Skip to content
How to ThinkIn the Age of AI

Question

What is polyhierarchy?

beginner

Quick Answer

Real knowledge often has items that belong to multiple parent categories. When you force every concept into a single branch of a tree, you destroy information. Lattice structures — where a node can have multiple parents — preserve the multidimensional nature of knowledge. The tree is a special case.

The lattice is the general case.

Example: A physician diagnosing a patient with liver cancer opens SNOMED CT, the clinical terminology system used in healthcare worldwide. She navigates to "Neoplasm of liver." In SNOMED CT, this concept does not sit in one branch. It is simultaneously a subtype of "Disorder of liver" and a subtype of "Neoplasm." She can reach it by navigating down from organ systems or down from disease types. This is not a quirk of SNOMED CT — it is a design principle called polyhierarchy. The concept genuinely belongs in both places because it genuinely is both things: a liver disorder and a neoplasm. Forcing it into one branch would destroy clinical information. A physician searching by organ would miss it if it were filed only under neoplasm type. A oncologist searching by disease category would miss it if it were filed only under liver disorders. The polyhierarchy preserves both access paths because both are true.

This concept is part of Hierarchy and Nesting in the How to Think curriculum.

Learn more in these lessons

  • Hierarchies are not trees they are lattices
  • Multiple valid hierarchies for the same data

    The same set of items can often be organized in several equally valid hierarchical structures. Each hierarchy foregrounds different relationships and obscures others. No single arrangement is canonical — the right hierarchy depends on what you are trying to see, find, or do. Recognizing this multiplicity is a precondition for deliberate knowledge design.