The average student does not exist

In the late 1940s the US Air Force had a problem it could not explain. Planes were crashing at an alarming rate, and it was not mechanical failure and it did not look like pilot error. Something else was wrong.

The cockpits had been designed to fit the average pilot, using body measurements taken from a survey in the 1920s. A reasonable first suspicion was that pilots had simply got bigger since, so in 1950 the Air Force measured 4,063 pilots across 140 dimensions to update the numbers.

A young lieutenant working on the project, Gilbert S. Daniels, asked a different question. Instead of computing the new average, he took the ten dimensions most relevant to cockpit design — height, chest circumference, arm length, and so on — defined “average” generously as the middle thirty per cent of the range on a given dimension, and then counted how many of the 4,063 pilots were average on all ten at once.

The answer was zero. Not a small number. None of them.

Even on just three of the ten dimensions, fewer than one pilot in twenty qualified. The cockpit had been carefully fitted to a man who did not exist, and it therefore fitted nobody — including, crucially, the pilots who were close to average on most dimensions and badly off on one, which is almost everyone.

The Air Force’s response is the interesting part. It did not commission a better average. It banned the concept and required adjustability instead — adjustable seats, pedals, straps, helmet sizes. The adjustable seat in your car descends directly from that decision.

Every course you have taken was fitted to that pilot

A course is one artefact serving many people. That is not a flaw in any particular course; it is the definition of the format, and it is what makes courses affordable. But it forces the same design move the Air Force made in 1926: pick a representative reader and build for them.

Now list the dimensions that actually determine whether a course works for a given person:

• How much of this specific subject they already know
• What adjacent things they know that could be built on
• Whether they have a working mental model or a wrong one to be dismantled
• Why they are learning it — curiosity, a deadline, a job change, an argument
• How much depth they want, which is not implied by any of the above
• How fast they read, and whether they want the derivation or the result
• What they will do with it within a month

Seven dimensions, and I have picked conservatively. Run Daniels’ test on any population of learners along those seven and you get the same answer he got. There is no average student. There is a composite that describes nobody in the room.

What “wrong for you” actually feels like

The failure is not usually experienced as a mismatch. It is experienced as a verdict about yourself.

Course too slow for your background and you conclude the subject is shallow, or you are wasting your time, and you leave — usually without noticing that you left at the point where it was still covering things you knew.

Course starts above you on one dimension and you conclude you are not smart enough, or not ready, and you go looking for a more basic one — which will restart from zero on all seven dimensions, including the six where you were fine.

Both of those read as facts about the learner. Both are facts about the fit.

Courses have almost no adjustability

Here is the thing that should be strange and somehow isn’t. The Air Force’s fix was not a better fixed design. It was making the thing adjustable along the dimensions that varied.

What can you adjust in an online course?

Playback speed. That is essentially the entire list. Some let you skip around, which is adjustment by manual labour and requires knowing in advance what you can safely miss — that is, requires already understanding the material well enough to judge, which is the thing you came here not to have.

You cannot tell a course you already know linear algebra and have it stop explaining matrices. You cannot tell it you want the mechanism rather than the analogy. You cannot tell it you are coming from a related field and would understand the whole thing faster through that lens. The seat does not move.

Why it was built that way

Because until very recently the economics allowed nothing else.

Writing a course is expensive — months of an expert’s time. That cost can only be recovered by serving the same artefact to a large number of people, which means the content has to be fixed before any of them arrive. Fixed content means one assumed reader. One assumed reader means the average. This is not a failure of imagination by course authors; it is the direct consequence of writing costing what it cost.

Tutors do not have this problem, which is why one-to-one tutoring has been the gold standard for as long as teaching has existed and has never once been affordable at scale. A tutor asks what you know, watches you work, and adjusts continuously. They are the adjustable seat.

The interesting question now is not whether personalised instruction works — that has never been in doubt — but whether the cost of producing a fitted course has fallen far enough that it no longer has to be written once for a pilot who does not exist.