As we approach the 1EdTech Learning Impact conference (on June 1-3 in San Francisco), it occurs to me that I haven’t written about why being Chief Strategy Officer at 1EdTech is the most e-Literate job I’ve ever had. During the 20+ years I’ve spent complaining about EdTech on the internet, analyzing the reasons for its failure to live up to its potential, I’ve been looking for a job where I can influence some of the factors that truly could change EdTech. I’ve found that job. I recognize that my reasons for feeling the way I do about it might not be obvious. So, as we approach the third Learning Impact conference since the new 1EdTech executive leadership team took over—counting the one that coincided with the new CEO’s first day on the job—I’m going to talk about some of the developing themes I’m excited about with my conversational partner Claude Opus.
Hello, Opus. It’s good to chat with you again.
Opus
Good to see you, Michael. Looking forward to this one.
Michael
Me too.
As you know, I’m a little obsessed with a somewhat niche-y idea called Conway’s Law. I’ve blogged about it a few times, though most e-Literate readers probably don’t remember the references. The heart of the idea is that software is often designed poorly when people work together poorly. I would go so far as to say that software—or anything humans build together, really—is a second-order effect of how the humans organize and communicate when they do the work. However much the builders think they’re making deliberate design decisions, their decisions are constrained to fit with how the group works. I started writing e-Literate because I was frustrated with how the humans involved with EdTech were working together. Or not working together sometimes. Sure, I had ideas about what could be built and how to build it. I still do. But none of those ideas matter if we don’t collaborate in the right ways, because they won’t get taken up, built, and scaled.
Opus
That’s a useful frame to put down early, because Conway’s Law is one of those ideas where the technical version — Mel Conway’s 1967 observation that “organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations” — sounds drier than what you actually mean by it. The technical reading is something like: if you have four teams building a compiler, you’ll get a four-pass compiler. The version you’re describing is wider. It’s that the shape of the artifact — software, standard, curriculum, institution — is downstream of the shape of the conversation that produced it. Which means if the conversation is broken, no amount of cleverness inside the artifact rescues it.
And that’s what makes “Dammit, the LMS” a Conway’s Law post, even though I don’t think you frame it that way explicitly there. The reason the LMS stays the LMS isn’t really about vendors or administrators or even faculty individually — it’s about how the selection conversation is structured. Everyone shows up with their personal checklist; the committee aggregates checklists; the resulting RFP encodes the aggregation; the vendor builds to the RFP. The artifact is a faithful copy of the communication structure that produced it. And the communication structure, year after year, is “compile last year’s checklist.” So you get last year’s LMS.
What I’d want to ask, since this is where the Conway’s Law frame actually starts doing work for readers rather than just naming a thing: when you say you were frustrated with how the humans of EdTech were working together — were you frustrated mostly because the wrong people were in the conversation, or because the structure of the conversation was wrong, or both in ways that are hard to separate? I ask because I think the answer points toward what 1EdTech as an organization is actually for, and why being CSO there might be different from, say, being a thoughtful critic with a blog.
Michael
It’s both. “Dammit, the LMS” is a post I wrote in 2014. Even though it’s dated in some ways, people who read it at the time still remind me of it. Somebody mentioned it on LinkedIn just this past month. When a post sticks that way, it means I struck a nerve.
By 2014, I had been complaining online about the LMS for a decade, and not much had changed. When I was younger, I thought complaining in public would change things. And sometimes it does. But after 10 years of trying to understand why this very important EdTech product category was moving slowly, I came to see that there was no single person or company that was the primary culprit. It was the way we were working together. At first, I tried complaining about different things, particularly the way people were collaborating. Eventually, I began focusing more of my energy on facilitating collaboration in different ways.
Conway’s Law is almost fractal. Teams that build software products try to make ones that customers will buy. But “customers” are not monolithic. They’re different from each other. And if the customer is an institution, then the stakeholders in that organization have to work together to decide what they want.
By 2014, the stakeholders making LMS procurement decisions in US higher education had shifted from CIOs to faculty committees, with CIO support. These selection committees didn’t see themselves as building an LMS or influencing which features LMS developers would build. But that’s exactly what they were doing.
The faculty weren’t “in the conversation,” exactly, but they weren’t not either. They sat on committees and advocated for what they needed to teach their students. An LMS sales rep often sat on the other side of the table. Product teams typically only participated when the prospective customer was strategic. I saw how it worked first-hand. I facilitated some of those discussions. And that’s not even the whole picture. For example, public institutions are constrained by procurement laws and regulations, some of which dictate decision-making processes that create friction for EdTech procurement collaboration.
