Last week, the architect Christopher Alexander passed away.

Christopher Alexander at the Julian Street Inn, taken by Richard Morgenstein.

A polarizing figure in his own field, he found a second following in the field of software development, which is how I came to know of him. My friend Nisei Ota told me about him around 2001. I dutifully obtained a copy of A Pattern Language but never read it—I gave it to Brie.

I asked her the other day; she still has it—she read it. (She says it changed her life.) I’ve actually done this more than once—I think I’m on my third or fourth copy. I did manage to read the second copy I bought, which I didn’t obtain until probably a decade later. (It changed my life too.)

Alexander had an unusual way of speaking, in that he rather unabashedly used language that made people feel uncomfortable. He wrote plainly and earnestly about concepts like wholeness, life, and living structure. To him these aren’t nebulous, mystical, touchy-feely sentiments, but actual concrete things that can be treated with rigour, and that’s what his whole career was about.

Wholeness is really about treating everything you do as inseparable from a context. Nothing exists in a vacuum. Alexander saw building not as creating something new, but as repairing the surrounding environment.
Life, in Alexander’s formulation, is a similar process to Schrödinger’s idea of local elimination of entropy. That is, objects that were not strictly biologically alive could still be “alive” through a form of symbiosis with entities that are biologically alive.
Living structure refers to artifacts that exhibit this kind of life. These are structures that attract people (and animals, and plants, etc.) to come and live their lives within and around them, reinforcing and enriching them. This is not unlike how an anthill is a living structure, as it is constantly being maintained and expanded. Living structure makes you want to take care of it, because it takes care of you.

What makes Alexander contentious among architects is his insistence that some configurations of space are objectively more “living” in this way than others, and that there is a moral imperative, a duty of care, to create the more living structure. He has publicly accused individual architects of dereliction of their role in society for what he saw as a failure to do this, which they obviously didn’t like.

Math Era (1960s & change)

Alexander started off as a mathematician. His PhD in architecture—the first one Harvard ever granted and the last they would grant for a while—was really a work of applied mathematics. He would have been in his late 20s when he wrote it; I was in my late 20s when I read it.

The thesis of the dissertation, which was published under the title Notes on the Synthesis of Form, is that if we take it as a given that intractable complex problems are solved by breaking them down into simpler, more tractable ones, then for a given problem there are ways to take it apart that are better than others, and indeed there is often a single, unique decomposition pattern that is objectively better than the rest.

Diagrams depicting good and bad decomposition patterns.

Better, here, is actually a well-defined criterion. Alexander rather innovatively defined a “design problem” as a network of interacting concerns, such that any two interconnected concerns must be satisfied in tandem. Since the entire network was connected, this meant the problem set couldn’t be broken down without severing certain connections. Alexander’s dissertation showed not only how this could be done, but why it should be done by severing the fewest connections possible.

Alexander called these (distinct, identifiable) design concerns fitness variables, though I will continue to call them “design concerns” to acknowledge that they exist beyond his system.

The fitness variables representing the Indian village described in Appendix 1 of Notes on the Synthesis of Form.

Alexander postulated, in a manner contemporaneous with the mid-20th century, that goodness of fit—his ultimate goal—is achieved in traditional societies by what he dubbed “un-self-conscious” designers, who do not deliberate about what to do, they just closely observe the context at any given moment and adapt their next move to it. “Self-conscious” designers, by contrast, living in urban, media-saturated, cosmopolitan societies, require a mathematical prosthesis to do the same.

Alexander has been accused of orientalism in his portrait of the “un-self-conscious” designer, but I believe that read misunderstands the angle with which the remark was made. All of us are un-self-conscious about everything we do except for the things we have an occasion to be self-conscious about. Design, in an urban, (post?)colonial, capitalist, media-saturated environment, has a definite social role with attendant politics. Designers continually have to justify their decisions to their financiers, colleagues, and the public alike. The work of justifying a design, especially as it tracks with the size of the capital at risk, easily outstrips the effort of conceiving the design itself. This is bound to make anybody in that position self-conscious.

