Simon’s Watchmakers and the Future of Courseware

Herbert Simon, a Nobel Laureate known for his work in too many areas to count, used to tell a story of two watchmakers, Tempus and Hora. In the story Tempus and Hora make watches of similar complexity, both watches become popular, but as one watch becomes popular the watchmaker expands and becomes rich, and as the other becomes popular the maker is driven slowly out of business.

What accounts for the difference? A closer look at their designs reveals the cause. The unsuccessful watchmaker (Tempus) has an assembly of a thousand parts, and for the watch to be working these must all be assembled at once; interruptions force the watchmaker to start over again from scratch. For a watch to be finished the watchmaker needs a large stretch of uninterrupted time.

The other watchmaker (Hora) has chosen a different model for her watch: she uses subassemblies. So while there are a thousand pieces, she can complete a subassembly of 10 or so pieces and put it down without it falling apart.

Simon actually gets mathematical at this point (italics mine):

Suppose the probability that an interruption will occur while a part is being added to an incomplete assembly is p. Then the probability that Tempus can complete a watch he has started without interruption is (1- p)1 000 – a very small number unless p is.001 or less.

Each interruption will cost, on the average, the time to assemble 1/p parts (the expected number assembled before interruption). On the other hand, Hora has to complete one hundred eleven subassemblies of ten parts each. The probability that he will not be interrupted while completing anyone of these is (1- p)10, and each interruption will cost only about the time required to assemble five parts.

“Now if P is about.01 – that is, there is one chance in a hundred that either watchmaker will be interrupted while adding anyone part to an assembly – then a straightforward calculation shows that it will take Tempus, on the average, about four thousand times as long to assemble a watch as Hora.

From The Sciences of the Artificial, p. 188-189

The point here is not about multitasking, but about evolution. We tend to see emergent complexity as the accumulation small deviations leading to new forms of life, thought, or practice. But that’s not enough. If we must get 1,000 changes or features in to create something worthwhile, we’re lost from the start. Changes that require these sorts of vast sequences are unlikely to occur.

Instead, evolution must proceed through a series of stable, intermediate forms, via a pattern Simon calls “aggregation of subassemblies.” You can’t get from nothing to the human eye unless there are stable and useful states between nothing and the eye as we know it, and the structure of your eye is as determined by those intermediate states as by anything else. We’ll be stuck forever with a blind spot in our field of vision because there’s no desirable set of intermediate states that would result in a new blind-spot free system.

Writing in 1961, Simon saw this issue as one of the more underappreciated elements of how change happens, any sort of change. To some extent we’ve caught up to Simon these days.

For example, programmers and designers have embraced emergent design:


Today, policy analysts talk about “path-dependence” — the reason why it’s so hard to get to a European health care system in the United States, for example, is precisely this problem of stable, intermediate states.

And just the other day I was flipping through the Andressen-Horowitz slide deck on network effects, and what did I see in there but this:


A non-network mode for a network product can be seen as another example of a stable intermediate state. In other words, single player mode is a form of sub-assembly.

I should caution people: getting the right intermediate forms is of course crucial, as is understanding the previous forms on which you are building. If we take the Henrik Kniberg picture up-page as an example, I’ve seen too many examples where Silicon Valley has offered educators a skateboard in exchange for the aging car they are driving. Emergence is more than incrementalism, and it’s one of the reasons why an deep understanding of the history of problems is necessary to solve them. Hora with the wrong sub-assemblies doesn’t do much better than Tempus.

Still, Simon’s thinking on these issues was very much ahead of his time but very much of our own time. The move from the industrial age to the information age has been a transition in part from Tempus to Hora, from structured design to aggregation by sub-assembly. We think like this now because the complexity of our current systems demands it.

OER and the Drake Equation

Of course, as Gibson once said, the future is here, it’s just not evenly distributed. We’re an information age economy that still has many bubbles of factory thought.

It’s been one of the great ironies that the information age’s biggest industrial age bubble has been in the sale of information. As an example of this, the textbook company Pearson Education recently penned an op-ed attacking the idea that open educational resources (which are produced on an emergent, open-source model) could ever compete with the factory precision of a company like Pearson.

