The actor holds two or more things — versions of one item, instances of a kind, subsets of a collection — and the question is the difference. Most tools answer by rendering each thing fully and leaving the comparison to the eye: two windows, two printouts, back and forth, the distinction assembled in the actor’s head and lost when attention moves.
Comparing gives the difference its own representation. Shared features recede, distinctions stay salient, and the magnitude of each discrepancy is readable — the comparison is read, not performed.
Three ways to show a difference
The visual-comparison literature names three families, combinable:
- Juxtaposition — the things side by side, alignment doing the work. The cheapest form and the weakest: the actor still carries the difference between glances.
- Superposition — the things overlaid in one space, difference read as displacement. Works where the compared things share a coordinate system.
- Explicit encoding — the difference computed and drawn as its own object: a diff, a delta column, a changed-region highlight. The strongest form and the most editorial, since something decided what counts as a change.
Consequences
- the difference is a representation of its own — shared features recede, distinctions stay salient — so the comparison is read rather than performed
- what is compared is an editorial claim: the bases of comparison someone chose decide which differences can be seen at all, and a basis left out is a difference that cannot surface
- versions of one thing and instances of a kind compare differently — a version diff can assume shared structure and show only the delta, while instances need their common attributes aligned before any difference reads
Resources & references
- Gleicher, Albers, Walker, Berman, Hansen & Roberts (2011) Visual comparison for information visualization — the juxtaposition / superposition / explicit-encoding taxonomy
- Burghardt (2009) Working Through Screens, F5 “Comparative representations” — comparison as manual juxtaposition in unsupported practice, and the case for purpose-built comparative displays
- Supporting sensemaking of large language model outputs at scale (CHI 2024) — computed similarities and differences across many generated responses as the thing that made the task tractable
Appears in
Making sense at scale
- Choosing how to show the collection, step 2 — Show the difference directly
Related patterns
Complements
- Selection — the items compared are first designated; comparing consumes a selection of two or more
- Reconciliation — showing the colliding versions is comparing's version case put to work