If humans stopped imitating, civilization would vanish.
It is 6:37 in the morning as I’m sitting down to write this, late in August, and I and the three-year-old are down in the old stable we use as study. As I look over my notes, she observes me. Mist is leaking through a crack in the window. Then she gets up, finds a pocket calculator, rummages around with the notes on the table a bit, and starts typing.
“The unicorn ate the human beings”, she says aloud. “Like wheat.”
Human beings cannot help picking up skills if they live immersed in communities where the skills are practiced.
If that statement is correct, or at least a good first approximation of the truth, as I believe it to be, the aim of education is to make sure as many people as possible can immerse themselves in skillful communities. We should, as far as possible, extract learners out of schools – where the incompetent live surrounded by, and therefore imitate, others at the same level of incompetence – and into productive environments.
Historically, this approach to learning has been the norm. Our grandparents lived in direct contact with the environments they would have to navigate as adults, and they could observe and imitate those who had already mastered it. But as modernity, working like a giant centrifuge, started separating everything that in traditional societies co-mingled, it became impossible for children to learn through pure imitation. They were segregated by age groups and denied access to the offices, factories, workshops, stores, whose skills they would have to master to become successful adults.
But maybe we can return to our premodern condition? Maybe we can build systems that allow novices to reintegrate themselves into communities of practice - but at scale?
It is a question of architecture, and can be phrased as such: How do we design workspaces where novices can enter at scale, and where they can easily observe the knowledge that guides experienced practitioners? And how can we do that without obstructing the work being done?
Historically, the rooms we’ve been able to build have been too limited for this. They put a low upper bound on how many novices could be absorbed by productive communities. Peter Paul Rubens, while developing his baroque masterpieces in the early 1600s, could only allow a handful of apprentices in his atelier. How would it have looked with hundreds! The demands of the incompetent would have overwhelmed Rubens, making work impossible, turning his workshop into a school!
In scaling access to valued knowledge, states, starting in the 18th century, opted for specialized learning environments. Fenced-in houses where novices lived segregated from the environments they were trying to master – called schools. By severing the bond between novices and masters, schools allowed us to increase the spread of select knowledge.
But schools have frustratingly slow rates of knowledge transfer and rapid learning loss. And they are marked by various psychological issues: disorders, boredom. It seems sitting in long rows of benches studying books is a bad fit for a species with learning instincts attuned to experiential learning.
We could have apprenticeships, which were effective but didn’t scale, or we could have schools, which were ineffective but scaleable. We couldn’t have both. This was the trade-off we faced. We choose school.
But the constraints that forced that trade-off on us are being overcome. We no longer need to separate novices from the environments they are to master. At this point in our technological development, it is possible to architect workspaces where novices can learn by doing, and more importantly by doing in the presence of competent practitioners. We can construct digitally mediated rooms where experts can coexist with large numbers of novices, allowing for a return to premodern forms of knowledge transfer. In this essay, I will explore how that would look.
Humans are Culture-Seeking
If you measure two-and-a-half-year-old children against chimpanzees and orangutans, they are about even in their capacity to handle tools and solve problems on their own. Only when it comes to observing others and repeating their actions is there a noticeable difference. Humans rapidly internalize the skills that surround them, other primates can’t compete.
There have been cultures that have used this, our human propensity to imitate, as their sole learning strategy, stripping away all didactics. When Alan Howard studied the inhabitants of Rotuma, an island in the South Pacific Ocean, he noted that they had a different mental model of how children become competent compared to the West. They did not talk about raising children, but rather about something similar to what Fiske has called "culture-seeking". The Rotumans did not see children as people who needed to be brought up - they did not need to be helped along by adults - but were more akin to some kind of dogs. They sniffed out what was useful and valued in the culture, and then they devoured it. They believed our instinct to imitate was enough to turn children into adults. Teaching was discouraged.