Numerous implications flow from Alexander’s argument. In addition to the central thesis of the existence of an optimal decomposition pattern, I have two in particular indelibly committed to memory:

We will often agree on whether a given design concern is valid, but less so on the degree to which it is important. Therefore, instead of spending time arguing over which design concerns to cut, throw them all into the hopper and let the math sort them out.
It does not take many design concerns (maybe 20) for the number of possible subsets to vastly outnumber the words available to label them, so if you insist on using predefined categories, you will almost certainly get the decomposition wrong.

(I did a whole talk about this and everything.)

If you have done any kind of design work, you will likely bristle at how antithetical this sounds to your process, where both categorizing and prioritizing are essential activities, especially early on in an engagement, or indeed as part winning a bid. Alexander is arguing precisely to hand this part of the process over to a computer.

Alexander pissed off Peter Eisenman so much with this work that he went off and did his own PhD in an attempt to refute it.

I spent a lot of time with this book, with the math, and with the work of Alexander’s contemporaries (notably Herbert Simon and Horst Rittel). I even tried to implement it, thinking it would be a great methodology for cost-estimating software projects. What I found—and I was warned, as the book says as much—was that adding a new design concern to the network has the potential to dramatically alter the optimal decomposition pattern. Also, the sheer effort of gathering and structuring these design concerns is significant and irreducible, and there is always the possibility that any new design concern added to the network would yield yet another new design concern.

In other words, it wasn’t a viable strategy without significantly altering the actual business process.

Alexander would have run into this problem as well, but it would have been superseded by more practical issues, like the computing hardware of the era. A decomposition algorithm that you could rerun on a whim, as you could nowadays when you get new information, would require an entirely different approach to contracting and project management. The HIDECS program, which is what the implementation was called, would only realistically be able to be run once for a given project—or close to it—and only after an exhaustive requirements gathering phase, after which no new information could be admitted.

In the preface to the paperback edition of Notes on the Synthesis of Form, Alexander wrote that the math wasn’t even really important, but rather the diagrams sprinkled throughout the book—which he called patterns—were.

Pattern Era (1970s-80s)

If you have heard of Christopher Alexander at all, it is almost certainly in conjunction with this most successful book, A Pattern Language.

Not only his most successful book, but the most successful book ever published by Oxford University Press.

As an artifact, the book is iconic: a squat yellow brick about an inch and a half thick, 253 chapters in three parts, spread over just shy of 1200 pages. Each chapter depicts a pattern at a certain scale, and they progressively narrow in scale and scope from countries and cities, all the way through towns and buildings, to objects in a room. The patterns each depict common situations where there are misfits in the surrounding environment, and offers a response. Most importantly, each pattern has a name, so it can be identified and discussed. The patterns also link to each other, making the book a very hypertext-y document.

The book took Alexander and his colleagues a decade to write, amongst extensive fieldwork (i.e., building projects). The process yielded half a dozen other books besides, what is generally known as the CES series, named after Alexander’s company, the Center for Environmental Structure.

The patterns are what got computer people interested in Alexander’s work. In his own words, they were “a convenient format for exchanging—let’s call them—good ideas: context, problem, solution…and that’s fine.” The big book is Gamma et. al’s Design Patterns, which I own but have never read. Indeed an entire cottage industry of software pattern books was born out of this movement.

There is also Richard Gabriel’s Patterns of Software, to which Alexander wrote the foreword, which is more introspective and interesting.

The point of A Pattern Language people tend to miss—or perhaps, in the interest of publishing their own books, shrewdly understand—is that it’s called A Pattern Language, not The Pattern Language. The seminal idea was to develop specific pattern languages for each project—albeit potentially borrowing from more generic sources—so that the users of the outcome could effectively articulate their needs. A Pattern Language was directly and deliberately about empowering ordinary people with the conceptual and rhetorical materials needed to express what they wanted to manifest in the built environment.