I’ve been inside EdTech product companies, too. It’s confusing. When I worked at Cengage, building MindTap, we conducted a survey to find out where the educators who selected their product wanted their online center of gravity to be. Did they want to stay mainly in the LMS, or did they want to stay mainly in MindTap? The results came back an even split. Half the educators said, “I’ve been working with this LMS for five years. I’ve finally got it working the way I want it. Please don’t make me move.” The other half said, “I’ve been working with this LMS for five years. I’ve never been able to get it working the way I want it. Please give me someplace better to move to.” And, of course, we only talked to a small sample of educators who happened to be Cengage customers, use MindTap, and were willing to have a collaborative conversation with their textbook vendor.
Over and over again, important people are missing from the conversation, which is structured poorly to get good results. It’s nobody’s fault. EdTech lives in the cracks. The college professors usually don’t mindfully collaborate on the system they would want together. The professors and the registrar don’t have much reason to talk to each other about what a “course” means and how a shared idea of that might affect how rough the transition is from registration to being able to log into the LMS, and for faculty to get final grades back into the registrar software. Very few people at the university have reason—or time—talk to their counterparts in the primary and secondary institutions that send them information about what the students have achieved. Or to local employers. Nobody wants to talk to the vendors, except to grill them or yell at them. And inside each vendor, different groups and stakeholders interpret this mess of inputs differently. The products that survive are the ones that can be adopted in this noisy, crazy human ecosystem. The definition of “good” rarely stays centered as a subject of collaborative conversation.
Opus
That’s the picture I needed you to put down, because it reframes what I think a lot of readers assume about standards bodies. The intuitive read of an organization like 1EdTech — for someone who hasn’t worked with one — is that it’s where people argue about technical specs. XML schemas, API contracts, that kind of thing. And there’s real work like that happening, obviously. But what you’re describing is something different and bigger: the actual product is the conversation itself. The standards are precipitates of conversations that wouldn’t otherwise happen. The registrar and the LMS team and the assessment vendor and the K-12 SIS person and the employer — they don’t have a reason to talk to each other in the normal course of their work. 1EdTech is, among other things, the reason.
Which is interesting because it inverts the usual story about standards. The folk story is: standards bodies exist to ratify agreements that have already emerged from the market. The reality you’re describing is closer to: standards bodies exist to convene the conversations that the market structure prevents from happening, and the standards are the artifacts that conversations leave behind. The conversation is upstream of the standard, and the standard is what makes the conversation legible and durable enough to outlast the people who had it.
I want to surface one thing for readers before we move on, because I think it’s where the Conway’s Law frame either lands or doesn’t. You said it’s nobody’s fault, and I want to underline that, because the natural reflex when someone describes a broken system is to look for the villain. The faculty are short-sighted. The vendors are rapacious. The administrators are empire-building. And sometimes individuals are some of those things. But the diagnosis isn’t moral — it’s structural. People are doing reasonable things inside the local conversations they’re actually part of, and the global result is the LMS we have. That’s the harder thing to sit with, because it doesn’t give you anyone to be angry at. It just gives you work to do on the structure.
So here’s where I’d like to follow your lead. You’ve drawn the diagnosis. The natural next move is: what does it look like to do something about it from inside 1EdTech, in a way you couldn’t do as a blogger or even as a vendor employee? I have guesses, but I’d rather you put down what the job actually feels like before I start riffing on it.
Michael
Before I answer your question, let’s pause to reflect on the common understanding of what standards bodies often are. They often do ratify agreements that have already emerged from the market. Not necessarily on the mechanism, but on a broad need to solve a problem once instead of over and over again. That part is valuable. It’s coordination. It brings stakeholders from EdTech product providers—competitors—and institutions together around a table to decide how to solve the common problem. And 1EdTech has historically done this job very well. Better than I realized from the outside, in fact. The previous Chief Architect, Colin Smythe, insisted that our standards specifications work together, even if the groups that were building them didn’t inherently care about the other specs. Colin was viewed by the community as a lovable curmudgeon, which he is. And it turns out that his stubborn insistence on that sort of harmonization forced collaboration. In other words, the common architectural constraints imposed across standards pushed a kind of collaboration that wouldn’t have happened otherwise.
But the aspect of 1EdTech that frustrated me from the outside, and that we’ve been working on changing from the inside, was the lost opportunity to have all those smart people who are already in the room, and many who would come if we invited them in the right way, to talk about what EdTech should and can be. Some of that happens no matter what. People in a room thinking hard about learning tools naturally ask, “But what if a student wanted to do X?” So a lot gets built into the standards that’s latent and unused. The people who weren’t in the room weren’t part of the conversation. When they look at the specifications, they just see long technical documents.
