Book Review: Deschooling Society

Want some career advice? Make sure that your name becomes synonymous with a key area of the company. It’s unlikely that they will fire John-the-IT-guy. You’ve reached your goal, if your colleagues and boss can’t imagine how the IT infrastructure of the company could ever function without you. At this point, it doesn’t matter if you’re lazy or screw up regularly. The association John<->IT becomes so ingrained in everyone’s head that whenever there’s a serious problem, you can easily convince your colleagues that it’s not your fault. You just didn’t have the resources you needed. Or a third party screwed up. In fact, serious problems are good for you. Every time there is a crisis, you can demand more money. 

This may seem pretty obvious. What is less obvious to most people is exactly this trick is used by all kinds of institutions. The “company” in that case is society. 

The military has managed to become synonymous with national security, churches with religion, universities with research, the police and prisons with security, and schools with education.

Most people can’t imagine a functioning society without these institutions. This is completely analogous to how in the example above, people can’t imagine a company without John-the-IT-guy. 

And even if any of these institutions screw up badly or produce unsatisfactory results, their existence is never questioned. Instead, they are rewarded with more money – just as John-the-IT-guy. For example:

  • Too many national security threats -> more money for the military.
  • Too much crime -> more money for the police.
  • Too little exciting research results -> more money for universities.
  • Too many badly educated people -> more money for schools. 

Ivan Illich calls this process escalation or social addiction and it can be characterized by the “tendency to prescribe increased treatment if smaller quantities have not yielded the desired results”.

If you don’t think too hard about it, this seems reasonable. In particular, you might want to argue that if John screws up, he certainly will be replaced by a more competent person. Of course! Let’s say that Bob replaces John. Bob-the-IT-guy might be a bit more competent or slightly less competent than John. But what really matters is that the company will always have IT problems and an IT guy. As long as the company remains trapped in the paradigm that it needs an IT guy, all improvements will be marginal at best. No IT-guy will be so stupid to improve the IT infrastructure in such a way that he becomes dispensable. 

Applied to institutions, this observation is known as the Shirky principle, after Clay Shirky, who observed that “institutions will try to preserve the problem to which they are the solution.”

For example, for the military it was certainly true that “the higher the body count of dead Vietnamese, the more enemies the United States acquires around the world” and this, in turn, requires an even larger military. Similarly, “jail increases both the quality and the quantity of criminals, that, in fact, it often creates them out of mere nonconformists”. And “mental hospitals, nursing homes, and orphan asylums provide their clients with the destructive self-image of the psychotic, the overaged, or the waif, and provide a rationale for the existence of entire professions”. 

With the Shirky principle in mind, it becomes obvious that “better” institutions never can bring satisfying solutions. Instead, we have to drop the paradigm that the institutions are synonymous with certain aspects of society. Once we do that, a world of possibilities opens up. If we move beyond the framework of institutions, there is a real chance at significant progress compared to the incremental progress that is possible in the institutional framework.  

This is one of the key arguments Ivan Illich lays out in his book Deschooling Society. Interestingly, he completely overlooks the role of the state which, in some sense, is the meta institution that enables most other institutions. This logical gap was filled only three years after the publication of Deschooling Society by Murray Rothbard in his book The Anatomy of the State. Both books are similar in spirit, driven by indefinite optimism, which isn’t too surprising given that they were written shortly after the first moon landing. The main difference is that Rothbard declares “the state is the devil”, whereas Illich declares “schools are the devil”. 

Illich passionately claims “deschooling is, therefore, at the root of any movement for human liberation.” and “School is the advertising agency which makes you believe that you need the society as it is.”

I think the case for these claims is much weaker and they’re less interesting than the general claim that the institutional framework is problematic. Nevertheless, it’s fortunate that Illich decides to focus on one key institution. Putting the magnifying glass on one of the institutions allows us to grasp the larger patterns and to think about concrete alternatives. And schools are a great choice because the existence of educational problems is something that everyone can agree on.