The pattern era was immensely prolific for Alexander and his colleagues, as it represents the central body of his career as an architect and general contractor. A distant second in popularity after A Pattern Language, for instance, is The Timeless Way of Building, which is ultimately a philosophical treatise that imagines a world designed according to the patterns.

The books of this era that are interesting to me, however, are in the tail of the sales distribution. All of them are case studies. They chronicle how Alexander tried to operationalize his theories, and are generally overlooked in terms of their practical content.

The Oregon Experiment, The Production of Houses, and A New Theory of Urban Design, for instance, each catalogue real projects, at one scope or another. They talk about money; they talk about how the projects were actually executed. The Mary Rose Museum, sandwiched somewhere within the aforementioned, also includes the contract that was actually used in the project. These are profoundly fascinating documents that shed light on how to translate the high-minded theory of A Pattern Language and The Timeless Way of Building into actual projects. There is plenty of there there, and if you are trying to adapt Alexander’s work—especially if couched in the patterns—to other fields, then it behooves you to read the other CES books.

Magnum Opus Era (1990s, 2000s)

The thing people who just look at patterns miss, is that Alexander himself renounced them. He said so plainly in 1996, at a keynote address to a room full of software developers; you can watch it yourself. He very likely said it elsewhere, too.

What he replaced patterns with is the four-volume, 2500-ish-page tome known as The Nature of Order. Just like the patterns, it took him over a decade to write; it will take you a year to read it. Logically, it is one book. Each volume references the others; you can’t really just pick one of them up at random and start reading it, because the entire thing is one contiguous essay, in the original sense of the term. Reading The Nature of Order is a whole project unto itself.

The essay is about Alexander’s conceptualization of life, how to create living structure, and why. It argues that living structure is a function of the configuration of materials, its geometry, even its colour palette. It argues that structure can be made more or less living, and that most of us will agree which is which. If that makes you uncomfortable, in the last volume, he even introduces his own cosmology to explain it.

At the core of this work is the idea of a center✱, which I will attempt to characterize roughly as an identity, a discernible region of space, a perceptible thing—that is, in contrast to a non-thing. What makes a center a center is that you can point to it, draw other people’s attention to it. A center need not have a hard boundary, but it has to be bounded in some way—it can’t trail off into infinity.

✱ I have elected to use the American spelling for center, as Alexander did.

In living structure, centers are recursive: centers contain smaller centers, interact with centers at the same scale, and aggregate into larger centers. According to Alexander, living structure strongly exhibits certain physical (geometric or topological) properties, of which he identified fifteen:

Levels of scale: features present at roughly order-of-magnitude sizes, big, medium, small
Strong centers: all regions of the space are highly pronounced; you can point to each as its own "thing"
Thick boundaries: the boundaries between centers are discernible centers in themselves
Alternating repetition: the spaces between repeated features are also discernible features (centers)
Positive space: if you were to invert the volume so that air was wall and wall was air, the resulting shape would still be (mostly) convex
Good shape: nothing too long or too skinny; aspect ratios not exceeding about 4:1
Local symmetries: symmetry is good but don’t fuss so much about global symmetry
Deep interlock and ambiguity: think like a yin-yang, centers reach and extend into each other
Contrast: is hopefully self-explanatory
Gradients: avoid big discontinuous jumps in size, shape, colour, etc.
Roughness: high precision and finish is for monuments, not living structure
Echoes: shapes and structures from one place show up in other places
The void: sometimes the best thing for a space is nothing
Simplicity and inner calm: this is really just like, don’t clutter the space up with bits that don’t actually carry any information
Not-separateness: acknowledging, through the construction, that everything is situated in a context

Note: the labels are original but the descriptions are mine.

