Trust Cascades

Why Airplanes Are Reliable and Online Information Is Not

As you board the plane, you realize that five million things have to go right for you to reach your destination safely. That sounds almost impossibly daunting. Yet there’s a 99.99999% chance that your flight will be a safe one. How is that possible?

An Airbus 350 requires five million parts in its assembly. And fortunately for everyone who flies, the level of quality control on those parts is exceptionally high. The result is that the manufacturing process is incredibly functional and flying is extremely safe.

Modern manufacturing of products like airplanes, smartphones, and large appliances works astoundingly well despite their complexity. Considering how many thousands — or even millions — of parts are assembled, this is an amazing aspect of modern life.

Trust Cascades

Trust cascades refer to the layered dependencies of trust, where each link in the chain relies on the integrity and reliability of the preceding one. This concept is crucial in understanding how complex systems, like manufacturing or information dissemination, maintain functionality and credibility.

Manufacturing is built on top of cascading trust. If my company is producing a part for Airbus, made up of ten sub-components, I'm going to go to great lengths to ensure:

1. I can trust the manufacturing process for each of those ten sub-components. If one of the companies producing them is doing shoddy work, it has the potential to destroy Airbus' trust and my company's reputation.

2. Airbus continues to trust my company. That means that we need to manufacture an excellent product for Airbus.

This trust cascade starts with the tiniest part of an engine. Rolls Royce -- who manufactures the A350 engine -- relies on tests, audits, and other quality control measures to ensure the manufacturer of the tiny part is delivering a working product. Airbus has to trust Rolls Royce to manufacture the engine properly. United Airlines has to trust Airbus, and I as a passenger have to trust United Airlines.

That's an effective trust cascade. Not all industries have developed such rigorous trust systems, and other trust cascades are considerably less robust.

The Modern Media Environment

The modern media environment provides a striking contrast. Here are some examples of trust cascades in media:

1. An article from the New York Times posits that a narrative is true, interviewing a few people who vouch for it (sometimes people considered experts, sometimes typical people who are an example of a trend). The Times' editor trusts that the journalist has in fact interviewed the subject(s) of the story, the Times' owner places trust in the editor to make a decision about publishing the story, and I trust that the New York Times is a reliable filter of accurate information about the world.

2. Someone I know (or follow) on social media posts a story or a video from an unknown source. I trust that that person is a reliable filter of accurate information about the world.

3. Someone I don't know posts a story or a video from an unknown source. I guess I could trust that that unknown video from an unknown source is trustworthy? Clearly a lot of people wind up trusting things they see in random videos online.

There are two things that differentiate this trust cascade from the Airbus 350 trust cascade.

First, the percentages are going to be a lot smaller. Depending on the friend, the video source, the journalist, and the editor, the likelihood of being true might be 99% if a journalist is being careful (still much lower than 99.99999%) or well under 50% if someone is peddling crap on TikTok.

Second, the trust cascade for online information is far less fine-grained. An airplane has a supply chain and an accountability process to ensure the quality of millions of small parts; in modern companies, millions of checks are made both by humans and by software.

Trust Cascades for Online Information

Trust cascades for online information are severely lacking, because the costs are relatively high and the benefits are relatively low:

• It’s time consuming and annoying for a journalist or a YouTuber or a Substacker to “show their work” and document/link back to the source of their claims. Investigative journalists often have to spend hours, if not days, gathering evidence, verifying facts, and cross-referencing sources, only for this work to be summarized in a few sentences. Similarly, YouTubers who provide commentary on current events may need to dig through numerous articles and interviews to establish credibility, but the payoff is often minimal. I can relate — I found it annoying to show my work as a kid, and as an adult, it’s way more interesting to make arguments than to add citations.

• There’s very little financial (or non-financial) upside in creating the kind of trust cascade that exists for airplanes. People want a safe flight across the country, but they want an article or video that will entertain them and maybe reinforce their pre-existing beliefs.

This is a huge problem. Why is it that we can be 99.99999% confident that our flight will arrive safely, but have so little confidence or transparency in the information we encounter online?

Can we apply the same principles to digital information? It won’t be easy, but I have a few ideas that I think could take us there. More on this soon.