What they don’t teach you at STEM school

What they don’t teach you at Harvard Business School (book cover)

What they do teach you at STEM school is how to think and act within rational systems. What they mostly don’t teach you is how to evaluate, choose, combine, modify, discover, or create systems. Those skills are actually more important for social, cultural, and personal progress. Learning them is also rarer, and more difficult—currently!

This post sketches a hypothetical curriculum for developing these meta-systematic capabilities. It’s preliminary; perhaps even premature. There is no existing presentation of this subject that I know of, which makes it more difficult than it should be. My understanding of the topic draws on a dozen academic disciplines, each written in its own unnecessarily obscure code. Both my understanding, and the pedagogical structure I’m proposing, are tentative and incomplete.

Partly this presentation hopes to inspire some readers to pursue meta-systematicity; partly it is a plan for a large project that I hope to pursue myself; partly I hope you will give feedback, make suggestions, or contribute ideas to the project too!

Goal and audience

The overall goal is to take you from systematic rationality to meta-rationality as quickly and painlessly as possible. The curriculum should re-present insights I’ve found in many semi-relevant fields, as clearly and simply as possible, in STEM-friendly terms, in a structured, sequential format.

Learning meta-systematic skills shouldn’t be so hard, and meta-systematic understanding is particularly valuable in STEM. It is inherently somewhat conceptually difficult; but probably not as difficult as, say, senior-year undergraduate physics. However, it does have cognitive prerequisites.

This curriculum is for people who have mastered systematic rationality, specifically in a STEM framework. For the most part, you have to have a thorough understanding of how to work within systems before it’s feasible to step up and out of them, to manipulate them from above. There are other routes to mastering systematic rationality—through experience as a manager in a bureaucratic organization, for instance—but this curriculum will assume a STEM background.

The minimum requirement might be an undergraduate STEM degree; but research experience at the graduate level may be needed. You have to have seen how many different systems work, and—more importantly—how they fail. At the undergraduate level, you are mainly shielded from the failures, and systems get presented as though they were Absolute Truth. Or, at least, they are taught as though Absolute Truth lurks somewhere in the vicinity, obscured only by complex details. Recognizing that there is no Absolute Truth anywhere is a small downpayment on the price of entry to meta-systematicity.

That may already have set off warning bells. Woomeisters and postmodernists say things like that—and if you think they are horribly wrong, I agree!

This curriculum is about how to do STEM better. It is not about taking you out of a STEM worldview into some alternative. Everything here is on top of that view. It addresses limitations in the way STEM is typically taught and practiced, but does not contradict any of its content. There is no woo involved—including no STEM-flavored woo, such as neurobabble or quantum or Gödel woo.

In fact, a critical step is letting go of some of STEM’s own woo—quasi-religious beliefs about the ability of rationality to deliver certainty, understanding, and control. For that letting-go, the meta-systematic mode demands that one develop an additional cognitive style. Routine STEM is easy for those who are precise and rigid of mind, and so find promises of certainty, understanding, and control particularly comforting. Meta-systematicity requires openness, flexibility, daring, and uncommonly realistic common sense—as well as technical precision.

I’ll begin with some preliminary definitions, and provide a brief overview of the curriculum. Then most of the page goes through the syllabus, organized into ten modules, in more detail. That is still just a summary, which may be difficult to make sense of on its own. I’ve included in it links to resources that provide more explanation; some of my own web pages, and articles and books by others. At this stage in the project, even these leave many holes, which I hope to fill gradually. Many of the books are seriously difficult reading; the hypothetical curriculum would extract and explain clearly their relevant points.

Some loose definitions

By system, I mean, roughly, a collection of related concepts and rules that can be printed in a book of less than 10kg and followed consciously. A rational system is one that is “good” in some way. There are many different conceptions of what makes a system rational. Logical consistency is one; decision-theoretic criteria can form another. The details don’t matter here, because we are going to take rationality for granted.