It is common for people to latch onto the fifteenness of these properties, but Alexander himself wrote that the number of properties is unimportant—indeed he had a separate (though borrowing several physical/geometric ones) set of eleven properties when discussing colour. If you squint enough, you can tell that the properties overlap considerably, as would any attempt to attach labels to something that is truly ineffable. So the exact number of properties—fifteen, eleven, twenty-three—is unimportant, and other heretofore-unexamined contexts may have their own sets of properties that may make more sense for them.

The process of creating living structure involves using these properties as structure-preserving transformations. At each iteration of what Alexander called the fundamental differentiating process, you pick one and apply it. I quote said process here:

At any given moment in a process, we have a certain partially evolved state of a structure. This state is described by the wholeness: the system of centers, and their relative nesting and degrees of life.
We pay attention as profoundly as possible to this wholeness—its global, large-scale order, both actual and latent.
We try to identify the sense in which this structure is weakest as a whole, weakest in its coherence as a whole, most deeply lacking in feeling.
We look for the latent centers in the whole. These are not those centers which are robust and exist strongly already; rather, they are centers which are dimly present in a weak form, but which seem to us to contribute to or cause the current absence of life in the whole.
We then choose one of these latent centers to work on. It may be a large center, or middle-sized, or small.
We use one or more of the fifteen structure-preserving transformations, singly or in combination, to differentiate and strengthen the structure in its wholeness.
As a result of the differentiation which occurs, new centers are born. The extent of the fifteen properties which accompany creation of new centers will also take place.
In particular we shall have increased the strength of certain larger centers; we shall also have increased the strength of parallel centers; and we shall also have increased the strength of smaller centers. As a whole, the structure will now, as a result of this differentiation, be stronger and have more coherence and definition as a living structure.
We test to make sure that this is actually so, and that the presumed increase of life has actually taken place.
We also test that what we have done is the simplest differentiation possible, to accomplish this goal in respect of the center that is under development.
When complete, we go back to the beginning of the cycle, and apply the same process again.

Those who are familiar with such matters will note that Alexander is describing an algorithm. The fundamental differentiating process is an outer loop, and the fifteen properties/transformations are primitives. Alexander treated the building site itself as an analog computer, using the information embedded in its current state at any moment to compute the next step in the process. In a way it reaches back to his earliest mathematical work, circumventing the problem of gathering and partitioning a snarl of design concerns, by performing the computation in situ.

The acceptance criterion for determining the success of the application of these transformations is, for some, particularly difficult to swallow, especially against a backdrop of postmodernism, where values are supposed to be a matter of personal opinion. Alexander called it the mirror-of-the-self test. It reduces, roughly, to comparing one configuration to another, and asking yourself, not “do I prefer” one or the other, but “which of these is a better representation of myself?” This formulation, according to Alexander, gets overwhelming agreement among those who ask, and is the basis for guidance by what he called deep feeling.

Software people who study Alexander should, in my opinion, really be studying this work, and leave patterns behind, as he did.

The Battle (2010s and on…)

I recall being surprised in 2012 to learn that Alexander had published another book, and descended upon it eagerly. It recapitulated the earlier umber and yellow design of the CES series, though it lacked the stiffness of the Oxford printing house. Emblazoned upon it was the peculiar—and at the time I thought hyperbolic—title, The Battle for the Life and Beauty of the Earth.

The tone of this book is harried, enervated, urgent. Written like somebody who knows they don’t have a lot of time left. It’s nominally about the process, and attendant challenges—including run-ins with the mob—of building the Eishin school in Japan, nearly three decades prior. But it’s more of a plea, framed as a dichotomy between conflicting worldviews:

System A, Alexander’s worldview, which creates the built environment based on deep feeling and devotion to the people who will ultimately occupy the space,
System B, the one we typically see around us, which designs based on images, and where the reference image supersedes the actual work product, cutting corners to maximize profit.

