Distance

Version

v3 · 2026-09-28 · History

Prime #

1593

Domain group

Formal Sciences

Origin domain

Mathematics

Subdomains

Metric Spaces, Geometry → Mathematics

Also from

Physics

Core Idea

Distance abstracts the question of separation away from any one substrate. The relata may be physical points, graph vertices, strings, probability distributions, social actors, or conceptual states; a rule and comparison frame turn the pair into a magnitude or ordered category.

Different distances answer different questions. Straight-line, geodesic, travel, edit, statistical, and network distances can disagree without contradiction. Nonnegativity is common, while symmetry, identity of indiscernibles, and triangle inequality must be declared rather than smuggled in.

How would you explain it like I'm…

How Far Apart?

How far is the park from your house? A bird flying straight there might go a short way, but you have to walk around the streets, so your trip is longer. Both answers are right; they just measure "how far" in different ways.

Many Ways to Measure Far

Distance is a way of saying how far apart two things are. The things don't have to be places: two words can be far apart if you need lots of letter changes to turn one into the other, and two friends can be close or far in a friendship network. To get a distance, you need a rule for measuring. Different rules can give different answers without anyone being wrong, like the straight-line distance to school versus the walking distance. So you always need to ask: distance measured how?

Separation Measured by a Rule

Distance is a general way of turning a pair of things into a measure of how separated they are, using a chosen rule and frame of comparison. The things can be points in space, stops in a network, strings of text, probability distributions, or even people or ideas. Different distance rules answer different questions: straight-line distance, path distance along a road, edit distance between words, or number of steps in a social network can disagree without any contradiction. Most distances are never negative. Other properties, like being the same in both directions or obeying the triangle inequality (a detour is never shorter than going direct), hold for some distances but not others, so you have to check them rather than assume them.

 

Distance abstracts separation away from any particular substrate: given two relata, whether physical points, graph vertices, strings, probability distributions, social actors, or conceptual states, a rule within a comparison frame maps the pair to a magnitude or an ordered category. Different distance functions answer different questions: Euclidean, geodesic, travel-cost, edit (e.g. Levenshtein), statistical (e.g. total variation or Kullback-Leibler divergence), and network shortest-path distances can disagree on the same pair without contradiction. Nonnegativity is almost universal, but the other metric axioms are not guaranteed. Symmetry fails for one-way street travel or for KL divergence; identity of indiscernibles fails for pseudometrics, where distinct items can have zero distance; and the triangle inequality fails for many dissimilarity measures. A mathematical metric is the special case in which all of these hold, so each property must be declared for the distance at hand rather than silently assumed.

Broad Use

• Geometry. Measures straight or geodesic separation of points.
• Networks. Counts or weights paths between vertices.
• Computation. Measures edit operations or representation differences.
• Statistics. Compares distributions under defined divergences or metrics.
• Social inquiry. Operationalizes relational or perceived separation.
• Control and planning. Evaluates displacement between states and goals.

Clarity

Always name the pair, representation, distance rule, admissible paths, normalization, scale, units, and formal axioms. Report estimation and model uncertainty separately from the resulting value. Inclusion test: Require two typed relata and an explicit rule that assigns their separation within a declared frame, including units or ordering and any metric axioms claimed. Exclusion test: Exclude one-place size measurements, uncalibrated feelings of difference, proximity labels with no comparison rule, and similarity scores mislabeled without specifying their conversion. Nearest boundary: Similarity increases with likeness and often requires a transformation to become distance; the relation can be monotone without satisfying metric axioms. Exit condition: The identity collapses when the pair, frame, or separation rule is removed, leaving only an unsupported assertion that things are far apart. Common misclassifications: It is not a one-place measure of size. It is not synonymous with difference or similarity. It is not necessarily physical length. It is not necessarily a mathematical metric. Nearest named distinctions: Displacement: A directed change vector can contain orientation that scalar distance omits. Similarity: Usually increases as items resemble one another; its conversion to distance is conventional. Metric: A distance satisfying specified identity, symmetry, and triangle axioms. Length: Measures extent or path size and becomes distance only in a two-relata comparison. Divergence: May be asymmetric or violate triangle inequality while still measuring distributional difference.

Manages Complexity

Distance reduces a potentially complicated relationship to an interpretable separation while retaining the comparison rule as part of its meaning. It supports ordering, clustering, neighborhoods, optimization, and thresholds, but hides direction and causal explanation unless separately represented.

Abstract Reasoning

1. Type the two relata and map them into a common space.
2. Choose a separation rule suited to the intended question.
3. Declare paths, weights, units, normalization, and axioms.
4. Compute or estimate the pairwise value.
5. Test sensitivity to representation and rule choices.
6. Use comparisons only where the scale and uncertainty support them.

Knowledge Transfer

Because the identity is role-based, distance transfers across physical, mathematical, computational, and social carriers when a pair and separation rule remain explicit. Numerical resemblance alone does not transfer geometry, units, or metric theorems.

Example

Euclidean distance compares two coordinate points by the length of their straight connecting segment. Mapped roles: carrier: points; rule: Euclidean norm.

Relationships to Other Abstractions

Current abstraction Distance Prime

Parents (1) — more general patterns this builds on

• Distance is a kind of Comparison Prime

  Distance is the Comparison species that places two relata in a shared space and reads off a nonnegative separation under a declared rule.

Children (3) — more specific cases that build on this

• Kendall tau distance Domain-specific is a kind of Distance

  Kendall tau distance is a distance function counting pairwise ordering disagreements.

• Spatial Observation Baseline Domain-specific is part of Distance

  The known distance between two observation positions is an identity-bearing constituent of a spatial observation baseline.

• Wind Fetch Domain-specific presupposes Distance

  Directional and effective fetch require measured open-water path lengths.

Hierarchy path (1) — routes to 1 parentless root

• Distance → Comparison → Self Checking

Not to Be Confused With

• Displacement. A directed change vector can contain orientation that scalar distance omits.
• Similarity. Usually increases as items resemble one another; its conversion to distance is conventional.
• Metric. A distance satisfying specified identity, symmetry, and triangle axioms.
• Length. Measures extent or path size and becomes distance only in a two-relata comparison.
• Divergence. May be asymmetric or violate triangle inequality while still measuring distributional difference.