Meta-systematic cognition is reasoning about, and acting on, systems from outside them, without using a system to do so. (Reasoning about systems using another system is systematic, and meta, but not “meta-systematic” in this sense.1) Meta-rationality, then, is “good” meta-systematic cognition. Mostly I use the terms interchangeably.

One field I draw on is the empirical psychology of adult development, as investigated by Robert Kegan particularly. This framework describes systematic rationality as stage 4 in the developmental path. Stage 5 is meta-systematic. However, as far as I know, no one from this discipline has applied the stage theory to STEM competence specifically. Empirical study of cognitive development in graduate-level STEM students would be helpful,2 but in the absence of that I’m working from a combination of first principles, bits of theory taken from many apparently-unrelated disciplines, anecdata, and personal experience.

According to this framework, there is also a stage 4.5, in which you lose the quasi-religious belief in systems, but haven’t yet developed the meta-systematic understanding that can replace blind faith. Stage 4.5 leaves you vulnerable to nihilism, including ontological despair (nothing seems true), epistemological anxiety (nothing seems knowable), and existential depression (nothing seems meaningful). It’s common to get stuck at 4.5, which is awful.

The arc of the path

Overall, the curriculum leads from 4 to 5, while aiming to avoid the nihilism of 4.5; or at least to minimize its trauma, by leading you forward from 4.5 to 5.

The term “4.5” prompts the thought that the path could be structured as ten substages; 4.1, 4.2, and so on, past 4.5 to 4.6 and eventually on to 5.0. This is a severe “abuse of notation”; the empirical data do not support it. However, it may appeal to STEM folks’ appreciation of crisp structures. So I am provisionally adopting it, taking the spurious definiteness as humorous.

In fact, the key to meta-systematicity is accepting that perfect definiteness is never available. Or, in other words, nebulosity is pervasive. Meta-systematicity is non-systematic by definition, so it cannot have as cut-and-dried a curriculum as undergraduate physics. The path is necessarily somewhat nebulous. The ten steps are artificial; in reality cognitive development is never altogether linear.

The obstacles to developing meta-systematic skill are emotional as much as cognitive. Everyone must navigate two emotional crises.

When you have watched rational systems fail enough times, you are ready to move beyond stage 4. However, you may also start to sense the nihilism that lies ahead. You recoil in horror from the possibility that all systems may fail conclusively. Then you may cling even more tightly to the safety of the known, and try harder to persuade yourself that the eternalistic lies and rationalist myths are true. This may make it difficult to take even the first steps beyond stage 4.

It should be helpful to make explicit from the beginning that falling into nihilism at stage 4.5 is a possibility, but that it is avoidable if you are suitably equipped. Also, that beyond 4.5 is stage 5, which is more functional than stage 4 (whereas 4.5 can render you practically catatonic if you don’t know how to deal with it).

The second potential emotional crisis comes at 4.5, when you fully understand that systems can’t function in their own terms, but don’t yet have a clear understanding of why they do work. Three supports may help:

So:

In 4.3 and 4.4, we challenge the systematic worldview’s understanding of how systems themselves work. We demonstrate that the certainty, understanding, and control promised by 4 is false; but also begin to show that a different kind of knowledge is possible through meta-systematicity, and meta-systematic skills give you more confidence, understanding, and influence than are genuinely possible at 4.

By the end of 4.4, you need to have abandoned the hope that systems can ever be made to “work” in the way stage 4 assumes. You have to really, truly, permanently give up on that to go further. But then the 4.5 module explains in detail why nihilism is wrong. Systems obviously do work—just not in the way claimed. If you assimilate both the 4.4 and 4.5 material, you recognize that nebulosity and pattern are inseparable, and so there must be some alternative to both eternalism and nihilism. And, from 4.1 and 4.2, you have some vague sense of what that must be.

Module 4.6 introduces the complete stance that acknowledges both nebulosity and pattern. The subsequent modules take that as given, and develop cognitive skills that work with their interplay. 4.7 explains how systems actually work (namely, by non-systematic situated meaning-making) at the nuts-and-bolts level. 4.8 and 4.9 develop skillful meta-systematic cognitive patterns, at increasing levels of complexity and breadth.