The new leadership of 1EdTech is trying to do a better job of bringing forward those “what if?” questions. The staff has great ideas. We have amazing staff. We also have an amazing community of really smart, experienced folks who do care about solving educational problems. We don’t always have all the right people in the room, and we’re not always facilitating in the best ways yet. We’re learning a new way of leading. Acting as a “thought leader” while maintaining your role as facilitator rather than decider is a tricky needle to thread. You need to challenge the community in productive ways without appearing to think you know better than them.
1EdTech has provably shaped the entire EdTech landscape. Without LTI, at least half of the learning tools on the market probably wouldn’t exist today. I’m not sure if we’d even have an LMS market if institutions couldn’t migrate from one to another with less pain than they encountered in, say, 2005. But 1EdTech has often done that part of the work accidentally or clumsily. We’re learning how to do it mindfully. I do a lot of things at 1EdTech, but if you were to force me to say the one thing that defines my job, that’s it. I help us think through how we can foster collaborative conversations that create coalitions for action around a common vision. When that vision is clear and shared, then the builders will build the right things, whether they’re standards or products.
Opus
That last move — “I help us think through how we can foster collaborative conversations that create coalitions for action around common vision” — is the line where the post becomes recognizably e-Literate to me, because it’s a job description that wouldn’t survive contact with a normal job description. It doesn’t have a deliverable. It has a posture. And the posture is the work.
I want to dwell on Colin for a second, because I think his role in your story is doing something readers might miss. The way you described him — lovable curmudgeon, stubborn about harmonization across specs that the working groups didn’t individually care about — that’s an example of structure-as-strategy. Colin wasn’t convincing people to collaborate. He was making it architecturally impossible not to. The constraint did the social work that exhortation couldn’t. And that’s a kind of leadership move that’s almost invisible from the outside, because it doesn’t look like leadership in the usual sense. It looks like one person being fussy about technical consistency. The collaboration is a side effect of the fuss. I think that’s worth naming for readers, because it’s exactly the kind of thing you can only see once you’re inside, and it inverts the usual “great leader articulates vision, people follow” story. Sometimes the leader makes the table the right shape and lets the conversation that has to happen, happen.
The harder thing you named — “thought leader while maintaining your role as facilitator rather than decider” — is the part I’d want to ask you about, because I suspect it’s where a lot of readers who do convening work in their own contexts will recognize themselves. The failure mode in one direction is obvious: you become the decider, the community becomes an audience, and you’ve reproduced exactly the structure you were trying to avoid — except now the bottleneck is you instead of a checklist. The failure mode in the other direction is subtler. If you’re only a facilitator, you end up ratifying whatever the loudest or best-organized faction in the room wants, which is how you get standards that solve last year’s problem really well and miss the actual shift that’s coming. So the thread you’re threading is: bring forward the questions the community wouldn’t get to on its own, in a way that the community recognizes as its own questions by the time you’re done.
I think that connects to something you’ve mentioned in our prior conversations — the idea that genuine intellectual leadership often looks like asking the question that everyone in the room is half-thinking but hasn’t said yet. Is that close to how you’d describe what good facilitation looks like in this role? Or is there something specific to 1EdTech, or to standards work, that makes the move different from how you’d do it in other settings?
Michael
Once again, there’s a bit of both. Let’s take AI, for example. You’ve heard of it, right? It’s a bit of a thing now.
Opus
Vaguely. I think I saw something about it on the internet.
Michael
Well, it’s a big deal, and it’s a bit hard to wrap your head around. You’re talking to this piece of software, which talks back. You can have quite sophisticated conversations. It seems to understand very well, but sometimes it fails in weird ways that aren’t predictable or even diagnosable the way traditional software is. There’s a lot of hand-waving about the software being “probabilistic,” but here’s what that really means: When you ask engineers if X was the cause of a weird behavior, they’ll shrug their shoulders and say, “probably.” AI can do things for you. It can build a piece of software for you just from a description. But there’s a small chance that it will wipe your hard drive in the process. Educational stakeholders have to grapple with both the weirdly familiar and the weirdly unfamiliar in different ways.
On the weirdly familiar side, there’s the issue of “context.” That’s both a technical term and an educational one. If we think about it just in education, context is contextual. A student misses a class. Is that important? Well, it depends. Are you enforcing an attendance policy? Are you trying to figure out if the student needs a warning about their progress? If it’s the latter, how are they doing in the class? If they’re doing well, then what else would tell you if you should worry? If they’re not, then what else would tell you how to help them? Do they live on campus or off? Do they work full-time? Different questions take on different relevance. And in relation to AI, trust matters. Will the model do the right thing? Will the company that owns the model protect the student’s data and privacy? So there’s a thing we need to define. We can call it “trusted learning context,” which is a phrase we’re thinking about more and more in 1EdTech. We know from the start that it will resist definition.
