelling Point for smart productive people. But all network effects are vulnerable to preference cascades. If enough elite performers decide that academia is no longer the best place to work, the network effect can unravel quickly. The creeping fear of every underpaid grad student is that those smart people are now somewhere else.