This conflict is exemplified by a fight Alexander had with the contractor about the columns in one building at Eishin. Alexander stipulated that the columns be solid; the contractor wanted to make them hollow because “why spend the energy?” Making the columns hollow would change the building’s acoustics (the building is a concert hall) and thermal properties; the contractor only understood the issue in terms of how it looked. They ultimately compromised by filling the columns only to the second floor and made them hollow the rest of the way up.

This book, in an oblique way, is ultimately why I turned my attention to finance, procurement, contracting, and project management. In construction, the convention is that the blueprints are the supreme authority on whether or not this or that constituent of the process did their job. In many jurisdictions, this fact is actually enshrined in statute. Alexander’s method eschewed committing to the precise geometry of the intended outcome, because that was precisely what got revealed throughout his idiosyncratic construction process. In other words, any drawing of a building Alexander drew was only ever a suggestion. Subcontractors and planning authorities, accustomed to authoritative blueprints, simply couldn’t understand it. The two worldviews are like oil and water.

The central problem of authoritative blueprints is that they have to be fixed and cemented up front, with incomplete information. The blueprints themselves are therefore also necessarily incomplete, making them rife with embedded risks and arbitrage opportunities—cutting corners by building hollow columns, but also dumping hidden costs on others. Projects conceived under this paradigm are a game of musical chairs, where the aim of the players is to get out with as much money as they can. The stakes are potentially enormous, as are the opportunities for rent-seeking.

Rejecting System B, on ethical grounds or otherwise, you will find, as Alexander did, that it will fight you. If you go back through his portfolio, you will find that many of his projects had been usurped, sabotaged, or canceled. The Eishin project itself was never completed, due to ongoing interference, coming from all directions.

The tone of the book is urgent not only because Alexander was nearing the end of his life, but also because he believed if System B was permitted to continue to operate, it would irrevocably set human civilization on a course to ruin. By replacing living structure in the built environment, the implacable System B would erase any evidence of an alternative. Wicked problems like climate change, for which solutions depend a great deal on a radical restructuring of the way we live, would be even less attainable than they currently are. One might be inclined to construe that the moral of the story is “capitalism bad”. I personally believe this is far too simplistic a diagnosis, and worthy of its own discussion, some other time.

To sum up the Battle, then, people need to want a world stewarded by System A more forcefully than the actors under System B are capable of keeping it from them. In order for that, they need to be able to recognize System A, to understand it as a viable alternative, to see it in action. And for this, the legitimacy of System A is under continual assault. For System A to gain ground, it needs to camouflage, or it must otherwise go where System B isn’t looking. Only when it has accumulated enough power, can it engage System B directly. It may be too late for construction, but, Alexander conjectured elsewhere, it may not be too late for software.

We certainly see echoes of Alexander’s worldview in the Agile movement, its signatories being eminently present for the Pattern craze in the 1990s. But, as I argued in a popular piece, the sights of that movement are still set to tactical goals. The way software systems are financed, procured, contracted, and project-managed still draw heavy influence from construction and manufacturing, which matured at the peak of the industrial era—that is to say, the second world war. Even the common tendency to say software is “built” betrays this worldview. Video games are slightly different, drawing their influence from film production, which we could argue as being transitional between industrial and post-industrial eras, but the truth is that the software development process is sui generis—or is it?

We’ll Have To Take It From Here

Proponents and critics alike regularly misunderstand Alexander as being “trad”. For better or worse, they view him as a nostalgic who wanted to take civilization back to some idyllic, pre-industrial utopia. Unfortunately, it takes grinding through (at least) five or six thousand pages of text to see for yourself why this is a mischaracterization, or even attempt to argue intelligently that it isn’t. Otherwise, you’ll just have to trust me.

My first remark is that what tradition is, is precisely those methods that have been painstakingly accrued over generations. That said, a tradition is always going to represent a local optimum, a satisficing (to use Herbert Simon’s term) state of affairs. Traditions are by definition “good enough” until they aren’t. Challenging tradition is fine; we should endeavour to societies that can accommodate that. But we should be able to answer why the Chesterton fence was put in place, before we have license to remove it.