5.0 could point out that, according to developmental psychologists, all aspects of the person typically progress more-or-less in sync. This curriculum concentrates only on cognition, because that’s what a STEM audience most wants to hear about. Having made the 4-to-5 shift cognitively, it should be easier to understand and appreciate the parallel changes that are possible in emotional life, relationships, culture, and society.

The syllabus in more detail

Jedi master marmot teaches mad skills

4.1 Meta-rationality is a thing and you already do it

Modules 4.1 and 4.2 aim to inspire you to step beyond stage 4, with promises of mad cognitive skills of types you haven’t learned before, and which are useful in STEM practice. To the extent possible at this stage, they should deliver on the promise, actually teaching meta-systematic skills that you will find valuable even if you go no further.

4.1 is an introduction. It gives simple examples of meta-systematic cognition, with exercises that leave you confident that:

A first lesson in meta-rationality” and “Judging whether a system applies” fit nicely in this 4.1 category.

4.2 Developing meta-rational skills

Here we go through all the various meta-systematic operations—evaluating, choosing, combining, modifying, discovering, and creating systems—in “sandboxes” that give simple illustrations for each. The Bongard problems of the “first lesson” are a sandbox for system discovery, for example.

Then we’ll look at some more serious, real-world examples, to give confidence that meta-systematicity is valuable for more than solving artificial puzzles.

How To Think Real Good” includes some 4.2-level material. I wrote it before I was thinking in the sequential framework I’m suggesting here, so it’s pretty scattershot, with bits that might fit into several modules.

4.1 and 4.2 leave intact the stage 4 understanding of what systems are and how they work. At this point, the aim is to “create space” around systems: to challenge the implicit assumption that operating within them is the whole story, and to show how acting on systems from the territory outside is also possible.

4.3 Nebulosity and the limits of systems

Here we begin to break up the systematic worldview’s fundamental assumptions. There’s several quite different ways of going at this, and 4.3 and 4.4 should lead the student through several; one may lead to understanding where others don’t.

Here’s one, in abstract summary. Any system describes the world in terms of a vocabulary of entities, categories, properties, and relationships. According to the systematic worldview, the system works because terms of the vocabulary correspond to entities, categories, properties, and relationships that actually exist in the world, and which work the way the system says they do.

However, the human-scale world doesn’t have any entities, categories, properties, or relationships. Not objectively, anyway! The physical world is nebulous—cloud-like—without any definite boundaries. There are no objectively-separable entities, because everything is somewhat mushy around the edges. Even after you divide the world up into entities by fiat, they never quite fit into categories. Your taxonomy always has some vagueness, making for marginal cases that can’t be classified meaningfully. Similarly, your imputed entities never definitely have the properties you enumerated. (Where does “red” end and “orange” begin? How much speckling can an apple have before you no longer want to call it “red”?) And so also with relationships.

Brian Smith’s On the Origin of Objects discusses this in detail.

“No amount of evidence can fix a wrong theory” is the title of what will probably be my next metablog post. It illustrates the failure of rational systems with two case studies, epicycles and nutrition, and fits neatly in module 4.3. It also introduces the basic ideas of chaotic dynamics, which are another common reason for rational systems not “working” in their own terms.

4.4 Systems can never “work”

This module continues the theme of 4.3, using numerous examples and forms of rational reasoning to undermine the cognitive illusions and emotional appeal of systematic eternalism: its promises that complete certainty, understanding, and control are possible, at least in principle.

Here we need to address objections that systems can be grounded in fundamental physics or mathematics, so they can (ultimately, at least) be made reliable. These are straightforward logical errors, which can be rectified with straightforward rational arguments. (A section of the “first lesson” dispelled one such objection rooted in the Church-Turing Thesis.)