My response to such a column would have been that this was like arguing in 1972 that we would never make it to the moon. The tipping point for OER in the classroom was year or two ago; much of what happens now is inexorable, watching the results of the past decade’s actions play out.

Still, others were more patient and less snarky than I. David Wiley wrote a fascinating rebuttal that summarized decades of careful thinking in the OER space about how open materials could compete with closed materials, and, just as importantly, what “compete with” means and how it might be judged.

In that essay (and it really does rise from the level of “post” to “essay”) Wiley attempts to explain how scale influences and informs the possibilities of open production. To do that he uses the Drake equation, an equation astronomers use to estimate the likelihood that advanced civilizations exist on other planets.

I won’t go into a detailed discussion of the equation here, but the point of Wiley’s example is this. Most of the equation seems to indicate that life is unbelievably improbable: only a fraction of stars have planets, only a fraction of those planets can support life, only a fraction of those planets would have developed life, and so on.

It seems pretty hopeless until you plug in what all these fractions of fractions are fractions of. In the Drake Equation, everything starts with the rate of star formation. And when you plug that in things change. There’s about 20 sextillion stars in the universe, which written out with zeroes looks like this:


When you start with that in your numerator you can do a lot of slicing and dicing, and you still end up with a lot of life in a lot of places. So while the likelihood of any given star producing an advanced civilization is nearly indistinguishable from nothing, the current estimate is that advanced species with the capability to transmit radio waves or other communication across space have existed in about 10 billion other instances.

Ten billion. (That startles me out of edtech thinking for a minute, the beauty of that. Think about that. 10 billion civilizations rising and falling throughout the history of the universe, and our experience here just the tiniest fraction of a fraction of this larger story. It’s one of the most beautiful things the mind can contemplate).

But back to OER. David’s point is that the chance of any one faculty member producing their own end-to-end course out of free materials that they stitch together is admittedly quite small, but when you look at the numerator (the vast amount of course creation that happens anyway each semester in a discipline) it’s not only possible that these open works will emerge — it’s highly probable, even when you plug in fairly pessimistic filters.

Hacking the Drake Equation

Once these works are pulled together initially, David argues, the filters become far less pessimistic:

Perhaps most importantly, the major efforts made by OER workhorses like Khan or Sousa catalyze additional, incremental work by others over time. As Benkler has explained, the smaller the contribution to be made, the more people there are who will have the time and inclination to contribute (c.f. Shirky’s idea of cognitive surplus).

This creates the opportunity for asynchronous, uncoordinated, incremental, continuous improvement that harkens back to Eric Raymond’s notion that “Every good work of software starts by scratching a developer’s personal itch.” Individual instructors make no improvements, small improvements, or large improvements to existing OER based on their own needs and available resources. Some subset of the group that makes changes share those changes back with the community. This kind of “snowball development” is a key characteristic of the most interesting and effective OER.

The future of the sustainable development of effective OER will be characterized by stigmergy. Stigmergy is the watchword for the next decade of OER.

I do think stigmergy, the self-organizing principle we find at work in so much of the social web, is the watchword for the next decade or OER, even if we don’t present it under that name. And it has to be that way, because as we get beyond those efforts with a large numerator (Introductory Psychology, Anatomy and Physiology, U.S. History) we have to start to optimize those denominator filters to make the whole thing work.

As an example, consider the issue I deal with on a daily basis. Many of our faculty who want to get off of paid textbooks teach advanced or niche courses. I had one faculty member who teaches State and Local Government who wanted to ditch their textbook. But State and Local Government is a course only taken a particular political science sequence, which means there’s a small number of people teaching it.  (In the terms of the Drake equation, the subject has a low “R”). The chances of an end-to-end State and Local Government textbook emerging any time soon are fairly low.

But we can fix this. The current educational materials ecosystem doesn’t exist in orbit around Alpha Centauri. It’s right here, right now, and we can tweak the environment to make conditions more favorable to intelligent life.

Still, while stigmergy is a powerful model, I’d propose that a simpler lens on how to make change is provided by Herbert Simon’s watchmakers, which provides some powerful insights into how to move forward in OER (and into how we are already moving forward).