So in the following, I won’t treat Illich’s book as a pamphlet that declares “schools are the devil”. Instead, I’ll treat it as a case study for the general problems of institutions, and for how a society without institutions might function. 

The trouble with schools

Before we can discuss how well schools fulfill their educational goals, we have to clarify what we’re talking about. The three most commonly proposed educational goals are:

  • Homogenization.
  • Intellectual Cultivation (Plato).
  • Realization of Potential (Rousseau). 

Typically, people who learn towards different dimensions, argue for different kinds of schools. Schools try to find compromises and thus inevitably fail to meet any of the goals. Even though Illich is an outspoken Level 2 thinker since he dismisses homogenization and intellectual cultivation completely, he correctly points out that new schools inspired by Rousseau’s vision will not bring significant improvements. 

Before we can discuss how well schools fulfill their educational goals, we have to clarify what we’re talking about. The three most commonly proposed educational goals are:

One of the strongest arguments against schooling he discusses is discrimination. Currently, it’s completely normal to discriminate in hiring based on previous attendance at some curriculum. This, obviously, favors children with richer parents and also otherwise makes little sense.  School years can often be translated as an “enforced stay in the company of teachers” and rarely says anything about how much someone understands.

Illich calls for dramatic measures: “To detach competence from curriculum, inquiries into a man’s learning history must be made taboo, like inquiries into his political affiliation, church attendance, lineage, sex habits, or racial background. Laws forbidding discrimination on the basis of prior schooling must be enacted. Laws, of course, cannot stop prejudice against the unschooled – nor are they meant to force anyone to intermarry with an autodidact but they can discourage unjustified discrimination.”

It’s hardly a controversial claim anymore that modern schools are mostly about signalling. (This issue is discussed more extensively in the book The Elephant in the Brain by Kevin Simler and Robin Hanson.) However, by invoking the word “discrimination” Illich manages to amplify the message. 

Certification is not only problematic because it leads to discrimination but also because of the well-known cobra effect. One intention behind certification is certainly to motivate students to learn subjects more deeply. In reality, however, it’s completely normal that students are actively discouraged from doing the kind of curiosity driven deep dives into a subject that leads to truly deep understanding because it would affect their grades negatively. 

An important related aspect is that certificates also plays an important role on the teacher side. There are lots of people who would love to share what they know with others. But in the current system, they’re effectively locked out from participating. Illich concludes: “Certification constitutes a form of market manipulation and is plausible only to a schooled mind.”

The observation that schools produce “schooled minds” is a further key argument. On the one hand, there is what Illich calls a “hidden curriculum” which aims to fulfill the homogenization goal mentioned above. Students learn not just about the history of Greece but also to follow orders and to work in silence on obviously nonsensical projects. (A great book dedicated to this issue is Disciplined Minds by Jeff Schmidt.)

On the other hand, the longer people attend school the more likely it is that they will defend the school system. It’s really hard to admit to yourself that you’ve wasted years of your life. The more comfortable option is to defend the current system to justify your decisions in the past. This is the well-known consistency bias. Since most people invest years of their lives into the school system, there is arguably no other area of life that is so heavily dominated by it.

A third problematic aspect of schools for the minds of students is that they corrupt their attitude towards learning and learning objects. The focus in school is primarily on achieving, having and consuming and rarely on growing and developing. To use the terminology introduced by Eric Fromm, while educational needs are clearly being needs, schools forces students into a having mode and addressing being needs from a having mode inevitably leads to unsatisfactory results. In schools, students are taught to use books, teachers, and their peers to their advantage in order to have better grades. In contrast, the being mode is characterized by a symbiotic relationship of mutual transformation.

Now, of course, it’s always easy to criticize without proposing an alternative. Luckily, Illich doesn’t fall into this trap. He proposes a concrete alternative which, even 50 years later, makes a lot of sense. 

How can education work without schools?