When discussing contemporary hunter-gatherers, Suzanne Gaskins and Ruth Paradise observe that in the ethnographic record there are virtually no documented cases of hunter-gatherers actively teaching. (This claim is from The Anthropology of Learning in Childhood.) It seems broadly correct. The exceptions I’ve come across are Inuit mothers pointing out relatives, expecting their toddlers to produce the appropriate term; various groups living in rainforests make displays of poisonous mushrooms to teach kids what not to eat; and among Hadza, women, upon first menstruation, are brought out into the bush to get lessons about magic and mythology. In all other instances, learning is through osmosis, through observation and imitation. To whatever extent contemporary hunter-gatherers can be used as a proxy for our evolutionary environment, this indicates that we’ve likely adapted to learn in this way.
We also see the same thing in most, but not all, agricultural societies. Here’s an account of sorghum growing peasants in Ghana written by Meyer Fortes in 1938, which could as well have been written about my grandparents, peasants in 1930s Sweden:
Nothing in the universe of adult behaviour is hidden from Tallensi children or barred to them. They are actively and responsibly part of the social structure, of the economic system, the ritual and ideological system. […] Hence the children need not be coerced to take a share in economic and social activities. They are eager to do so.
The mental model of childrearing in 1930s Taleland was similar to that of the Rotumans. What governs what children learn is not what they are taught - that matters less. What matters is what culture they are in. The more time they spent near competent role models, the more effectively they learn. Peer group is everything.
This is incidentally the polar opposite of the mental model of adults in rural Sweden in the 1990s, when I grew up. They focused on content – on curriculum and compliance – but were entirely oblivious of the culture they created, the culture which then created us: that of the schoolyard.
It wasn’t a very successful model.
Structural Constraints on Learning
We need to be surrounded by the skills we are to learn.
For this to work, the environment must be structured in the right way. It must be possible for us to observe the skills being used in meaningful contexts. The environment needs to be a glass box, as opposed to a black box; it must be possible to observe the workings without breaking it. The children of Rotuma could do this; I could not.
Gaskins and Paradise:
While learning through observation is clearly a universal characteristic of our species, we argue that some cultures because of their understandings of childhood and learning, provide environments that maximize the opportunities for this kind of learning and lean more heavily on it as a tool of cultural transmission.
In cultures where adults are unwilling to teach, it is common to see people alter how they work to make it more transparent: Aché hunters walking less briskly to accommodate less experienced animal trackers; Mayan basket weavers sitting on the ground to make it easier for their daughters to observe. This is a glass box approach.
We too often see the opposite in modern vocational education. Though explicitly dedicated to learning by doing, workplace learning often implicitly rejects this model in favor of didactic schooling. Instead of putting novices in free contact with the environment, trusting their instincts to imitate, educators tightly control the transfer of knowledge through a curriculum. This isn’t just ineffective. It can also, just as soon, turn into a nightmare dreamed by Kafka. Have you ever had that experience? Employed in my early twenties as a quality controller at a medical factory, I had to spend four nights – by which I mean 7 pm to 7:20 am – in a windowless room, reading detailed instructions on how to safely wash my hands.
Learning in a windowless room: that is the black-box approach.
That’s the worst example I’ve experienced, so it's not a representative sample of contemporary vocational training. But milder versions of it are a staple in the literature on workplace training.
In a study of 1970s meat cutters, quoted in Lave and Wenger’s classic Situated Learning, Hannah Marshall observed that apprentices spent most of their time doing bookwork. The practical parts of the meat cutter education were conducted in “shop”. Which was a staged supermarket. They were taught, among other things, how to sharpen knives. In case you didn’t know, supermarkets have sent their dull knives away for factory sharpening for the last seventy years. When certified by the sharpening teacher, the apprentices were placed at the wrapping machines in the supermarket. That is: they were placed where they couldn't observe what the experienced meat cutters did.
For apprentices to learn effectively, educators must, in a sense, let go of control. Instead of controlling the transfer of knowledge through lectures and curricula, they need to focus on ensuring apprentices have meaningful access to the environment, handing out appropriate tasks, and trusting the learner to extract knowledge out of what is going on.
This is a hard enough problem in itself. To get access, the novices will have to coexist with the masters - and do so without impeding the work. If done naively, giving novices free access to experts – to observe and act alongside them – quickly exhausts the expert's attention, making work impossible.