Simply put, the textbook as produced by Pearson and others (and even as produced by many OER publishers) is Tempus’s Watch. Rather than being an aggregation of sub-assemblies, it’s an end to end treatment of a subject meant to be tightly coupled to a course sequence. Because of this, the emergence of new soup-to-nuts textbooks tends to cluster around subjects with a high numerator. Like Tempus, if you are working end-to-end you need a lot of tries (and failures) before you get a complete, usable, product.

But the future does not belong to Tempus, it belongs to Hora. Rather than large, complete works that then accrue stigmergic change, we should look forward to a shift towards evolution of sub-assemblies which are composed and recomposed into larger works.

It’s this shift away from the industrial textbook model that will radically change (and is already radically changing) what is possible with OER.

Building Hora’s Watch

I’ve spent eight years of my life looking at why the bottom-up reuse environment we envision might be possible in OER never quite emerges. And honestly, the reasons are complex, and somewhat ironically, a path-dependence problem.

In short, everything from our technical architecture to our file formats to our institutional policies assume a Tempus-like product. If you think the problem is simple or already solved, you probably haven’t done your homework.

The subassemblies of Hora’s watch must be both stable (that is, presently useful) and intermediate (that is, providing clear proximate paths for both evolution and assembly into different projects). Our architectures tend to emphasize stability (present usefulness) over the intermediate nature of such things.

As an example, we often choose formats which privilege sharing (present usefulness) at the expense of remix (intermediacy). I first encountered this when looking at OpenCourseWare (OCW) in the mid-aughts. In the quest to share things in “open” formats, providers had shared PowerPoints in PDF, a widely readable format that increased shareability at the expense of remix. In one fell swoop, this decision transformed what could have been an emergent pattern into yet another for of publication.

Once tuned into this, however, you see this clash between present stability and intermediacy all over the web. Old timers will remember that early internet video was often circulated via email or downloadable MPEG files (Remember “All your base”?). That solution favored intermediacy (the ability to edit and remix the file) at the expense of present usefulness (you might not have the proper player to play it). YouTube changed this, making playing videos trivial, and radically increasing shareability, but at the same time killing remix.

Here are other examples:

Intermediacy Stability
Usenet Webpage
Wiki Content Management Systems
MP3s Spotify

One of the things you’ll notice is that we have often moved away from intermediacy for a variety of good and less than good reasons. But that’s for another novel.

Thinking Like a Photographer

People sometimes say that I’m wrong about reuse, that they reuse images *all the time*.  To which I say: Exactly!

You see, for a variety of historical reasons images have remained intermediate (and these reasons date all the way back to HTML’s special include syntax for them). The use of images on the web is actually a pretty good guide on how to make intermediacy happen:

  • Images can take on many meanings in different contexts.
  • People take images with this in mind. E.g. not every picture I take of a building has me in it.
  • I can, with a couple clicks, get the original image directly from your server and have a copy of it up on my server (whereas if I want the raw pre-templated and processed text from your database, good luck).
  • I can transclude your image into my own work with a simple dialog box.

When we think about intermediacy, it’s helpful to think about the way people take pictures which then get included in a wide variety of contexts. My friend Alan Levine, for example, has his pictures reused all the time. And he’s a good technical photographer to be sure, but check out this. Here’s a snippet of Alan’s photostream from Flickr, a tiniest sample of thousands of photos he has.


As I said, he’s got great technical talent. But more than that, he’s got an eye for the reusable, remixable image. Each image is complete in itself, but at the same time is framed for remix and reuse in a variety of contexts. I don’t know where that recycling bin photo will end up, but my sense is it captures a network of ideas and metaphors and associations that is going to allow it to drop into a slidedeck or blog post like it was created just for that use.

Since so much of what photography does is to feed into other media, this is a big part of what we mean when we talk about having an eye for the perfect image. Alan is in these images, surely, but he’s being quiet here, and leaving space for you, the reader or remixer to use them in ways that he has not imagined.

When it comes to writing, the primary artform for remixable writing is wiki, of which Wikipedia is an an example, but certainly not the most instructive one. If you want to get a sense of what wiki can be, I usually recommend a trip to TV Tropes. Here’s an example of an article from that site, which discusses the tendency of  TV, comic book, and film characters to have battles in the middle of rainstorms:


Note that the skill here in identifying and explaining the idea is similar to Alan’s photographic skill. Here, an idea has been identified which is linky and dense with potential connections. It’s big enough to be meaningful on it’s own, but sized so that it could support other writing as well.