The alternative Ivan Illich proposes, which he calls “learning web” or “learning network” or “educational web”, has four key components:

  • Reference Services to Educational Objects – which facilitate access to things or processes used for formal learning. Some of these things can be reserved for this purpose, stored in libraries, rental agencies, laboratories, and showrooms like museums and theaters; others can be in daily use in factories, airports, or on farms, but made available to students as apprentices or on off hours.”
  • Skill Exchanges – which permit persons to list their skills, the conditions under which they are willing to serve as models for others who want to learn these skills, and the addresses at which they can be reached.”
  • Peer-Matching – a communications network which permits persons to describe the learning activity in which they wish to engage, in the hope of finding a partner for the inquiry.”
  • Reference Services to Educators-at-Large – who can be listed in a directory giving the addresses and self-descriptions of professionals, paraprofessionals, and freelancers, along with conditions of access to their services. Such educators, as we will see, could be chosen by polling or consulting their former clients.”

Thanks to the internet, access to learning objects has already become much easier and there are all kinds of skill exchanges like Codementor and italki. But what about the remaining two puzzle pieces? 

To some extent, peer-matching is happening on social platforms like Twitter. But many people don’t have a large enough following on these platforms to find peers or are too introverted to reach out. A dedicated platform would be amazing and I’m not sure why it currently doesn’t exist.

Illich has a concrete vision for how it might function: “The operation of a peer-matching network would be simple. The user would identify himself by name and address and describe the activity for which he sought a peer. A computer would send him back the names and addresses of all those who had inserted the same description. […]  People using the system would become known only to their potential peers. […] A complement to the computer could be a network of bulletin boards and classified newspaper ads, listing the activities for which the computer could not produce a match. No names would have to be given. Interested readers would then introduce their names into the system.”

Moreover, he emphasizes that the matchmaking should be based on common interest in a book, article, film or recording. Less specific matchmaking based on an idea, a topic, or issue are necessarily teacher-centered and hence less suitable for a peer-to-peer framework. “Theme-matching is by definition teacher-centered: it requires an authoritarian presence to define for the participants the starting point for their discussion.”

This matches my experience. I organized different “theme-matched” learning groups in the past, and they only worked well if there is someone more experienced who guides the group. 

Another important observation by him is that whoever operates the learning web, should do their best to stay out of people’s way. “Today’s educational administrators are concerned with controlling teachers and students to the satisfaction of others-trustees, legislatures, and corporate executives. Network builders and administrators would have to demonstrate genius at keeping themselves, and others, out of people’s way, at facilitating encounters among students, skill models, educational leaders, and educational objects. Many persons now attracted to teaching are profoundly authoritarian and would not be able to assume this task: building educational exchanges would mean making it easy for people – especially the young – to pursue goals which might contradict the ideals of the traffic manager who makes the pursuit possible.”

The final puzzle piece, “Reference Services to Educators-at-Large”, currently doesn’t exist either although it’s arguably, the most important part. Most learning journeys will fail unless there is proper guidance, encouragement and support. These are things that teachers, at least in theory, provide in the current system. Unfortunately, most teachers are far too busy with other tasks to fulfill this role properly. By unbundling this piece of the educational puzzle, teachers who are passionate about guidance, encouragement and support could dedicate all their energy in that direction.

The only thing I can think of that currently offers something along these lines is StackExchange and related platforms. But while these websites are quite good at providing support, they don’t provide any form of guidance and have many flaws that could be fixed by a proper reference service. In particular, sites like StackExchange are completely impersonal and you only get answers from people who don’t know your background. Explaining for each question what you’re currently doing and what you already know is extremely laborious. Moreover, the quality of the answers certainly would be higher if people were paid for them. And last but not least, most of the sites suffer from an overly strict moderation policy that contradicts the principle of “staying out of people’s way”. 

A “Reference Services to Educators-at-Large” would allow learners to get support, guidance and encouragement from someone more experienced who knows their background and is paid for his or her service. Moreover, if the system is based on reputation, not credentials, there’s certainly no shortage of people who can help. 

Making the deschooled society a reality

Let’s imagine that all the puzzle pieces discussed in the previous section are in place. This would provide a complete framework for serious learners that has nothing to do with schools. Would schools then become completely irrelevant? Certainly not.