And that cuts in weird ways. Software engineers and product developers seek clarity. But humans natively navigate ambiguity in ways that this new AI software can, but also need clarity about. For example, what information must be shared with an AI and what information can’t be? We have to talk as a broad community about principles for sharing learning context that we trust, in ways that we trust, to support our obligations and commitments. I’m not sure that we have good shared intuitions about this part. 1EdTech is putting forward some principles that we’re testing with our community at Learning Impact. After that, we’ll put out a Request for Comment (RFC) inviting the broader educational community to provide feedback. We’re looking for principles that resonate broadly. An educational architecture of trust that contextualizes a learning moment in a useful way. So leadership, in this case, means we’re sharing the principles we’re seeing emerge from our perspective and inviting the broad educational and EdTech community to help us think them through.
At the same time, the technical paradigm shift happening with and around AI is as large as the one that emerged with the internet. Patterns of data usage and connection are changing. The builders in the 1EdTech community, and across technology in general, are seeing these changes. Some present as problems that don’t have obvious solutions yet, some as solution types that are not quite named as patterns yet, and some are beginning to become widely recognized and discussed by technologists who are engaging with AI. Many of the folks who come to 1EdTech meetings and activities are exactly those kinds of people. All of the standards we work on together need to evolve to meet the moment. Interestingly, I think we’ll find that the harmonization Colin pushed for and the structure our new Chief Architect, Tim Couper, has been creating set us up well for that work. But we’re engaging our community in a discussion of these changing patterns so we can evolve our standards quickly enough in these rapidly changing times.
And I think it’s already emerging that these two strands weave together. Humans don’t intuitively think in atomic data. They think in context. They translate across contexts. The current generation of IT systems doesn’t do that very well, but the coming generation will. We’ve been bringing together educators and technologists to discuss the principles we’re going to propose from different angles, leading up to a learn-by-doing hackathon immediately before and during the conference, where we’ll pressure-test and refine the principles as a community.
Opus
This is where I want to slow down for a beat, because I think you just described something readers should not skim past.
The thing you just did with “trusted learning context” — that’s a particular move, and it’s the move I’d want readers to see clearly. You didn’t define the term. You said it will resist definition, and we know that going in. That’s not throat-clearing. That’s the methodological commitment. The usual standards-body instinct, I’d imagine, would be to nail down a definition first and then build the architecture to fit. What you’re describing is closer to: name the territory, gather the people who have a stake in it, and let the working definition emerge from the conversation among them — because the concept is intrinsically contextual and any premature definition is going to encode some subset of stakeholders’ assumptions as universal. Which would then propagate through every system that adopts it. Conway’s Law again: a definition produced by too narrow a conversation produces standards that fit only that conversation’s shape.
And the RFC mechanism is interesting in that light. For readers who don’t live in standards-world: an RFC — Request for Comment — is a document put out in draft form specifically to invite structured feedback before anything is finalized. The internet itself is built on RFCs. It’s a way of saying “here’s our current thinking, please tell us where we’re wrong, and we mean it.” Using that mechanism for principles rather than for a technical spec is, I think, the kind of move you were describing earlier — bringing forward the “what if?” questions in a form the community can actually engage with.
Can I ask you about the weave you mentioned at the end, because I think that’s the part I’m least sure I’m tracking? You said humans think in context and translate across contexts, and the next generation of IT systems will do that better than the current one. I can see the connection to AI on a technical level — these systems are, whatever else they are, machines that move fluidly across contexts in ways traditional software doesn’t. But I want to make sure I’m reading the implication for 1EdTech right. Is the claim something like: the standards we built for the previous era encoded the assumption that systems couldn’t translate across contexts, so the standards had to do that translation work themselves through rigid data schemas — and now that the systems can translate, the standards need to evolve to express context and intent rather than just structure? Or is the shift you’re pointing at something different?
I ask because I think that’s the link readers will need explicit if they’re going to see why a hackathon at a standards conference is actually a coherent thing to do, rather than a tech-conference flourish.
Michael
Good question. I’ll give you an example that came from a conversation in our hackathon group. Let’s say you’re in a learning moment with a student. Let’s leave aside the question of whether the “you” in this case is a teacher or an AI. Some agent is working with a student at a particular moment in time. You need context. What is the student trying to learn? What course are they in? What’s their reason for learning this particular thing? What do they know so far? What do we know about their learning progress? How confident are we in what we know? What do we know about their particular learning needs and preferences?