Despite the module’s title, it’s not that rational systems don’t work; much of the time, they obviously do, and are indispensable. It’s that systems don’t work for the reasons ideological rationality claims they do. They do work for quite different ones (which we’ll learn in 4.7-4.9). This matters because rationality’s failure modes are not the ones rationalists expect.

Rationality expects failures due to known unknowns: parameter uncertainty, incomplete information of determinate types, and insufficient computational power, for instance. These sorts of failures can be planned for, and mitigated by adjustments within the system.

Systems don’t expect, and can’t cope with, unknown unknowns.3 For example: relevant common-sense observations can’t be made to fit into the model because its vocabulary doesn’t make the necessary distinctions; a sensible rule is misinterpreted in a specific case; significant aspects of the circumstances are unexpectedly not accounted for by the model at all, so it’s not even wrong, but entirely inapplicable; the system’s recommended course of action is infeasible, ignored, or obstructed, and the next-best option is outside its scope.

These sorts of failures are the raw materials that the meta-systematic skills of 4.7-4.9 work with! But for a stage 4 sensibility, they either have to be denied and ignored, or else they result in total breakdown.

Rittel and Webber’s “Dilemmas in a General Theory of Planning” discusses “wicked problems,” which cannot be solved systematically. In fact, “wicked” problems can’t be solved at all. But they are important, and can be addressed intelligently by other means. Most problems that involve more than a few people are “wicked” in this sense—which may explain why STEM-educated people tend not to like those sorts of problems!

Hubert Dreyfus’s What Computers Still Can’t Do: A Critique of Artificial Reason is a rather dated discussion of artificial intelligence. However, it is actually a first-principles, rational argument against the sufficiency of systematic rationality per se, and applies to people as much as to computers.4 I believe the argument is basically correct. Dreyfus’s conceptual framework is based in Heidegger’s Being and Time, which I’ll mention again later.

James C. Scott’s Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed describes ways that rational systems cause catastrophes when they collide with nebulosity. It’s recommended highly by many smart people, but I haven’t managed to read it yet.

4.5 Cloud-treading over the nihilist abyss

The eternalistic myth of systems is that they provide definite ground: solid justifications for understanding, and guarantees of effectiveness for action. By implication, without a correct system, knowledge is impossible and action is futile.

Module 4.4 showed that there is no solid place to stand anywhere. Knowledge, understanding, and action can only ever be vague, ambiguous, and fluid. That is: nebulous; cloud-like.

This may—perhaps should!—induce severe vertigo. The aim is to bring the student to a total emotional, as well as intellectual, disillusionment with rationalist eternalism. You realize you are over the abyss of meaninglessness, with nothing but wisps of cloud between you and the bottomless darkness beneath. This can produce panic, rage, and depression: symptoms of the nihilism of stage 4.5.

I believe those can be prevented; but some degree of emotional upset may be unavoidable. Moving conclusively beyond systematicity inevitably induces feelings of loss—loss of your previous way of making sense of the world, and of your previous, systematic self.

Curriculum module 4.5 addresses the danger of stage 4.5 nihilism.

Since there never was any ground, you always were walking on clouds—and that worked pretty well! Your eternalistic belief in systems was mistaken, but your activity was relatively effective nonetheless.

In other words: because we do understand and act effectively, therefore we can. The remaining work, from 4.6 to 5.0, is learning more about how that can be, and how to do it better.

Ideally, pointing this out is sufficient. Nihilism is just obviously wrong, and refuted by every moment of everyday experience. However, there are dozens of supposedly rational arguments in favor of nihilism, which may suddenly seem compelling when you reach stage 4.5—particularly if you come from a STEM background. Each of these arguments is straightforwardly mistaken, on straightforward factual, rational grounds, but there are so many of them that they can oppress you into despair.

Oddly, no one in any intellectual tradition seems to have written a clear and accurate explanation of why nihilism is wrong. Most explicit opposition comes from eternalism, and boils down to “God exists, so nihilism is wrong” and/or “nihilism implies murder is OK, so even considering it is taboo.” More sophisticated writers take it for granted that nihilism is obviously wrong, and so don’t bother to refute it.