At this stage, arguably, the biggest remaining obstacle would still be the “discrimination-based-on-credentials” issue. Most people would continue to go to school because this is the only way to get the credentials that you need to avoid being discriminated. 

I don’t think it’s realistic that there will be laws against “discrimination-based-on-credentials” anytime soon. Nevertheless, there is a realistic chance that the “learning web” becomes a success.

Some brave people would recognize its potential. These early adopters would become role models and more and more people would be attracted to the freedom offered by the alternative path. Simultaneously, companies will recognize that learning-web-learners are better hires  such that credentials eventually become irrelevant. In particular, more and more companies will understand that “a test of a current skill level is much more useful than the information that twenty years ago a person satisfied his teacher in a curriculum in which typing, stenography, and accounting were taught.”

Moreover, the best teachers will welcome the opportunities offered by the learning web. In the current system, “schoolteachers are overwhelmingly badly paid and frustrated by the tight control of the school system.” Therefore, “the most enterprising and gifted among them would probably find more congenial work, more independence, and even higher incomes by specializing as skill models, network administrators, or guidance specialists.”

But even if all this happens, it would be foolish to believe that schools become completely irrelevant. Humans are primed to think in terms of binary opposites: schools = bad, learning web = good. In the real world, there is rarely such a clear line. Moreover, progress usually occurs in a manner of transcending and including not by wiping out what came before. So it’s likely that schools will continue to play an important role. However, they will be embedded into a larger ecosystem. This will allow them to focus on what they’re best at (e.g. homogenization) while the fulfillment of other educational goals is delegated to different kinds of structures. 

In summary, irrespective of what you make of his other ideas, I’m convinced that Ivan Illich’s proposal for a learning web is a fantastic idea. I’m not entirely sure why not all the puzzle pieces have been built so far. But I would love to help to make it happen. 

Why is learning still hard and what can we do about it?

How many people truly understand general relativity? Or quantum field theory? Or any other challenging subject?

If you spend enough time in academia you slowly but steadily realize that the number is tiny. Even most students and professors of physics don’t understand the best fundamental theories of nature that we have.

And what about the billions of people who never studied physics at a university? There might be a few autodidacts here and there but the number remains tiny.

This is puzzling if you think about it. Many people have a deep interest in understanding fundamental ideas that explain how the world works. Yet hardly anyone reaches an understanding beyond a superficial level. Many would love to contribute to our communal quest to decipher the source code of the universe. Yet an incredibly large talent pool is locked out from participating.

The only possible explanation is that it’s still incredibly hard to learn challenging subjects.

But why? And how can we improve the situation?

Let’s start by talking about three observations that answer the first questions.

1.) To some extent, learning has to be challenging. However, this idea has been fetishized.

To quote Sönke Ahrens:

“Learning requires effort, because we have to think to understand and we need to actively retrieve old knowledge to convince our brains to connect it with new ideas as cues.”

Unfortunately, lots of teachers (especially at the university level) take this whole idea to a level that I can only as absurd. They leave out crucial details and tell the students they should figure it out themselves knowing fully well that this is too difficult of a problem and the students don’t have time for it. They select homework problems solely under the aspect that they’re sufficiently difficult irrespective of how much students can learn from them. They don’t follow any grand plan that cumulates in a series of “aha”-moments. They don’t even try to design the lectures in a way that makes them enjoyable. The mantra seems to be “the more I confuse the students the better”.

Sometimes this is intentional, sometimes it isn’t. There are many facets of this problem:

  • If you confuse the students they are less likely to ask questions. Each questions potentially challenges your status and hence it’s a good strategy to avoid them.
  • Most university professors are too busy with other obligations to have sufficient time to prepare for their lectures.
  • Professors don’t get selected for their teaching skills. Metrics like the number of citations are far more important.

These observations lead us directly to the next point.

2. Current educational institutions are mostly about signalling.

Institutions focus on metrics (grades, citations) that primarily select people who are conscientious and conformist and tells us little about understanding. (If you doubt this, just try take a slightly non-standard position in an oral exam.)

