Skip to content

It Is Not a Tree

Tree structures have the advantage of being easy for humans to understand, but the world rarely contains counterparts that fit neatly into a tree.

Designing on principle

In Inventing on Principle, Bret Victor explains what it means to invent or create on the basis of a principle. Instead of making whatever comes to mind in whatever way, you establish a principle and make things that accord with it, in ways that accord with it.

After years of making tools for creators and thinking about them, Victor says he arrived at the principle that “Creators need an immediate connection to what they’re creating.” He uses this principle to give a consistent account of the various things he has tried.

As an example of someone else’s principle, he introduces Larry Tesler’s “No mode.” Tesler was so committed to interfaces without modes that he even changed his car’s license plate to “NOMODES.”

Do I have a principle like this? Can I say that I have a principle, rather than just making whatever I feel like making? Perhaps “It is not a tree” could be mine.

A city is not a tree

In A City Is Not a Tree (1965), the architect Christopher Alexander argued that naturally formed cities have a semilattice structure, whereas artificial cities designed by humans have a tree structure. He criticized the latter for severing the connections life requires and, as a result, fragmenting people’s lives. Here is a summary of his argument:

Trees are useful for holding complex ideas in the human mind, but a city is not a tree. A city is a vessel for life. A city designed as a tree is like a vessel full of razor blades: whatever you put into it will be cut to pieces. Designing a city as a tree fragments the lives of the people who live there. #

I think this is true of many things besides cities.

Things that are not trees

Information architecture is not a tree

Folders are fundamentally trees. If you try to classify information into folders, fragmentation is hard to avoid. For example, should cognitive neuroscience go under cognitive science or neuroscience? The moment you set classification criteria according to a single hierarchy, information that does not fit neatly into it gets fragmented and scattered.

Tree-shaped designs also shut out opportunities for happy collisions. They prevent discussions from different fields from meeting and producing interesting conversations. The same goes for Slack channels, community forums, and the like that are divided too finely.

Design is not a tree

Design is not a tree either. For example, CSS was deliberately designed to make it easy to cross the boundaries of a tree structure (the DOM tree) through selectors. Yet many people define this as a “problem to solve,” break design into components, and aim to minimize the ways design crosses those boundaries.

This lets designers and developers put the problem away at a “manageable” level without communicating as much as they need to. Instead of solving the problem, they chop it into small pieces and tuck it away in the corners of components.

Software design is not a tree

When software is modularized according to a single primary concern (usually one related to the business domain), other concerns, such as logging and security, are scattered throughout the system. This is called the tyranny of the dominant decomposition.

Aspect-oriented programming mitigates this problem through multidimensional separation of concerns.

Identity is not a tree

In the court cases K. W. Crenshaw analyzed in Demarginalizing the Intersection of Race and Sex, the courts held that Black women could not bring a distinct claim combining racial and sex discrimination. In other words, “discrimination against Black people” or “discrimination against women” could be recognized, but “discrimination against Black women” was held not to constitute a legally valid concept.

This problem stems from a way of thinking that tries to decompose human identity or structures of social oppression into a single-axis tree. Intersectional analysis is an attempt to move beyond this limitation.

Trees are not trees

Even biological trees do not follow a mathematical tree structure. The leaf veins of many species form networks, branches sometimes grow back together through natural grafting, and roots either form networks themselves or are connected by fungal mycelia into networks across individuals and species.

Tree structures are tidy and help humans understand things, but the world rarely contains neat tree structures. As always, we are merely cutting the world to fit our cognitive convenience.

How to respond to things that go against the principle

This principle is not a principle because it is superior to, or more objective than, other principles. It is merely a subjective criterion that applies only to me because I have decided to adopt it.

When I encounter something that goes against this principle, I would like to be someone who first takes time to contemplate it and gather my thoughts, or, depending on the situation, feels interest or curiosity first, rather than anger or superiority. (Be a Slow Opinionist)

See also

By
ak
Created
Updated

Machine-translated from the Korean original.

2026 © ak | markdown | 한국어