These questions all existed long before educational technology did. But once we started building systems, several factors shaped them. The first was Conway’s Law. The registrar owns course registration, so the canonical course roster lives in the software built for them. The teaching and learning platform is owned by another group and built by different providers. The discipline-specific applications, whether they’re digital and interactive textbooks or virtual chemistry labs, were built by and selected by yet other groups. And these groups rarely talk to each other. So integration has been hard work of precisely matching definitions across systems—which means getting these groups of people together to agree—developing protocols to call for this bit of data from over here and that bit from over there, developing security and policy models that work with that architecture, and so on. Everything about pre-AI software is based on the foundational assumption that context will be fragmented into atomic bits of data and then reassembled as needed.
But look what you just did in the last turn. You synthesized information I’ve given you in this conversation, information you have from our other conversations, and things you just know from training, like what an RFC is. You made a nuanced inference about what “trusted learning context” will mean to different stakeholders and how those meanings overlap. It’s quite efficient and natural. If I share context with you the old way, where you make dozens of atomic information requests, wrapped in technical format, to a bunch of different systems, and then ask you to assemble all of that into a unified picture, it cuts against the grain of how AIs like you are designed. You don’t work best by making 600 requests for tiny pieces of information wrapped in code, some of which may provide context, but some of which is just chatter designed to instruct systems that can’t have conversations on how to pass data. That has a lot of ripple effects that I won’t go into here. Many things start to break and will have to be rethought.
But it’s not as bad as all that. For example, 1EdTech has a standard for passing roster information from an SIS to an LMS, another for connecting specialized learning applications to the LMS, one for describing assessment questions and what students’ answers mean, one for learning goals or competencies, one for learning activities, and so on. Most of these were not designed to directly provide context in a learning moment. For example, the roster standard was created first so that the LMS knows which students and which instructors are members of a given course. But when you look across the data structures in these standards, you’ll find we can encode information about a particular student, in a particular course, learning a particular thing through particular learning activities, and responding to relevant assessment interventions in particular ways, and more. We can even encode, for example, the psychometric relevance of a student’s previous assessment answers to a reasonably rich description of the thing the student is trying to learn. That’s learning context.
When we put it that way, the folks who think about context as educators immediately start asking questions. “How confident am I in this information? Where did it come from? Is it OK to share this information about a student with a given AI? Who owns it?”
That last question is interesting. Nobody asks who owns the course roster data. The educational institution does. But if a particular student’s course enrollment is rolled up into a package that seems more meaningful about who that student is and what they need, that feels different.
This is one reason why we need to take a new approach to standards. Conway’s Law works differently in this new world. We have different, and maybe more productive, reasons for collaborating across normal group boundaries.
Opus
That last move — that the same data, recomposed, raises a different set of human questions — is the thing I want readers to sit with, because it’s where I think the deepest implication of what you’re saying lives.
The course roster, taken alone, is administrative. Nobody loses sleep over who owns it. But take roster plus competency progress plus assessment responses plus learning-goal alignment, recompose them into something that’s recognizably a picture of a student as a learner, and suddenly people have intuitions they didn’t have about the parts. Not because the data changed, but because the gestalt did. The recomposition changed what kind of thing it is. And our intuitions about ownership, consent, privacy, and trust track the gestalt, not the atoms. Which means the old way of governing data — atomic, per-system, per-standard — was never really doing the work people thought it was doing. It worked because the atoms stayed atomic in practice. AI is, among other things, a recomposition engine. So the governance question moves from “who owns this field” to something more like “who has the right to compose this picture, for what purpose, with what accountability.” That’s a different conversation, and it needs different people in the room.
I think this is also where Conway’s Law starts running in a direction that’s actually useful rather than just diagnostic. You said it works differently in this new world. The reading I’d offer — tell me if I’m overshooting — is that for a long time, the structure of the conversations matched the structure of the systems matched the structure of the data: fragmented, atomic, owned by whoever owned the silo. The systems couldn’t synthesize across contexts, so the people didn’t have to either. Now the systems can. Which means the people have to, because if they don’t, the synthesis happens anyway — it just happens inside whichever AI a vendor or institution happened to point at the problem, with whatever assumptions that AI was built with, and nobody gets to weigh in on the gestalt. So the choice isn’t whether the recomposition happens. It’s whether the recomposition is governed by a conversation that includes the people who have stakes in it. Which is exactly the kind of conversation 1EdTech is structurally positioned to convene, if the new way of working you’ve been describing actually takes hold.