Or as Sanjoy Mahajan puts it:

“Traditionally taught science and mathematics teach little except obedience.”

This point has been elaborated on by many much more gifted writers, so I’ll just refer you to them in case this is a new idea for you. Good starting points are:

One often overlooked aspect I want to emphasize is that this also affects textbooks. Most textbooks are not written to help the reader understand. Instead, they’re written to signal how smart the author is. That’s why most textbooks are impossible to understand. And to make matters worse, these textbooks get recommended over and over again because if you recommend a book that’s difficult to understand, you signal how smart you are.

The problems of the traditional learning institutions lead us to the next observation.

3. There hasn’t been a revolution in learning despite the internet

Despite all the possibilities offered by the internet, no one has yet come up with a framework that actually puts them to use to make it dramatically easier to learn challenging subjects.

Yes, there are a few nice lectures freely available online. But this was never the bottlenecks and thus hasn’t improved the situation in any significant way. Excellent textbooks have been available since ages and are superior to lectures anyway.

A book allows you to go jump around and move through the information on your own pace. This is extremely cumbersome in video lectures. And if anything, online lectures are worse than real world lectures (which are already an awful medium) since it’s much easier to get distracted. Funny cat videos are always just one click away. (And no, there aren’t different learning types.)

Moreover, the flexibility to start the course whenever you want and to pick from lots of different lectures usually leads to analysis paralysis and hinders learning.

For quite a while I thought that there is nothing we can do about it. I was convinced that there are some intrinsic flaws (e.g. proximity to cat videos) that makes online learning ineffective.

But this isn’t true. It’s just that all current online learning frameworks are missing one or several essential puzzle pieces. Online learning has not to be synonymous with recording a bunch of videos and putting them in a MOOC.

Essential Aspects of Effective Learning Environments

To understand which puzzle pieces are potentially essential to create an environment that makes it possible to understand even the most challenging topics, it makes sense to analyze what universities are doing. While universities are far from optimal learning environments, this analysis will allow us to understand what is missing in current online learning frameworks. Once we’ve identified these puzzle pieces we can discuss how they can be realized and improved upon in an online setup.

So, let’s ask ourselves: What does a university provide to help students learn challenging subjects? There are at least the following eight puzzle pieces:

  • Community
  • Support
  • Guidance
  • Accountability
  • Credentials
  • Feedback
  • Content

With this list in mind, it becomes clear why most online learning experiments have failed so spectacularly. Many of them primarily focussed on providing content online, which is really just one tiny aspect and certainly not the bottleneck.

MOOCs typically don’t foster any sense of community, offer only limited support, feedback, guidance, or accountability (if any), and only provide credentials that are hardly worth anything. If we then add to this that the content is usually just a recording of mediocre lectures, the package doesn’t look enticing.

Of course, as discussed above, universities are hardly any better in most aspects. And some MOOCs are excellent at mimicking the university experience online. But the difference is that universities get away with their far from optimal package since they provide sufficient accountability and incentives (credentials) to make students endure the whole thing.

Universities are simply not the gold standard of learning experiences that they are often thought to be. Moreover, if you try to mimic something there are necessarily some losses along the way. Just to name two examples:

  • A video recording of a lecture is never as good as a live lecture since you can’t see everything at once.
  • Online certificates simply don’t have the same standing as established offline certificates.

So while the university package might be just good enough, due to these losses the online equivalent no longer crosses the threshold.

To illustrate this point, let’s rate how good universities are at providing the essential puzzle pieces listed above. I would say:

  • Community: 7/10
  • Support: 3/10
  • Guidance: 3/10
  • Accountability: 10/10
  • Credentials: 10/10
  • Feedback: 5/10
  • Content: 4/10

In total, we have 42 out of 70 possible points. (Feel free to put in your own ratings.)

Now what happens if we mimic the same thing in an online environment? In my experience, a typical result can be rated as follows:

  • Community: 2/10
  • Support: 3/10
  • Guidance: 2/10
  • Accountability: 1/10
  • Credentials: 2/10
  • Feedback: 2/10
  • Content: 3/10

Certain aspects are mimicked quite okay while others are almost completely lost. I might be a bit negative here but in my book we end up with 15 out of 70 possible points. Even if you’re a bit more generous, the result is certainly below the threshold (say 40 points) that separates effective and ineffective learning experiences.

