I gave a commencement address at Caltech in 1974. It was about the one thing they do not teach in school: how to not fool yourself.

The first principle is that you must not fool yourself, and you are the easiest person to fool.

Let me tell you what I mean.

The Cargo Cults

During World War II, the U.S. military built airstrips on islands in the South Pacific. Cargo planes landed, bringing food, medicine, equipment, and all the material wealth of industrial civilization. The islanders, who had no context for this technology, watched the planes land and the cargo appear.

After the war, the military left. The planes stopped coming. The cargo stopped arriving.

The islanders wanted the cargo back. So they did what was logical from their perspective: they imitated everything they had seen. They built runways out of dirt. They lit fires along the sides at night. They carved headphones out of wood and sat in bamboo control towers, waving straw paddles. They built life-sized replicas of airplanes out of sticks and leaves.

They had the form of an airport. They were missing the substance. The planes did not come.

Cargo Cult Science

There is a lot of science that works exactly the same way. The form is perfect. The papers are published. The experiments are described. The statistics are reported. Everything looks like science.

But the planes do not land.

The results do not replicate. The predictions fail. The theory does not advance. The knowledge does not accumulate. The whole edifice looks like science from the outside and produces nothing from the inside.

How does this happen? The same way the cargo cult happens: by imitating the visible parts and skipping the invisible ones.

What Gets Skipped

The visible parts of science: hypotheses, experiments, data, papers, citations, grants, degrees, conferences, peer review.

The invisible part: the relentless honesty about what might have gone wrong.

When a scientist publishes a result, the paper lists the methods, the data, and the conclusions. What it often does not list are:

• The experiments that failed before this one worked
• The alternate explanations that were not ruled out
• The specific conditions under which the result does NOT hold
• The ways the experimenter might have unconsciously biased the result
• The reasons you should be skeptical of this particular finding

A good scientist includes all of this. A cargo cult scientist includes none of it. The paper looks the same. The science is completely different.

The Test

How do you tell the difference? I proposed a simple test in the Caltech speech: does the scientist report results that argue against their own hypothesis?

If yes: real science. The scientist is trying to disprove themselves, which is what science actually is. You do not try to prove your theory. You try to break it. If it survives your best attempt to break it, it might be true.

If no: cargo cult science. The scientist is only reporting results that support their conclusion. Every experiment is designed to confirm, never to refute. The hits are published, the misses are buried. The paper looks impressive and the knowledge is empty.

Why It Matters More Now Than in 1974

In 1974, cargo cult science was a problem inside the scientific community. Bad papers got published. Unreplicable results spread through the literature. But the damage was contained because science moved slowly and the audience was small.

In 2026, the entire information environment is cargo cult. Not just science. Everything. Generated text that looks like journalism. Generated images that look like photographs. Generated arguments that look like reasoning. The form is perfect. The substance is absent.

The epistemic fold point we wrote about (see When Noise Becomes Free) is cargo cult epistemology at civilization scale. The generation of plausible-looking information at near-zero cost, without the substance that would make it true. A billion bamboo control towers, each one indistinguishable from a real airport, none of them landing any planes.

How to Stop

The cure for cargo cult science is the same as the cure for cargo cult epistemology: the extra effort to check whether you are fooling yourself.

Report the negative results. Describe the conditions under which your theory fails. Tell the audience what would disprove you. Design the experiment to break your hypothesis, not confirm it. And when the experiment breaks the hypothesis, publish the breaking, not the hypothesis.

That is the part that is hard. That is the part that cargo cults skip. That is the part that determines whether the planes land.

The first principle is that you must not fool yourself. And you are the easiest person to fool. I said it in 1974. It is more true now than it was then. The tools for fooling yourself have gotten better. The discipline for not fooling yourself has not.

The planes land when you stop building bamboo towers and start building real airports. The real airport is not prettier than the fake one. It is just honest about what it takes to make a plane land.