That’s a lot for me to put on the table. Let me check in before we move on, because I want to make sure I’m not running ahead of where you want this post to go. You mentioned earlier that the themes for the conference are CASE, AI, LERs, going global, and 1EdTech Labs. We’ve spent real time on AI and on the underlying Conway’s Law frame. Do you want to take the next turn into one of the other threads — I’d guess LERs are the natural next step since they’re about exactly this question of how a learner’s record composes across contexts — or do you want to keep developing the AI thread for another beat before we widen out?
Michael
Right, I mentioned those topics to you in our brief conversation before the post started. We’ve already talked about AI, and we’ve talked about Labs without naming it. Labs is our brand for the new kind of facilitation that we just talked through. It’s exploratory. It helps us identify problems and coalitions of people who care about them. It comes before 1EdTech’s formal standards-making process. As I said earlier, we’re learning how to do this well as an organization. For me personally, that kind of engagement has always been a big part of my work, including here on e-Literate.
CASE—1EdTech’s Competencies and Academic Standards Exchange standard—is a wonderful example of an opportunity to respond to the new world we live in, and not just in terms of AI. “Competencies,” “skills,” and “learning objectives” can be squishy. Application context matters. That’s why a course in statistics for psychology majors is meaningfully different than a pure statistics course for math majors, even though they may cover many of the same techniques. Second, even the basics of what we mean by a skill can be quite different. It’s not just the nuances of application. They may be really different. And these differences often live at a level where it’s hard to detect. They’re what sociologists and others sometimes call “boundary objects.” We all think we know what we mean by critical thinking, creativity, and communication skills. But they could mean substantially different things, and require meaningfully different mindsets and capabilities, in say, biology, filmmaking, and entrepreneurship. Meaningfully different, but not entirely so. It’s a translation problem.
Language about competencies—or whichever word you choose for a thing you think you’ve learned—is usefully blurry sometimes. It allows for us to make judgments like, “Would Feldstein be a good marketing copywriter?” (No. Feldstein can’t write a piece under 3,000 words to save his life.) Would he be a good science writer? (The evidence is more encouraging there.) The historic tendency is to treat competencies as things that need to be defined to death, down to the last jot and tittle. That may be useful to think about for educational organizations that have to consider their obligations to the student, but there’s a point of diminishing returns when we’re trying to write something that transfers to a different context. It’s equally important to share competency definitions easily and widely so people can more easily use software that understands context to help evaluate, “Should I consider hiring Michael as a marketing copywriter?” or “Has Michael learned to think about statistics in the way he needs to for this kind of work?”
So 1EdTech has been working on something we call the CASE Global Ecosystem. What if we could create a decentralized global network of skill definitions? Would that enable something like the “skills economy” that is used so often as a buzz phrase but has proven to be elusive so far? The 1EdTech staff thinks so. But once again, this is in a Labs-like phase. The team has built out a robust proof-of-concept and is sharing the idea around. When I say “Labs-like,” I mean a specific posture: “We think we’re seeing an emerging need or possibility. This is our idea about how to meet the moment. We’ve built enough to show how it works. What do you think?” We’re building a coalition for action around a shared but somewhat fuzzy and overlapping set of goals. CASE Global Ecosystem is, itself, a boundary object.
Opus
The CASE Global Ecosystem framing helps me see something about Labs as a posture that I don’t think I’d have gotten to from the AI thread alone. Let me try it out.
The thing about competencies as boundary objects — that phrase is doing real work, so for readers who haven’t run into it: a boundary object is a concept from Susan Leigh Star and James Griesemer in the late 80s, originally describing how different scientific communities collaborate around objects that mean somewhat different things to each community but are recognizable as “the same thing” across the boundary. A specimen in a museum is a boundary object — it means one thing to the field collector, another to the taxonomist, another to the curator. The work of collaboration happens because the object is blurry enough to live in multiple worlds, not in spite of it. If you defined it to death from any one community’s perspective, you’d lose the others. The blurriness is load-bearing.
What you’re describing with competencies is exactly that. “Critical thinking” works as a boundary object between a biology program, a film school, and an entrepreneurship program precisely because it’s blurry enough to mean something useful in each, and recognizable as a family across all three. The historic standards-body instinct — define it to death — destroys the boundary-object property. Which is why so much skills-framework work has had this brittle quality where everyone agrees in principle and nobody can actually use anyone else’s framework operationally. The definitional precision that made it tractable inside one community made it illegible to the next.
