Tensions in Practice: Equal coordinate treatment in tension with task-weighted similarity¶
Two normalized attributes of catalog options
A requested profile is (0, 0). Option A is (3, 0); option B is (0, 2). With distance |Δx| + |Δy|, B is nearer. With distance |Δx| + 3|Δy|, A is nearer. Both are valid weighted Manhattan metrics: the positive weights preserve the distance axioms. The disagreement comes from how heavily a gap in y counts, not from changed options.
Treat dimensions symmetrically
Use equal weights after the declared normalization.
Protect a sensitive dimension
Charge more for a gap in the task-critical y attribute.
Why these aims pull against each other
A weighted comparison makes priorities explicit but embeds them in every nearest-match decision. Equal weights also encode a priority choice.
Choose an arrangement to see what changes and what remains difficult.
Fixed rows show the two gap vectors. The distance column recalculates the same gaps under the stated rule, and text marks the selected nearest option.
What this choice protects
What it costs
When it fits
Compare the arrangements
Use equal weights
Add the absolute gaps in the two normalized coordinates.
| Gaps x, y | Distance | Nearest? | |
|---|---|---|---|
| Option A | 3, 0 | 3 | No |
| Option B | 0, 2 | 2 | Yes |
- What it protects
- Neither normalized dimension gets extra weight.
- What it costs
- A two-unit y gap beats a three-unit x gap even if y matters more for a particular use.
- When it fits
- Fits tasks for which the two normalized gaps deserve equal treatment.
Illustration note: A distance of 2 versus 3 is correct under this declared metric, not a claim of objective overall superiority.
Weight the sensitive gap
Multiply the y gap by three before adding the x gap.
| Gaps x, y | Distance | Nearest? | |
|---|---|---|---|
| Option A | 3, 0 | 3 | Yes |
| Option B | 0, 2 | 6 | No |
- What it protects
- The chosen option matches y exactly.
- What it costs
- It accepts a larger x gap and requires a defensible weight and normalization.
- When it fits
- Fits a task where a unit of y mismatch is judged three times as costly as x mismatch.
Illustration note: The metric changes; the points do not. This is neither a causal estimate nor a learned preference.
What this illustration does—and does not—establish
The source supplies the structural tension; the invented example makes one relation inspectable. Costs and conditions are part of each arrangement, not exceptions to a universal recommendation.
- Coordinates are already normalized by an explicit toy convention. Real mixed units need a justified scaling rule.
- Both weights are strictly positive. A zero weight would generally produce a pseudometric instead.
- This example changes which candidate is closest, not whether similarity is transitive.
Source entries
Metric
The canonical tension motivates this comparison. The setting, finite values and arrangements are declared editorial illustrations, not measured findings.
Choice of Metric versus Objective Similarity (the modelling choice hides)
"How similar are these?" has no answer until a distance function is chosen, and the choice determines the verdict — Euclidean and cosine disagree about which songs are alike, edit and Hamming disagree about which genomes are kin. The tension is that the metric is a substantive modelling decision disguised as a neutral fact.
The source operation
A metric is a rule that assigns a non-negative distance to every pair of objects in a set, subject to three constraints. The distance is zero exactly when the objects coincide (identity of indiscernibles); it is symmetric, so the distance from a to b equals the distance from b to a; and it satisfies the triangle inequality, so the direct distance from a to c never exceeds the sum of the distances from a to b and from b to c.