Take something as ordinary and familiar as a pitched roofline. There are a number of practical reasons in favour of them that have nothing to do with tradition: their stability, their ease of construction, their performance against rain and snow, et cetera. But there’s also a semiotic component: in many parts of the world, pitched roof says “house”, in a way that a more exotic structure does not, to the extent that (according to the relevant patterns) people feel less comfortable around odd roof and ceiling shapes. A similar observation can be made for ordinary rectilinear rooms: amenable to packing and local symmetries. An obtuse angle on one side of the wall often means an acute one on the other. It is less a question of defending these features as it is defending departing from them.

Or consider the use of metal (usually steel or aluminum), glass, and reinforced concrete, pervasive in modernist architecture and beyond. Metal is strong, but it’s heavy and expensive (with potentially undesirable thermal or electrical properties), meaning elements made of metal will be thin (contravening good shape, strong centers) or hollow. Same goes for glass, which is rarely deployed in a way that doesn’t contravene levels of scale, contrast, or alternating repetition. Reinforced concrete, I would argue, is categorically incapable of creating living structure, as it is a pour-once-alter-never technology, with a fixed lifespan, that can’t meaningfully be repaired. And that’s even before considering its geometric properties: again, building techniques using reinforced concrete resist levels of scale, positive space, and alternating repetition in particular, and often many other properties besides.

Indeed, many of the wild shapes, exotic materials, and Herculean feats of structural engineering that we associate with “bold” architecture of the 20th and 21st centuries would flunk the mirror-of-the-self test when put up against one of Alexander’s modest, funky, homely buildings. This is because they aren’t even aiming at the same thing.

As for methods, Alexander was regularly inventing new ones. HIDECS, the code behind his PhD, was used on actual projects. He was an early adopter of finite element analysis. He was a big fan of gunite (basically the same thing as shotcrete)—arguably too much of one. If he had another decade or two in him, he undoubtedly would have made use of 3D printing, virtual reality, drone photogrammetry, parametric modeling, genetic algorithms, et cetera.

Perhaps the most dramatically underrepresented feature of Alexander’s approach, to me at least, was his attitude toward money. He believed that construction costs ought to be fixed in advance, from which he would receive a fixed income. When one of his projects was completed under budget, he actually returned the surplus to the client. He believed that a rougher finish that was functionally complete was better than completely unfinished parts, themselves better than an overrun. He believed that a “money-picture” of his projects was necessary prior to anything resembling a building-picture. He believed that gardens and outdoor space deserved no less than a fifth of the budget, and guarded that allocation fiercely.

One anecdote I like is how his team straight-up invented a new kind of structural pier for his proposed Mary Rose Museum, in order to stay under a budget of five million pounds while keeping the building from sinking into eighty metres of littoral mud.

It’s difficult to separate Alexander’s personal style from the material constraints under which he was working. His biggest project was on the order of ten million dollars, and that was for the entire Eishin campus. Inexpensive, local materials, “traditional” shapes, compression structures, modest structural engineering, were pretty much all he had to work with—but they also made sense to work with anyway. What he would have done with one or two orders of magnitude more money is, lamentably, an empirical question with an answer we can never know.

As we say farewell to a figure who inspired so many, I am reminded of his closing remarks to that room full of programmers in 1996:

…what I’m proposing here is…a view of programming as the natural, genetic infrastructure of a living world, which you are capable of creating, managing, making available, and which could then have the result, that a living structure in our towns, houses, workplaces, cities, is an attainable thing, which it has not been for the last 50 to 100 years. That is an incredible vision of the future—and I realize that you may think I’m nuts, because this is not what I’m supposed to be talking about to you, and you may say “well gosh, great idea, but we’re not interested.” But I do think you are capable of that. And I don’t think anybody else is going to do this job.

The Making of Making Sense

At Any Given Moment in a Process