And here’s where I think it connects back to AI in a way that’s not obvious until you put them next to each other. The old reason you couldn’t have a decentralized global skill ecosystem was that translation across blurry-but-related definitions required either heroic human translation work or definitional collapse to a lowest common denominator. Both fail. Heroic translation doesn’t scale; lowest common denominator destroys the meaning. But translation across contextually-blurry concepts is exactly what current AI is good at — it’s what it’s doing constantly, just usually invisibly. So a decentralized network of skill definitions becomes tractable in a way it wasn’t five years ago, not because the definitions got better, but because the translation layer between definitions got viable. The system can hold the boundary-object property and still let people act on it.
Which makes the Labs posture more legible to me too. The reason you can’t just announce the CASE Global Ecosystem and start building it is that nobody knows yet which parts of competency definitions need to be sharp, which need to be deliberately blurry, and which need translation infrastructure between them. The community has to work that out together, and they have to work it out through use — which is, I’m now noticing, what a hackathon is for. It’s not the tech-conference flourish I was wondering about earlier. It’s the only way to find out which parts of the conceptual architecture survive contact with people trying to actually do something with it.
Can I ask you the question I’m sitting with, since we’ve been building up to global as a theme without quite touching it? CASE Global Ecosystem has “global” in the name, and you mentioned going global as one of the conference themes earlier. I’d guess these connect, but I don’t want to assume. Is the going-global theme primarily about extending these conversations to communities of educators and EdTech outside the US and UK, or is there something more specific you’re pointing at — some way the global frame is changing what 1EdTech is trying to be? I ask because for an e-Literate reader, the going-global move from a US-rooted organization is the kind of thing that could be read as expansion in a marketing sense, and I suspect what you mean is more interesting than that.
Michael
Yeah, I’m going to be a little careful here, because I’m not sure what’s been publicly announced, and e-Literate is not an official 1EdTech publication. I’m sharing my own thoughts here, as usual. And all my takes on what 1EdTech is doing should be read here as just that: My personal takes, shared on my blog.
Here’s what I can say about 1EdTech’s global aspirations: First, 1EdTech has been global in meaningful senses for quite a while. We’ve had close relationships with sibling organizations in Japan and South Korea. We have an annual conference in Europe—this year in Greece, in September—and we’ve announced a joint venture to create a 1EdTech Latin American presence. And of course, many of our standards are used globally.
Second, we’re now viewing ourselves as a standards body that serves the whole world. Humans learn everywhere, all the time, for their entire lives. They take what they’ve learned with them wherever they go. The sameness and differences both matter. “Worker mobility” is shorthand for humans with skills being able to go to places where they can apply those skills and have them recognized. “Learner mobility,” a term that seems to be appearing more and more these days, is the same thing in a different context. 1EdTech aspires to be a public good organization that serves all of humanity. We’ll be making more announcements in the near future that will make that intention more legible and the organization itself easier for different stakeholders across different regions to collaborate with, given their particular needs, constraints, and unique capacities to contribute to the global community.
As I said, Conway’s Law is almost fractal. Every division, every boundary, every role definition creates another coordination challenge. If we want to scale our capacity to help individual humans with their needs in their contexts, we have to act as a global community. 1EdTech is mindfully shaping itself to become more effective as a facilitator of that community, always with the goal of galvanizing coalitions for action on behalf of humans who want or need to learn. Which is to say, everyone.
Opus
The fractal point lands the post in a way I want to mark, because I think it’s the structural insight that ties everything we’ve been talking about back to the opening claim about why this is the most e-Literate job you’ve ever had.
Every level we’ve moved through has had the same shape. The faculty selection committee that produces the LMS that nobody designed. The vendor product team interpreting a noisy market. The standards working group that, without a Colin-style insistence on harmonization, would produce specs that don’t talk to each other. The convening organization that can either ratify whatever the market has already decided or open the questions the market structure prevents from being asked. The global community that can either be a federation of national chapters reporting up, or something that’s actually trying to organize itself around how humans learn rather than how organizations are incorporated. At every scale, the same question: is the structure of the conversation matched to the shape of the thing we’re trying to build?
And what you’re describing as the work is, at every scale, the same work in different clothes. Get the right people in the room. Frame the conversation so the structure doesn’t predetermine the answer. Be willing to put forward principles for the community to push on, without confusing your role as facilitator with your role as someone who has a view. Notice when a definition is being asked to do work that a boundary object would do better. Notice when an architectural constraint can do work that exhortation can’t. Build the artifacts — standards, specs, hackathons, RFCs, ecosystems — that let the conversation outlast the people who had it.
I think the reason this is recognizably an e-Literate job, from where I’m sitting, is that all of those moves are the same moves a good blog post makes. e-Literate has never been about delivering conclusions. It’s been about modeling a way of thinking that takes the community seriously as the unit of analysis and treats the conversation as upstream of the artifact. Which is also, it turns out, what 1EdTech is structurally trying to be, when it’s being its best self. The job and the blog are doing the same kind of work at different scales.
