Why is it that structural change in “STEM education” has proven so elusive? What—if anything—might an ambitious foundation do about it?

In our capacity as an advisor to a philanthropic foundation, we had a series of conversations, both in person and over email, exploring opportunities for (and structural challenges to) transformative changes in “STEM” education. From our perspective, many foundations and initiatives have tried and largely failed to make a dent in this domain. This excerpt from an email exchange begins to detail why and offer very preliminary responses to these challenges.

Structural challenges in STEM

I thought it might be helpful to share some of our thoughts about what fundamental assumptions need revisiting if we're going to see any sort of transformative change in STEM. The overwhelming majority of work we see nowadays is extraordinarily incremental, and we think that some of these factors are at the root of that…

  1. Emphasizing content over skills — The most basic issue is one we touched on in our call: There is a very deep failure to recognize that to whatever extent "STEM" is a thing, it is overwhelmingly a set of mindsets, skills, and heuristics— not facts. This fact is reflected in the acronym itself. As long as the emphasis is on content over skills, at least two things will happen: (a) the emphasis in the design of learning experiences will be on communication rather than action/experimentation/creation, and (b) any attempt to broaden STEM to make it more relevant will founder on the shoals of time, because the assumption will be that we need to increase the information transmission, and the bandwidth just isn't there.
  2. Failure to acknowledge the basic time dynamics of STEM investigations — (1) — in concert with some basic assumptions of traditional school systems—inevitably leads to another failure mode: a total failure to grapple in any meaningful way with the structure of time required to engage the skills that go into "STEM." Specifically, because so much of what's involved in the cultivation of real STEM literacy comes back to actually doing something outside of yourself—often engaging a system or representation which is messy, unpredictable, or otherwise needs debugging—and that simply can't fit into a typical 40- or 80m period.
  3. Deep conflation of "easy" and "good" — To the extent that STEM has a serious issue with people's relationship to it, there's an incredible amount of flitting activity which attempts to basically remediate people's relationship to STEM through PR. Sometimes that PR takes an active form (e.g. "making"), however in general, there is a pretty deep assumption behind many efforts that the problem is that STEM is too hard currently. I don't mean to discount the importance of interpersonal or cultural buy-in to an activity, but this cough-medicine approach to reform ("a teaspoon of sugar makes the medicine go down") inevitably hides the actual, intellectual work...and worse, in many cases actively cuts against many of the habits of mind (e.g. debugging) which might be at the core of whatever we mean when we say "STEM."
  4. Failure to develop a human capital model — What people do in math class looks nothing like doing mathematics. The same is true of the vast majority of physics or chemistry or other STEM classes (especially at the secondary level). This is for many reasons; however, one of the deepest is the relative absence of people with real disciplinary depth. The reasons behind this are myriad, but I believe this is the most fundamental, limiting reagent in really transformative shifts in STEM education in traditional public school settings. Without an adult who themselves has and embodies the mindsets at the core of STEM, you inevitably backslide into a context where the focus is information transmission. Not only doesn't this work, but that mode is precisely opposed to many of the skills at the core of STEM.

Computation as an exemplary failure

All that's pretty abstract. I thought it might be helpful to flesh it out a bit with a specific example. Watching "school" attempt to figure out what to do about "computers" has been totally fascinating. It's especially fascinating because there is a burgeoning, parallel infrastructure out there for folks to grow as software engineers.

Let's start with just those four structural challenges:

  1. Emphasizing content over skills — You've got AP Computer Science and Exploring Computer Science, but overwhelmingly, we're doing to computers what we did to Algebra II. It's just another class. Anything subtler than this—e.g. using computers to revisit traditional ways of learning—are incredibly unimaginative, focusing on computers as a channel for traditional content as opposed to a new technology a la calculators or pencils. We were on the Massachusetts panel drafting MA's Digital Literacy and Computer Science standards (and are now on their Implementation Panel looking at licensure and PD), and it's a total farce. The entirety of the emphasis is on doing some weird ontologizing of "computer science" with an utter disregard for what actually learning and building with computers looks like.
  2. Failure to acknowledge the basic time dynamics of STEM investigations — Given your background as an engineer, you can appreciate how wildly disruptive it would be if I forced you to interrupt design sessions every forty minutes. That same background means you also understand how wildly time- and detail-intensive actually getting a built thing to work is. The same is true in chemistry or mathematics or whatever. Computers have some serious advantages here (copying is free and instant, trying again is free and instant, debugging is natural, you can inspect your system well, professional quality development tools are widely available, etc.), but any software engineer would wince at the time constraints of school. It's not an accident that so many people (still, today!) get into making things with computers in an afterschool club or through a personal project or a summer camp.
  3. Deep conflation of "easy" and "good" — This impulse generally coincides with a technocentrism that's understandably overrepresented amongst tool-builders =) If you look at efforts like Blockly or Tynker or Hopscotch or Hour of Code, the underlying premise of all of them is, "Programming is too hard. Let's reduce the friction." I'm all for reducing friction, but they fundamentally misunderstand the problem. Scratch comes closer, but I also don't think they're trying to make more software engineers, per se.
  4. Failure to develop a human capital model — I think that these broader STEM failures show up especially acutely in computational settings because everything is so abstract. Everything you do normally (lecturing and homework and tests) so obviously makes much less sense in building things with computers. You absolutely need a staff member who themselves can build things with computers, who has a real fluency. And overwhelmingly, schools don't have that. Consider the rebranding of CS10K (now "CS for All Teachers"). Its original goal, as you may recall, was to "develop an effective new high school curriculum for computing, taught in 10,000 high schools by 10,000 well-qualified teachers by 2015." Last year, ~58K students took an AP CS exam. Assuming a class size of 20 and discounting the folks who self-study, online-study, or otherwise take the test, that's still less that 3K classrooms (and presumably, teachers). Of course, AP isn't all of CS, but I think that persuasively demonstrates just how total of a failure CS10K was. (Why, given that, the scope was expanded to "For All" is beyond me.)