This is the problem of scarcity of expert attention.
It can be seen in open source. The open source movement has been important in spreading access to code, but also to the knowledge needed to produce it. Studying a repository is not quite being invited into Rubens atilie, but it is closer than novices get in most other areas of the modern economy. Many, such as Twitter’s Jack Dorsey, claim to owe their programming skills to observing how experienced coders solve problems in open repositories and pitching in. But this spread of knowledge has come at the expense of maintainers.
Nadia Eghbal, in her book Working in Public: The Making and Maintenance of Open Source Software, writes:
In my conversations with maintainers [of popular GitHub repositories], I heard them express a genuine conflict between wanting to encourage newcomers to participate in open source and feeling unable to personally take on that work. Maintainers simply don't have the energy to onboard every person who shows passing interest.
No matter how hard they worked, most “still felt underwater in some shape or form”.
There is always a tension between giving novices access to work environments and doing productive work. For apprenticeships to work, the architecture, unlike that of the meat cutters, needs to be constructed in such a way that the novices can observe the experts, while at the same time, unlike GitHub, shielding the experts from excessive demands on their attention.
Those two things. Access for novices. A productive working environment for experts. How would such an architecture look? How do we build rooms that let novices and masters coexist at scale?
Stuck in the physical world, what we can do is limited. We can unlearn the worst fail modes of schooling; we can study the architectures of traditional workshops; we can try to divide novices into smaller groups so that their numbers do not overwhelm the experts. I’m not too optimistic about reforming the world of atoms given our track record.
In the digital realm, on the other hand, almost all of the constraints that have limited us have been lifted.
Social Media as Architecture
If we consider the spaces we inhabit online as rooms, their architecture has far more degrees of freedom than physical rooms.
We can build rooms millions of people can enter at the same time. We can limit visibility to a select few – the person live streaming on YouTube, say, or the group invited by the moderator to turn on their microphones in a Clubhouse room. A novice observing an expert can stop time, reverse it, study difficult passages endlessly, or speed up to get past the fluff. He can create an alternate universe, by forking, a universe where he can break the master’s work into pieces, destroying it to see how it works. On Twitter, a user with a large following can rely on others to upvote the most interesting comments and questions, sparing her the burnout of trying to process all her replies. We can have algorithms, like Twitter’s, optimized for polarizing the rooms we work in, and we have algorithms that depolarize and filter for consensus like Pol.is’.
The constraints that made it hard for experts and novices to exist are less severe in the digital realm. The constraints that made schools the only viable means of mass education – we no longer need to let them limit our thinking.
Yet so far, online education has operated as if these constraints were still in place. The same way we use the QWERTY set-up on the keyboard – explicitly set up to make writing optimally slow, so as not to jam typewriters – we impose outdated constraints on our learning environments. We recreate schools in digital rooms.
Importing unnecessary constraints from the physical world is what Balaji Srinivasan refers to as scanning education, as opposed to building a digital native solution:
Balaji Srinivasan: So you can think of a piece of paper, and then you can think of a scanner and then you can think of a text file. Right. So you go from physical to this hybrid, like physical to digital scanner and then digital native. Another example would be you’ve got a face-to-face meeting, you’ve got Zoom, and then you’ve got virtual reality. Zoom is also like a scanner. It’s taking the offline and putting it online. But then VR is like digital native. With me?
Tim Ferriss: I am.
Balaji Srinivasan: Third example, physical cash, then something like PayPal or fintech, which is just basically the scanner, taking that and putting it online. And then you have crypto, which swaps out the backend is just now natively digital, it doesn’t have a physical form.
And with education, at least higher education, offline, you’ve got the physical college, then you’ve got the scanner, which is like Udacity, and Coursera and so on, which are fine companies, good companies, billion dollar companies, but they’re basically taking the college experience and putting it online. And then there’s this really interesting Terra Incognita of what does digital native education look like?
This new continent of learning is vast and uncharted. But one important trend is the return to immersive learning, which characterized the premodern world and is now growing back from the edges. This is a hopeful sign. To encourage it, we need to build the infrastructure that lets more people apprentice themselves to skillful communities online.