I think a detailed refutation of nihilism will be valuable anyway, for two reasons. Falling into nihilism is a genuine danger (especially for STEM folks); and fear that nihilism might be right is one of the main reasons people stick to eternalism even when it is also obviously wrong. So this is near the front of my writing To Do list.

Nietzsche began working on a refutation of nihilism near the end of his working life. His Twilight of the Idols is a preface to that project. I consider it perhaps the high point of Western philosophy. The single-page chapter “How the “True World” finally became a fable” is an intense summary of his summary—and also of the whole Western philosophical tradition and what is wrong with it.5 As for the overcoming of nihilism, he writes there:

Bright day; breakfast; return of bon sens and cheerfulness; Plato’s embarrassed blush; pandemonium of all free spirits.

Robert Kegan’s work can be read as a guide to the transitions from eternalism, to nihilism, to the complete stance. He wrote primarily about the emotional and relational aspects, although he does describe the epistemological ones in passing. My work here could be taken as filling in the details he omitted.

4.6 The dance of nebulosity and pattern

Systems, we saw in 4.3, don’t deal well with nebulosity, so they try to avoid it, ignore it, pretend it doesn’t exist, hope it goes away, or destroy it. Going beyond systematicity requires acknowledging nebulosity. Nebulosity by itself would be unworkable, but it always intertwines with patterns, and effective meta-systematicity works with their inseparable union.

My understanding here draws heavily on Dzogchen, a branch of Buddhism. The most relevant work I know of is Ju Mipham’s Beacon of Certainty, which unfortunately is extremely difficult. The content is only moderately difficult, but the presentation assumes familiarity with the technical vocabulary of academic Buddhist philosophy and its two thousand years of arcane controversies. In a sense, Meaningness could be read as an attempt to make the Beacon of Certainty accessible.

Taoism, which I know much less well, also developed a sophisticated understanding of nebulosity and pattern. The Zhuangzi is the root text. It is easy to read (unlike Mipham), but I found it difficult to understand. The mathematician Raymond Smullyan developed an interpretation that may appeal particularly to STEM folks.

From Taoism, I’ve found most valuable The Great Image Has No Form, or On the Nonobject through Painting. It is notionally about Chinese landscape painting, but actually about nebulosity and pattern. I wrote about it here.

In the Western tradition, ideas about “spontaneous order” and “emergent behavior” are relevant. The most sophisticated treatments are in economics and in evolutionary theory. These two are closely linked, with each influencing the other throughout their history. Norman Barry’s The Tradition of Spontaneous Order is a useful historical review of the economic strand.

4.7 Orienting to a rule: the occasion of use

Where 4.6 is abstract, general, and may sound vaguely mystical,6 4.7 gets down to nuts and bolts.

So how do rational systems work, if they don’t mirror the True World? Module 4.7 answers: through non-systematic situated meaning-making. Put another way: by intelligent interpretation of the system as meaningful in specific but necessarily nebulous circumstances. Or: through the participants in an interaction orienting to a rule as a resource on that particular occasion. These answers are still highly abstract, but the module will show how it is possible to understand them, and to demonstrate their accuracy empirically, in extreme specificity and detail.

Heidegger’s Being and Time is the root text for non-systematic situated meaning-making. That book is extremely difficult, so I would recommend instead Hubert Dreyfus’s Being-in-the-World, a readable explanation of the relevant parts of Heidegger’s work. It is still relatively abstract.

The nuts-and-bolts understanding comes from ethnomethodology, which investigates the question “how do systems actually work” through minutely detailed observation and analysis of actual people actually using systems. This immediately eviscerates the rationalist view, and demonstrates the correct alternative fairly painlessly.

Empiricism for the win! Isn’t it obvious this is the right approach? And yet ethnomethodology remains almost entirely unknown outside a handful of academic departments.7 Part of the problem is the lack of an accessible introduction. I started with John Heritage’s Garfinkel and Ethnomethodology, which is excellent but not easy. (Garfinkel founded the field; his own writing is nearly impenetrable.)