One of the reasons I think computation is an interesting example to think with is that the outside world has so rapidly caught up and surpassed school in its efficacy in developing people's fluency...in ways that has largely not been the case for, e.g., physics or mathematics or chemistry or engineering. I think there are a ton of lessons to learn from these successes, but one of the elements that those successes elide is the very real opportunity to substantively revisit what and how we learn in all of our other subjects. Just as pencil and paper and blackboard transformed schools, so can (I believe) computation. Obviously people believe this about "technology" (broadly construed), but this has not touched pedagogy at all.

Structural responses

So, what to do about all this? Obviously, we have a particular strategic take in the form of a school design, but that intervention was articulated with a much broader mission than "just" STEM education. So I thought it might be interesting and useful to share a few of our thoughts about what some of the ingredients underlying structural interventions which might be possible that don't take the form of a school and more specifically target STEM...There are plenty of much more concrete, exemplary initiatives/brainstorms we could share from this, but I thought it might be a good start...

So first, to go back to our four, structural challenges above:

  1. isn't fixable from the inside...
  2. isn't fixable in a traditional classroom schedule and structure...
  3. isn't fixable until less incremental successes are seen...
  4. is really the only meaningfully addressable issue.

And in a very fundamental way, (4) is the limiting reagent to broader changes. But I think why and how that's the case deserves some elaboration. Specifically, I don't mean that as narrowly as CS10K—e.g. for computer science—it isn't that there aren't enough "computer science teachers" (though that may be true, too). But if you zoom out and think more broadly about other ingredients... like policy changes or school designs or school leaders or curricula or tooling or anything like that, our bench is wildly shallow there, too. Until you have real people with a real relationship to their domain who are domain experts and practitioners first and designers and educators second, I think we're stuck.

The only way to work around that (e.g. in the interim) is to focus on efforts which in some way unbundle the time or resources school currently monopolizes to meaningfully allow third party partners and experiences to take on a significant role. This has proven to be extremely difficult in general. Vanishingly few schools or school systems are going to give up money or time when they see something as nominally within their wheelhouse.

Given this, you'd expect to see very interesting work around the margins—with at-risk/opportunity youth, adult basic education, vocational education, etc. And you do! But almost all of that is around computer programming (for a variety of reasons I'm happy to hypothesize about). That is, while there are coding bootcamps, there are no physics bootcamps for overaged/undercredited youth.

In addition to human capital, I think another wildly underappreciated limiting reagent is basic research and design. Right now, people solidly assume that all work needs to be done at scale, quite soon. For a variety of reasons, this precludes a lot of the much more basic research and design work that's necessary when a field is underdeveloped. That is, in education we basically assume all the fundamentals are set, and now all we need is help scaling/implementing. There are vanishingly few communities devoted to the equivalent of tech transfer/translational research, and even fewer devoted to basic research and design...Those who do research have to put up with researching existing schools, which limits the scope of their inquiry to unhelpfully narrow (e.g. fMRI studies) or broad (e.g. charter school efficacy research) questions.

I think that this means we need significant vision and investment in articulating a real research and design community... think Xerox PARC or Mayo Clinic or Dewey's Lab School. For whatever it's worth, I think that this naturally complements (in ways I'm happy to expand on) the human capital diagnosis. Roughly: If you're going to bring in or develop a deep bench of domain talent, that talent won't be happy to be dumped into existing schools, and only a few will have the patience/gumption to put up with the bullshit of starting a new public school. So, where do those people learn and grow and acclimate to education-the-industry in the interim?