To get an idea of what that would look like, let’s return to open source.
Scaled Apprenticeships at Open Source Projects
The open source community is in many ways already a glass box, where a novice can look in and study the workings of the machinery. Anyone can inspect the code and read the issue log. You can fork the code and get your fingers busy, extending the program, using it as a Lego block, or breaking it apart to get a deeper feel for its dynamics.
But open source is not built to facilitate learning. Popular platforms, like GitHub, do not have the tools necessary to handle large numbers of inexperienced developers contributing to projects, nor the tools to facilitate their onboarding and learning or to shield maintainers from excessive demands on their attention.
For apprenticeships to work at scale, we need a better mediating layer. The mediating layer is the medium through which apprentices and masters interact. One example is the repositories themselves: these rooms that can branch off and multiply and that contain their history encoded in a chain of commits. Another is the bots that any maintainers use to answer frequently asked questions or handle pull requests. With the next generation of neural networks beyond GPT-3, we will likely see more mentor-type bots, scaling access to 1-1 mentoring.
A third type of mediating layer, not as widely used, is a video stream, letting people see the code development in real-time. This creates an even more high-resolution environment novices can observe. And you can go even further: developers can share not only their screen but their actual working environment, using tools like Replit. Novices can observe the experts working in real-time, but also at any moment jump in to code themselves, ideally in a way that does not affect the expert. One can imagine, for example, looking at a coding tutorial and then stopping the playback to test out the solution yourself before resuming to see how the expert solved it. Or, working in a live project, the novice could spin off an alternative universe through real-time forking, testing out some ideas or inspecting the code being built, before rejoining the main working space, where the expert has continued oblivious to the mess of activities occurring around her.
The mediating layer can also consist of human facilitators. One way to structure open source apprenticeships could be through cohort-based courses, similar to those of Lambda School and others, but arranged around real working projects instead of digital classrooms. A course provider teams up with a popular repository, taking care of the guidance needed to onboard apprentices, in exchange for being allowed to use the repository as a learning environment. Course facilitators help by answering the novice’s questions, handing out appropriate tasks, reviewing pull requests, and pointing apprentices to learning resources; thus guarding the maintainer’s attention. The maintainer gets a share of the course fee and free labor in exchange for sharing her screen, giving occasional feedback, doing a daily AMA or whatever. The apprentice gets high-resolution access to a community of experts and can later point to a series of valuable contributions to important projects when applying for jobs.
The apprenticeship can be financed by a course fee or an income sharing agreement, which could be kept lower than the fees of traditional education because the apprentice is providing value and can level up faster thanks to better input. In a web3 context, it might even make sense to provide this service free of charge - or to pay the apprentices for learning - to more rapidly scale up the number of skilled developers and thereby increase the value of the ecosystem.
These are just some first preliminary ideas. Open source will be a good testing ground for trying out and developing new modes of education. But it is not the endpoint. As the technology needed to scale apprenticeships matures, it will be possible to reintroduce immersive learning into more and more domains. At some point in the distant future, it might be possible for all learning, starting in childhood, to once again be immersive.
Three-Year-Old Cancer Researchers
It's been four weeks since I started this essay. Mushrooms the height of my daughter’s leg are growing along the path to the stable. In the last few weeks, she has grown increasingly obsessed with cancer. Drawing it, she makes it yellow and jagged, spilling out across the paper.
Zooming out, this is the context this essay must be read against. This is what I’m ultimately thinking about: why is there no way for her to use that passion? Why is it impossible for her to ride her current fascination into a productive research community, catching the wave before it passes, growing through observation and imitation, learning by doing?
A three-year-old doing cancer research is a vision on the borderland between the utopian and Monty Python; it is a vision to strive toward. As our technology stack matures, we should aim at including ever more people into productive environments, giving them the dignity of acting in the real world.
This ideas where developed in conversation with Johanna Wiberg and Torbjorn Elebjork. Miranda at LessWrong did the editing. I made all the mistakes.