With these observations in mind, we can finally think about how we can design more effective learning environments.

Revolutionizing Learning and Teaching

An exhaustive discussion of how each puzzle piece can be improved and implemented in an online setup requires several additional articles. The following list merely offers a few starting points.

  • Community. In theory, this should be the easiest part to improve in an online setup. The internet is a tool that connects people and hence should make it easier for communities to form. Unfortunately, this rarely happens when it comes to online learning. Most online communities are too open, too heavily moderated, too anonymous, and focus solely on text messages. Most people don’t feel comfortable shouting stuff from the rooftops for the whole world to hear. Moreover, it’s quite hard to make friends solely by exchanging topic-specific text messages. As a result, no real sense of community emerges. The rise of smaller private communities supplemented by virtual conferences, virtual book clubs and virtual co-studying places have the potential to improve the situation dramatically.
  • Network. Similar comments as for community building apply. An additional aspect is that it’s possible to network online simply by putting yourself out there. Most people remain invisible. Hence, if you’re interested in some topic and regularly publish things related to it, opportunities will show up at your door step.
  • Support. The way universities offer support is far from optimal. Lecturers are annoyed by questions since they make it harder to get through all the material. Tutoring sessions mostly focus on homework platforms and there is rarely any time for discussions. In contrast, online platforms like StackExchange are great at providing answers to specific questions. However, so far, the platforms function solely on a voluntary basis and are quite impersonal. Respondents typically don’t know your full background and the context of your question, and it’s simply a matter of luck if someone qualified feels motivated to answer your questions. This certainly limits the quality of answers. In addition, most platforms suffer from overmoderation. A paid full-time staff of tutors and a more decentralized, personal approach would improve the situation.
  • Guidance. There are quite a few learning curricula and learning roadmaps available online for most topics. However, these one-size-fits all curricula are far from optimal for individuals. Moreover, they are hardly ever detailed enough to provide enough guidance for individual decisions. Personal mentoring could solve these problems but also automated solutions could work.
  • Credentials: Online credentials exist but are not a good solution. They’re just the same (already bad) thing slightly worse. One proposed solution are college equivalence degrees. Although they could represent an improvement, they, like their offline equivalents, would inevitably get corrupted by the cobra effect. So, instead of the mimicking something intrinsically flawed, the opportunity should be used to replace it with something better. Credentials signal conformity and conscientiousness but certainly not deep understanding, creativity and high agency. To improve the situation, we could replace exams by projects and hence credentials by portfolios.
  • Feedback. Some MOOCs allow students to get feedback on their solutions of homework problems. However, homework problems (offline like online) are often primarily designed to keep students busy and to filter out certain types of students. Solutions would be to create exercises that help the students deepen their understanding (e.g. by explaining what they’ve learned to others) and to get real-world feedback on their projects.
  • Accountability. This is something universities are good at and that current online learning frameworks are lacking. A solution is to offer paid courses for small cohorts with a fixed starting and end date and regular virtual meetings.
  • Content. Despite of all tech innovations, books are still the best medium if you want to understand something deeply. You can read a book far away from all distractions, move through it at your own pace and jump between chapters. But this doesn’t mean that all books are automatically great or that no improvement is possible. In fact, as mentioned above most textbooks are impossible to learn from. And there are many ways of how even good textbooks can be further improved. A great example is to include multi-level content. Moreover, ideally textbooks are supplemented by something more human to provide all the little social cues that get lost when we try to transfer information in text form.

The message to take away is that attempts to create new learning environments are doomed to fail if they ignore one or several of the essential puzzle pieces.

But, of course, it’s not enough to just throw ideas out there. In the past, I’ve primarily focussed on the “content problem” by writing reader-friendly textbooks and by building a multi-layer wiki since I didn’t understand the bigger picture. Currently, I’m working on something that tackles remaining puzzle pieces too. So if you find the ideas outlined above interesting and want to get involved, feel free to send me a message.