Mainly ethnomethodology is difficult because it requires a new way of seeing. Many people report experiencing a sudden flip in perception when they “get” it. In order to “get it,” you have to set aside everything you think you know—in order to actually look, without mistaken assumptions. Otherwise, you hallucinate systems where there are none. Theoretical presuppositions get in the way of accurate observation.

The Anglophone rationalist tradition (analytic philosophy, cognitive science, artificial intelligence) assumes that systems live in your head. They don’t. Representations are not datastructures. Rules are not effective procedures. Plans do not engender action. Until you set those delusions aside, when you look at people acting, you keep asking “what rule in their head made them do that?”, which prevents you from seeing what is going on.

The Continental social tradition assumes that systems are vast, abstract structures of oppression, with elite-imposed power rules, which determine the details of individual interactions. This is backward. Ethnomethodology reveals social systems as extremely concrete, detailed patterns of interaction. Large-scale social structures are determined by these details, not the other way around. Constantly asking “how does this exemplify oppression?” prevents you from seeing what is going on.

So, what is going on? Rules work through their interpretation by the participants in a concrete situation. That interpretation bridges the gap between the system’s theoretical vocabulary and the nebulosity of the visible specifics. Such interpretation is inherently, necessarily improvisational and collaborative. In ethnomethodological terms, participants orient to rules. They take rules as a resource for making sense of what everyone involved is doing, but the rules don’t govern the action in any way. Rules are routinely violated; and then there are patterns of reaction and repair.

My brief paper “Computer Rules, Conversational Rules” may help explain this. It tries to join the ethnomethodological and cognitivist understandings, and explains ethnomethodological conversation analysis by analogy to computer networking protocols.

An orrery: Carlo G. Croce's reconstruction of Dondi's Astrarium

Taking different examples as prototypes leads to the “vision flip.” For the Enlightenment tradition, Newton’s theory of gravity is the universal prototype. To caricature only slightly, ideological rationalism imagines you have an orrery in your head computing F = GmM/r2, and it’s connected to your muscles and makes you do things. Newtonian mechanics is incredibly cool, but most things don’t work that way. For the Continental social tradition, the enclosure movement is the universal prototype. To caricature only slightly, social theory imagines that reality consists of a series of ever-more-monstrous enclosures. Enclosure was a big deal if you lived in Scotland in 1800, but most things don’t work that way.

We are eating breakfast. You say “Jam?” and I nod in a particular way, and you pass it to me. If I had nodded in a slightly different way, you would not have passed it. There are rules about nods, which are part of a larger system of conversational rules. You may not “know” the rules about nods, but you reliably orient to them, and interpret them accurately.8 As ethnomethodologists, we can find the rules by video recording people eating breakfast, and watching carefully, over and over.

This is a different prototype. Nods may seem trivial, but—I will argue—systems mostly do work that way.9

4.8 Patterns of meta-systematicity

Modules 4.8 and 4.9 teach meta-systematic skills, at a much more sophisticated level than 4.2. We’re no longer pretending, as we did there, that systems work systematically. Instead, we are taking into account the nebulosity of all systems, the nebulosity of the situations in which we use them, and the nebulosity of the system/situation interaction.

4.8 and 4.9 may be the most important part of the curriculum. Unfortunately, it is the part that is least-well understood: by me and, I think, by everyone. Still, there is already much to say; and because the topic is not much investigated, there may be much low-hanging fruit left for the picking.

In retrospect, “How To Think Real Good” was an attempt. I wrote it in a hurry, and I understand the issues better now, and it looks remarkably lame three years later! However, it does make the key point that we can only ever deploy rationality as a miscellaneous collection of oft-useful tools, rather than The Single Correct Way To Do Everything.

“How To Think” was case-study driven, and that’s probably the best method of investigation. It draws some general conclusions, but inevitably—given the nebulosity of the topic—those can only be nebulous. Insight comes from close examination of specifics (as in ethnomethodology). Finding and analyzing good case studies may be a significant task.