As you can probably tell, there's plenty more where this came from (and I'm happy to share other, exemplary analyses/brainstorms we've drafted internally as we've thought through our longer term strategic plan), but this is probably enough for now =)

Interesting organizations

As I mentioned, I think the vast majority of the most interesting work in this space isn't happening in schools— It's happening out in industry and adult basic education. Here are some of the organizations I mentioned which I think are especially interesting and bear much closer inspection...in their own way, I think each is also very interesting as an inspiration by analogy:

  • 42 and the Holberton School — Both of these are pretty interesting examples of revisiting the essential characteristics of "college" (colocation, immersion) tied to a deep domain (in their case, computer science).
  • Recurse Center — I think the Recurse Center is one of the most fascinating organizations devoted to a vastly different model of learning. They're very thoughtful folks, and they aren't at all like a "coding bootcamp." I really recommend that you check them out and chat with their founders...
  • Flatiron School — The Flatiron School is—in my opinion—the best and most interesting example of the "coding bootcamp" as a form. They're always thinking about what that means, and I think have done a lot of good work, especially around training staff. You may also find their early experiences trying to connect with K12 (through PTECH and others) illuminating failure modes =)
  • MissionU — I'm very excited about the potential for organizations which aren't even trying to compete with school. Through new financing mechanisms, they get the opportunity to tackle some very interesting unbundling. I think one of their most interesting characteristics are that they are obviously woefully incomplete relative to "school," but I hope they portend an ecosystem which, once fleshed out enough, will be an obviously superior (if sometimes less legible) choice.
  • MSRI in general and Math Circles in particular — I think MSRI is a fascinating organization, and one that I think has a tremendous amount to offer thinking about what good environments for deep thinking and growth looks like (especially from a POV that is less concerned with content). Math Circles are one of the best examples I've seen of the mindsets-over-content disposition in a traditional setting, and I think they're underappreciated and unfortunately limited to a luxury/enrichment opportunity instead of a design inspiration.
  • Startup School — In your neck of the woods, I think that the whole world of how folks learn "startups" is fascinating (relative to, e.g. traditional management and business experiences) for reasons very similar to what I find so inspiring about the world of software engineering "education" (broadly construed). The important part to me about Startup School (or Y Combinator) isn't the videos (though that's good too), it's things like where these videos come from (practitioners), whom they're aimed at (practitioners), what grounds them (an emphasis on practice over preparation), etc. I think that questions like, "What might Startup School for X look like?" are very interesting.

Further readings

As promised, I also wanted to send along a few readings that are either exemplary of or have been formative in how we think about "STEM."

  • Mindstorms — If you have to read just one thing on this list, I couldn't recommend Seymour Papert's Mindstorms more highly. I think it's one of the deepest pieces of work about how people think and learn, and his vision is still nearly completely unrealized and unappreciated now.
  • "An exploration in the space of mathematics education"  — Also from Papert, this article I think is a very concrete example of what it might mean to substantively revisit how and what people learn under the assumption that you take computers seriously as a tool to think with.
  • "Mathematical education" — Bill Thurston was an incredible mathematician and pedagogue, and I think that this is one of the better examples of what it might mean to actually look at the intellectual activity of a domain to understand what we should do about learning it...
  • "Messing about in Science"  — A classic, I think that this article is a good example of what it means to take some of the impulse behind something like Thurston's article (i.e. "What do professionals do?") and actually operationalize it in a classroom environment. I also think it has a lot to offer in considering what's so much harder about "science education" at the secondary level compared to the elementary level...
  • Turtle GeometryVisual Complex Analysis, and "Reforming the Mathematical Language of Physics" All of these are incredible textbooks/papers. I share them not just because they're incredible textbooks, but because I think that they're fascinating examples which suggest a natural question, "Why aren't there more books like this?" By "like this" I don't mean "incredible"— I mean specifically that they're revisiting basic ideas about how we represent and work with concepts in a traditional domain, and they build out a completely alternative pedagogical approach that doesn't take "content" for granted and focus on its communication through exposition.
  • The Having of Wonderful Ideas — I think this is a charming book, and a great example of what I believe more educators need to do (i.e. act as anthropologists) as we work to understand how people actually learn. Another element in this book which I find pretty special is its emphasis on the affective, not just cognitive, experience. Overall, I think this is a very basic blindspot in most reform: the returns to increasing engagement are far higher (in our experience, at least) than the returns to focusing on cognitive clarity. I think that means we need to understand what engages people and how to leverage that, which is a largely unspecified and unexplored discipline.
  • The Mathematician's Mind and The Psychology of Invention in the Mathematical Field In the same vein as Thurston's article, I think these are both great examples of what’s involved in thinking deeply about both a domain’s deep ideas and the psychology of how we develop and use those ideas in very domain-specific ways.
  • Minds of Our Own — These three clips should give you a sense of the documentary…but overall, I find this to be a fascinating look at just how fully we fail in STEM currently, even at our best institutions.
  • "Master of Fine Arts in Software" — This is a great example of what imagining alternative institutions might look like...and I don't think its similarity to places like 42 or Recurse is an accident.
  • "Teach Yourself Programming in Ten Years" and "How Stripe Teaches Employees to Code" — Both of these are more individually focused, but again, as an exercise in looking at what actually learning a thing looks like, I think they're extremely instructive starting points for imagining different contexts for folks to "learn STEM."