PS: It’s a fun exercise to rate different online learning products and platforms using the 8 criteria discussed above. I would love to hear who gets the highest rating in your book!

How I learned to learn physics

“You do not understand an argument, until you’ve found the major flaws in it. For any problem complex enough to be interesting, there is evidence pointing in multiple directions. ”


While there are many models that try to encapsulate how learning and understanding works, I recently came across one particular model that I keep thinking about and find extremely useful.

The model is a simple 3-level model and was proposed by Nat Eliason here.

The model describes remarkably well how I reached maturity in my thinking about different physics topics and since Nat didn’t mention physics, I want to discuss some examples below.

But first, a short summary of the model.


Level 1

Level 1 is called “Blind Ideology“. Everyone starts at this stage for any given topic. This stage is

“characterized by the wholesale adoption of the beliefs, attitudes, and lifestyles that were thrust onto you by your upbringing and environment. [….] Level 1 thinkers have an ideology they’re fixed to, and their blindness to it makes them throw out contrary opinions as heresy.”

A great example is diet. Here, Level 1 means that you eat what your parents taught you to eat, which in most cases is the standard Western diet.

Level 2

Level 2 is called “Chosen Ideology“. At this stage, people realize that the first best thing they were taught isn’t the best thing that exists and they become obsessed with another ideology. As Nat describes it

“If you know someone who believes in something and is annoying about it, they’re most likely at Level 2.”

We reach Level 2 after a “Moment of Clarity“. During such moments we realize that we have been driving with blinders on.

For the diet example above, Level 2 means that you become obsessed with something like Low-carb, Paleo, Veganism etc. At this stage, you are convinced that, for example, Paleo is the only way to go and every other way to eat is stupid.

Level 3

Finally, there is Level 3, which is called “Ideology Transcendence“. At this stage, we are able to sample the best bits from pre-packaged belief systems. At Level 3 we realize that no pre-packaged ideology is a perfect fit for us and we start developing our own. We start studying all ideologies that are out there and pick from each one only those parts that are of use for us.

The step from Level 2 to Level 3 is only possible through lots of Moments of Clarity. Only when we are exposed to lots of contrarian points of view, we can recognize the flaws in every pre-packaged belief systems. To reach Level 3 we must read books and articles that make us uncomfortable.

Regarding the diet example, Level 3 means that you recognize that different people respond differently to different diets. Everyone has different genes and therefore everyone has to experiment to find a diet that is a good fit. However, no pre-packaged diet can be a perfect fit for everyone.

A good test if you’ve already reached Level 3 are “Brake Lights”:

“When you react emotionally to information, any information, that’s a sign of Level 1 or Level 2 thinking. If you truly had a well-rounded stance on a topic and cared about enhancing your understanding of it, you would not react emotionally to anyone else’s opinion.”


It’s important to note that at Level 3 there is a “Strange Loop“. After enough time you will build an ideology of your own by picking the best stuff from other ideologies and adding something of your own. However, as soon as this happens you are again back at Level 2 since you are again following an ideology. Then, you must again search for flaws in your thinking and get exposure to contrarian points of view. In other words, Level 3 starts again. Level 3 is a stage of constant deliberate uncertainty.

The notion “Strange Loop” was coined by Douglas Hofstadter in his book “Gödel, Escher, Bach“:

“The “Strange Loop” phenomenon occurs whenever, by moving upwards (or downwards) through levels of some hierarchical system, we unexpectedly find ourselves right back where we started.”

In some sense this a miniature version of the whole scientific process. We can never know anything in the real world with 100% certainty. The only thing we can talk about is the level of confidence we have in a given theory, model or idea. Ultimately, today’s paradigm-shifting theory will become tomorrows standard theory and will again be replaced by another paradigm-shifting theory.

Nat discusses several other examples and most importantly ways to actively “level up”. It’s much better than this short summary and I highly recommend reading it.

But now, let’s discuss what all this means for physics.