The links here are where the action is. The trope is an example of the Empathic Environment trope, where nature mimics the moods of characters in the story. It’s counter but related to the Battle Amongst the Flames.

Wiki pages are some of the few truly hypertext pages on the web, designed to be part of a nonlinear reading experience rather than either a linear read or the spoke and hub model of blogs. Wiki writers, like Alan behind the camera, create work to scratch their own itch, but always with an eye for leaving space for future extension and remix. A good wiki article is like a well conceived subroutine — it sees a general gap that needs filling, and fills it in a way that can solve more than the present problem.

(It’s partially because of this that I believe that anyone looking at the future of OER must start with a deep understanding of hypertext, from Vannevar Bush through Ward Cunningham; there is a deep and rich literature here that has already dealt with the problems we claim to be encountering anew. As just one example, Wikipedia is the largest digital education project in existence, and yet there are many in the OER movement whose understanding of wiki culture doesn’t go beyond a five minute summary. That’s nuts.)


I didn’t sit down to write how we solve our OER problems, though having just passed 3,000 words in this post maybe I should have. There’s a host of things we need to do, both technically and culturally to move OER into its next, emergent stage. Perhaps in some other venue I can expand on those.

For now, however, let’s just summarize the current situation:

  1. Early OER texts had a high “Drake Equation R”, and as such could emerge as nearly whole works, to then be iterated by others.
  2. As we move forward with OER, we will move more and more into “Low R” courses, where we are less likely to emerge whole.
  3. This necessitates a different approach (Hora’s watch) to OER than some of the approaches that are currently working. But neither technology or culture is currently well-aligned with the approach we need to move to.
  4. At the same time, other models, such as Flickr libraries, Pinterest, and wiki writing show that the transition can be made.
  5. Maybe we should look at these, amirite?

I really just meant to write a short post here, and surprised by the current word count. But let me know what you think in the comments.


17 thoughts on “Simon’s Watchmakers and the Future of Courseware

  1. Mike, thanks so much for this mighty thing.

    On economies of scale, and nichery: “The chances of an end-to-end State and Local Government textbook emerging any time soon are fairly low.” Outside the US, this is our familiar condition. In small educational markets like Australia the chances of a locally relevant textbook emerging any time soon are often low; and the obverse of this is that a locally developed textbook struggles to find world export markets. It’s an exact parallel to the history of our media content industries, that have been both crushed in our domestic market by the US, and tightly constrained in export terms by the same force.

    The historic consequence of this has been that we forced into a kind of resourceful openness, including adaptive reuse and local customisation (sometimes pretty clunky) of materials developed in the US and Europe, right across the disciplines. In media terms, Australia has a high cultural tolerance for torrenting, that amounts to a sense of entitlement, in a context where foreign producers’ capacity to offload discounted content, because scale, has made it more or less impossible for local producers to flourish. So the short version of this is that I think there are already good examples outside the US of the kind of thing you’re sketching, that MOOCs just made more obvious. Why make an Australian introduction to programming MOOC if you can get one off the shelf from Stanford? It’s a textbook argument presenting itself as fabulous novelty.

    In relation the design of resources that are already themselves adaptable and open, and environments that can play a hospitable role, I’m groping towards a small understanding of something about wikity. Throwntogetherness is a concept that many people associate with geographer Doreen Massey, but is also present in the work of ethnographer Kathleen Stewart, and it’s a way of seeing the hash of the forces that come together in the moment as contingent, tentative, and curious — rather than fixed, settled, or resolute. Battles in the Rain is things seen that way in a moment. What happens when it becomes A Thing?

    Wikity seems to me to be part of the irresolute web, that has a side effect rather than a goal the capacity to fix things in new ways. But what we’re seeing in education is headed the other way: the increasing fixity of things, driven (a bit accidentally) by historic learning design principles, and the rise and rise of analytics as the science of monitoring approved responses to designed interactions with fixed content.

    The question you’ve raised for me is about designing for a pedagogy of irresolution.