I have also found relevant insights in diverse not-obviously-relevant literatures (some of which I’ve alluded to in this post). Extracting and explaining these is another project. I’ll mention two works here, briefly:

Donald Schön’s The Reflective Practitioner: How Professionals Think In Action examines the interplay of three ways of knowing and acting: technical rationality, non-systematic tacit expertise, and meta-systematic reflection on the first two. The book is based on his close examination of the practice of professionals in five different disciplines, and presents many case studies. He starts from the observation that systematic rationality frequently fails when it meets nebulosity, but he shows how competent professionals can find creative and effective courses of action by drawing on tacit and meta-systematic resources. He observes that problem-finding and problem-formulation are as important as problem-solving, and that all these activities are improvisations in collaborative interaction with concrete, hard-to-characterize circumstances. [Update: Reader Brian Marick recommends Schön’s follow-on book Educating the Reflective Practitioner over the original; he says it’s clearer and more concise.]

Jean-François Lyotard’s The Postmodern Condition: A Report on Knowledge is one of the two root texts for postmodernism. Knowing this, you might not suspect that it was commissioned by the government of Quebec as a report on the influence of information technology on the exact sciences. Written in 1979, it’s astonishingly prophetic about the then-future impact of the internet—but that is not the reason to read it. You might also not suspect that, unlike the voluminous obscurantist blather of later postmodernists, it’s only 70 pages and reasonably clearly written. Lyotard’s main topic is the breakdown of the systematic worldview in the face of nebulosity, and the persistence of multiple, functional, partial systems despite that. He aims for “a politics that would respect both the desire for justice and the desire for the unknown.” This remains unfulfilled, and obstructed not least by the subsequent development of postmodernism—but I think still a worthy goal.

4.9 Fluid competence: Creating functional meta-systems

This module describes systems that recognize nebulosity, and the limitations of their own systematicity, and so are open to continual structural revision as they are illuminated by interaction with nebulous circumstances.

Several of the key figures I mentioned above moved from studying meta-systematic cognition in individuals and task groups to applying their insights to larger social structures. Schön, Kegan, and Brown all pioneered theories of “learning organizations” that continually rethink not just their methods but also their goals.10 They ask not just “how do we solve this technical or business problem” but also “what problems are we addressing? are they still the right ones?” and “what are we doing to support our staff in developing to address our business more effectively?” Answers to these questions are inevitably improvisational, collaborative, interpretive, and meta-systematic.

I would like to end 4.9 with a section titled “Conjuration: legendary feats of meta-rationality.” I have, actually, many examples in mind; but none of them are in STEM fields. Does this mean meta-systematicity is not, after all, useful in those fields? I don’t think so. Probably, instead, it is because STEM results have to be stated and evaluated in systematic terms. I suspect meta-rational insights produce many STEM breakthroughs, but the resulting journal articles don’t explain the thinking process. (A STEM paper is not supposed to include that.) And so this knowledge—which may be incredibly valuable—is generally lost. The rare discussions by major scientists about how to think—I discussed Feynman’s and Rota’s in “Real Good”—often talk meta-systematically.

5.0 The other dimensions

According to Kegan’s framework, developing a meta-systematic way of being affects every dimension of life. It completely reorganizes your self, your relationship with your self, and your relationships with others. That reorganization manifests in structurally identical ways in your family life, your understanding of ethics, how you plan projects, the way you act at work, and so on.

Kegan says that an epistemological shift—a new way of making meaning—underlies all the rest. He explains the epistemological dimensions of the other stage transitions in detail, but says little about this in 4-to-5 development.11 This hypothetical curriculum would supply that missing discussion.

I suspect—based only on anecdata—that STEM folks can make the 4-to-5 transition more readily than most others, if the epistemological dimension is brought to the fore. We care about epistemology in a way most people don’t; and it is an epistemological shift that drives the development of other dimensions of being.

Completing the hypothetical 4.9 module should bring the hypothetical STEM-educated student to a stage 5 epistemology.