Physics beyond Ideologies

Quantum Mechanics

  • Level 1 is the standard “Shut up and Calculate” approach that everyone learns in the lectures and standard textbooks.
  •  Level 2 thinking is becoming obsessed with, for example, “Bohmian Mechanics” or the Everretian “Many-Worlds interpretation”.
  • Level 3 thinking is realizing that none of these approaches is entirely correct and starting to develop your own way of thinking about quantum mechanics.

Gauge Symmetry

  •  Level 1 thinking is that gauge symmetry is a neat trick to derive the Lagrangian of the Standard Model and otherwise only necessary to prove renormalizability.
  • Level 2 thinking is becoming obsessed with the geometrical interpretation of gauge symmetry in terms of fiber bundles or with the idea that gauge symmetries aren’t fundamentally important after all but merely redundancies in our description.
  •  Level 3 is when you realize that gauge symmetries are indeed only redundancies, but carry a lot of physical meaning that isn’t captured by fiber bundles or the “neat idea” narrative.

Quantum Field Theory

  •  Level 1 is again the standard “Shut up and Calculate” approach that everyone learns in the lectures and standard textbooks. For quantum field theory this meany learning how to calculate Feynman diagrams and path integrals without caring about their meaning.
  •  Level 2 thinking is becoming obsessed with, for example, Supersymmetric Quantum Field Theory or String Theory.
  •  Level 3 thinking is realizing that none of the existing “beyond QFT” frameworks is the final answer. Maybe there are no quantum fields after all, since every time we took the field idea seriously we ended up with horribly wrong predictions (Monopoles, Strong CP violation,  Domain Walls etc.).

General Relativity

  •  Level 1 is the conventional narrative that in General Relativity there is no longer a gravitational field, but instead, gravity is merely a result of the curvature of spacetime.
  • Level 2 is the realization that you can turn this whole idea around and argue that the essence of general relativity is that there is no spacetime at all but only interacting fields. The only thing that exists are points where spacetime trajectories of field excitations meet. Only this way spacetime emerges. Another possible Level 2 understanding is “GR is the unique theory with no absolute object”, as coined James L. Anderson in his book Principles of Relativity Physics. (I actually have a friend who is really obsessed with this idea.)
  •  Level 3 is… I have no idea. I find the level 2 idea outlines above extremely cool and I guess this means I am stuck at level 2 for now. But if you know any articles that could help me improve beyond Level 2, please send them my way.

Some thoughts on how to level up in physics

After reading Nat’s essay I started thinking about how I could actively improve my learning process by taking the various 3 levels into account.

I started by assessing at what level I current am for various topics. (It turned out I’m still at level 1 or 2 for many physics topics).

Then I started to think about how I can get from Level 1 to Level 2. The crucial step here is recognizing that there is more than what we learn in lectures and the standard textbooks.

Level 2 ideas usually can’t be found in textbooks. Instead, they must be actively discovered. Often it’s just a side remark in a paper, book, blog post or at StackExchange that initiates the moment of clarity. Afterward comes a period of “going down the rabbit hole” where I try to trace any reference and comment on the alternative approach.

Finally, after enough research, I slowly realize that the alternative approach I became obsessed with is not the final answer. Level 3 thinking requires that I recognize that there is more than one reasonable idea of how to go beyond what we learned in lectures and textbooks.

To stay at Level 3 I must be constantly exposed to ideas that challenge my current beliefs. If I become too certain of a given idea I fall back to Level 2. Level 3 is uncomfortable and lonely.

To summarize: To level up you must read broadly. If you only stick to the books that your professor recommends you will stay at Level 1. Read books and articles by experts, read blog posts, read comments at StackExchange or at the PhysicsForums, read stuff by weird unknown guys. It doesn’t matter as long as they do not all repeat the standard story over and over again. As soon as some alternative approach sparks your interest it is necessary to dig deep and understand it from all possible angles. While it is extremely helpful to become obsessed during this phase, this obsession should always end after some time. At some point, it is always necessary to recognize that there is no universal pre-packaged answer.