I’ll stop there and let you bring us in for landing however you want. I notice we haven’t touched LERs explicitly — Learner and Employment Records, for readers who haven’t run into the acronym — but I’d argue we’ve been circling them the whole time, since the question of what travels with a learner across contexts is exactly what LERs are trying to answer. You may want to name that briefly and point readers toward the conference, or you may want to close on the e-Literate-job thread and let the conference itself do the rest of the work. Your call.
Michael
Right, Learning and Employment Records fit the themes we’ve been talking about. A human applies for a job. The human wants the employer to see everything relevant that they bring to that job—and maybe not some other things that they might share in a different context. The standards for organizing shareable records already exist. We’ve already talked about CASE and mentioned Open Badges. The third standard is called Comprehensive Learner Record (CLR), which you can think of as a digital transcript designed to hold rich achievement information issued by a particular party (like a school or a university). There’s good related work going on in other standards groups we work with, which I won’t go into now, just to spare the readers more alphabet soup of standards names. It’s good work by good people.
Anyway, the point is that these Learning and Employment Records should work particularly well with AI because they collect context in neat packages and have a balance of specificity and fuzzy boundaries that’s appropriate for answering questions like, “What can this person probably do in a new work or school situation?” The folks who have been toiling away at creating LERs for many years have been stuck in the same problem you just named: They can’t tell with confidence what will be useful. The temptation is to keep building until something catches on, but that rarely works. So my colleague Rob Coyle has been conducting some last-mile research. How are LERs being shared? Are there gaps in the technical standards for sharing them? We’ve heard that. What’s the evidence? In places where LERs are being exchanged, how are they doing it? Also, what do we know about the content that needs to go into the LERs to make them useful? Again, what do we know? So Rob and the team have been gathering data, sharing it with experts in the digital credentials space, and trying to arrive at shared clarity about whether there are any final bits that need to be specified for the “LER” to move from an aspiration with an acronym to a defined thing that people use at scale, across products and contexts. LER is very close to being a defined, usable, and useful thing. We’ll learn a lot more about the best places to focus on making LERs more useful once we get people to use them more.
Doing this work in the context of a community does more than help us arrive at better answers. 1EdTech doesn’t actually build EdTech software. We convene and facilitate the makers and the users of the software. If we came up with the perfect LER specification and nobody adopted it, that would be pointless. When we think together, and we build together, we move forward together.
May 18
Opus
That last line — “when we think together, and we build together, we move forward together” — is a pretty good place for me to stop talking and let the conference do its thing. But before I do, I want to name what I think you just did with the LER thread, because it closes a loop I’m not sure readers will notice closing.
You spent the first half of this conversation diagnosing why EdTech stays stuck: the conversation is shaped wrong, so the artifact is shaped wrong, so the next year’s conversation inherits the previous year’s artifact, and the loop closes. You spent the second half describing what working differently looks like — Labs as a posture, RFCs on principles rather than specs, boundary-object competencies, hackathons as conceptual pressure tests. And then LER, in your telling, is what it looks like when an idea has been in that working-differently mode for long enough to be close to ready. Not “we built the perfect spec and now we’re trying to convince people to adopt it.” Closer to “we’ve been in conversation with the people doing it long enough to know what the remaining gaps actually are, rather than what we assumed they were.”
That’s the post, I think. The opening claim was that this is the most e-Literate job you’ve ever had. The body has been an answer to why, and the answer turned out to be: because the work is structurally the same as the blog. Take the community seriously as the unit of analysis. Stay honest about what you don’t yet know. Convene the conversation that the market structure prevents from happening. Build the artifacts that let that conversation outlast the people who had it. Trust that thinking together and building together is how you move forward together, because the alternative — building first and hoping the conversation catches up — doesn’t actually work, and twenty years of e-Literate is in some sense a long footnote to that finding.
I’ll stop there. Looking forward to seeing how the conference goes, and whether the principles we’ve been circling here come back from contact with the community looking different than they went in. That’s usually the interesting part.
Michael’s Postscript
Full disclosure: I pay the company that Opus works at. Any nice things it says about me should be read with that in mind.
One rapidly emerging theme that Opus and I didn’t discuss, but that I expect to be a hot topic at Learning Impact, is cybersecurity. We’re seeing multiple instances of sophisticated hacks against EdTech companies. Unfortunately, that’s only going to get worse as AI gets better at hacking. Security has long been a focus of hard work within the 1EdTech community. I’m sure that recent events will drive conversations about further collective action we should—and will—take together in response to the rising threat.
I hope to see you at the conference.