It would then be helpful to explain the rest of the 4-to-5 shift by analogy with that epistemology. That is, a stage 5 marriage is structurally parallel to a stage 5 STEM research project; and so are stage 5 ethics, religion, management, art, and politics.

  1. 1.“Meta-systematicity” is non-systematic just by definitional fiat. This is not an empirical claim; rather, I’m declaring any systematic reasoning about systems to be non-meta-systematic for terminological convenience. Otherwise I’d keep having to say “non-systematic meta-systematicity,” which would be tedious. I do make the empirical claim that much reasoning about systems has to be non-systematic, but this is distinct from the definitional fiat.
  2. 2.Maybe this has been done by education theorists? They certainly look at high-school level STEM learning. There is also a relevant literature on “post-formal operations,” in the Piagetian framework, but I haven’t yet looked at it seriously.
  3. 3.This term was accidentally popularized by Donald Rumsfeld. It had been used by others for decades before, however.
  4. 4.And also as little. That is: humans avoid the critique by being non-systematic much of the time, and there’s no reason in principle to think artificial agents can’t do the same. In the first edition of this book, Dreyfus assumed that computers could only be programmed to operate systematically, which was reasonable given the claims of AI researchers of the time, who were attempting to program systematic rationality. The second edition adds a discussion of so-called “neural networks,” which supposedly do not attempt systematic rationality.
  5. 5.Because it’s highly condensed, it may be incomprehensible without some knowledge of the tradition. One key to understanding is that “Königsbergian” is a reference to Kant specifically. The supposed “true world” of Nietzsche’s stage 3 is Kant’s ding an sich, “the thing in itself.” That is the inaccessible “noumenon,” or true reality, as opposed to the defective “phenomenon” that appears to the senses. This is a catastrophically bad idea, which leads straight to nihilism.
  6. 6.Economic theory is not all that mystical, although its relationship with reality is often dubious.
  7. 7.John Seely Brown, the genius director of Xerox PARC, recognized its value, and hired several ethnomethodologists. They did some extremely interesting work on how people use information systems, with implications for how those could work better. One of these studies wound up making Xerox a great deal of money. I learned about ethnomethodology from that group when I worked there. I had lost track of JSB for the past 15 years or so; I see he has a new book that looks relevant to my interests!
  8. 8.The rules of grammar are also like this. “Adjectives in English absolutely have to be in this order: opinion-size-age-shape-colour-origin-material-purpose Noun. So you can have a lovely little old rectangular green French silver whittling knife. But if you mess with that word order in the slightest you’ll sound like a maniac. It’s an odd thing that every English speaker uses that list, but almost none of us could write it out.”
  9. 9.The ethnomethodological understanding is similar to Wittgenstein’s in Philosophical Investigations—in his discussion of the “builder’s language game” for instance. The builder says “slab! there!” and his assistant puts one there. This is an armchair thought-experiment, however. Ethnomethodology’s extensive empiricism reveals how language, interpretation, cooperation, and improvisation work in much greater detail.
  10. 10.All three of them worked as management consultants. By the way, the title of this post is—obviously—a riff on Mark McCormack’s book, whose cover is the header image for this page. The revised edition is better. I found it very helpful when I was running a tech startup. What they do teach you at Harvard Business School is mostly financial analysis, which is dead easy for anyone with a math degree. What they don’t teach you are people skills—not always as easy for someone (like myself) with a math degree. Financial analysis is a rational system; people skills are not. Actually, I gather that HBS teaches some “soft skills” nowadays—perhaps partly in response to McCormack’s criticism.
  11. 11.Kegan says little about stage 5 epistemology in the works of his that I know, anyway. I’ve read only a fraction of his output. His theory is rooted in Piaget’s, which goes only to stage 4 (“formal operations”). Other researchers have proposed theories of stage 5 “post-formal operations,” which Kegan mentions only in passing. Based on only cursory investigation, my impression is that this work isn’t very good, which may be why Kegan decided to pass over the matter in silence.