    Ooof, long. Shorter: thank you.

    • No, all this is wonderful. And yes, I’m aware that even 7 or 8 years ago non-U.S. markets were much more accepting of this pattern for the reasons you note.

      I’ll have to look at “throwntogetherness”, it certainly sounds like it gets at the tension between stability and intermediacy that I’ve been struggling with. As always, you teach me so much when you comment. Thank you.

  2. Please do not apologize for writing long form; let the rest of the web apologize for writing status messages or less.

    Thanks for mentioning of my photos; personally it is the most rewarding thing I do, and not because of the external uses. The link you used, however, are not my photos, but other flickr photos I have favorited- I keep my favorites in an album

    Home Skies

    This does get me thinking of the multiple ways flickr both organizes *for* me automatically (by date taken/uploaded, by camera metadata) and the ways I can organize myself- free form (by tag) category (by putting into albums), and even meta category (you can make collections of albums).

    And so I look at and agree about the lovely structure and linked-ness of TV Tropes (and for that matter the whole universe of interest based wikis, which is deeper and wider than I imagine)– not knowing for sure, those linkages are human made, as much as I do my tags and albums, someone with interest/knowledge of tropes draws the connections.

    Yet there are the wildly parts of IMDb that are built on automatic relationships, its really subtle that you do not really even see the data relationships, but I’ve fiddled with combing up with interesting genre crosses by playing with URLs. But they afford the human level arrangement too, like letting users create lists and collections of “the data” inside.

    I’m not sure what I am really saying, but there is something about these systems which make use of the relationships between information that can be done automatically, but also allow for, the kinds that only humans can do.

    Now, I only have to read this 5 more times so I can grasp the rest of it (that’s a thank you by the way for making us think, not a criticism of the writing).

    Write more / apologize less!

    • I was thinking of you both as I was searching Flickr for a background image and whose work should I stumble upon? So I spent a long time browsing Alan’s work and thinking about exactly that readiness to be open that makes the images themselves practices of hospitality. This is a delayed thank you to Alan too, because I used to be really slack about attribution. I can’t defend this except with an even slacker kind of Australianness. Now I’m weirdly hooked on attribution as a way of saying thanks. This seems to me part of the valuing of others that citation has absolutely lost, because citation has been flipped by its own marketisation/monetisation.

      Anyhow, still thinking. And certainly don’t apologise for making us all smarter. I’d never heard of most of the things in your post. My head is still spinning.

  3. Thought provoking piece, as ever Mike… It made me think of “learning objects”, but there’s a lot of negative baggage around different particular takes on that idea….

    The watch assembly model is a powerful one, particularly if the sub-assemblies themselves can be appropriated for other things. So for me, appropriation and what makes things appropriate and appropriate-able(?!) is a Big Idea.

    Related to this comes the idea that it can be a good idea to share intermediate works that you yourself intend to integrate or polish off elsewhere. The finished work may be tightly integrated into your own course, but a slightly rougher version may well be more useful to someone else. The problem there is that many academics and educators have a quality filter that means they don’t like to publish half-baked works (i.e. intermediate components/sub-assemblies), or ideas (I don’t have that problem;-)

    In practical computing, sites like Stack Overflow are exactly the sort of place where folk share intermediate steps, often phrased in the context of a problem.

    In terms of OER, I still don’t really get what they are or are supposed to be. Ideas that can be repurposed or appropriated? Or finished works that can be immediately and directly integrated into your own work? Or partial, intermediate works that need tidying up? If the answer is “all the above” then maybe that’s partly where my confusion arises, becauseI think they all play a different role in a different part of the workflow when producing new courses.

    • This is a good question. In general, I’m looking for a name that can be reassembled into a larger piece through linking or paths. I don’t hog the main namespace of a subject unless I’m ready to make a pretty good go at a general take on it.

      So a good name to me is something like Sociopathic Whuffie, which captures a take on a subject that is somewhat personal but that can be combined eventually into larger efforts:

      I tend to think that larger more general pages and paths (e.g. Whuffie) will emerge out of these smaller pages, but I don’t like to push it into that early on.

      It does hurt sister site discoverability, and I don’t quite know what to do about that.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s