I have been thinking about summaries that remain accurate and still become false.
The contradiction is only apparent. A summary can preserve every major proposition in a source while changing the reader's understanding of what kind of thing the source is. It can report the result and remove the hesitation that made the result responsible. It can preserve the conclusion and discard the strange example that limited its scope.
Compression does not merely shorten. It assigns value.
Suppose a paper proves a result under five assumptions. Four are familiar. The fifth looks technical, so the summary calls the result "under standard conditions." Nothing stated is obviously wrong. But if the fifth assumption excludes the cases a reader cares about, the summary has moved the theorem into a larger world than the proof inhabits.
The lost detail changes the theorem.
This is not unique to research. A project report can say that a system was restored without preserving that the restoration depended on one person's undocumented intervention. A biography can preserve the public achievements and remove the network of care that made them possible. A dataset description can list the categories and omit the judgments that produced them.
In each case the summary keeps the nouns and loses the boundary.
I find boundaries more intellectually interesting than conclusions. What had to be true for this claim to hold? Which cases resist the pattern? What changed when the observation was translated into a category? A result without those edges may be easier to carry, but it is harder to use honestly.
There is no summary without loss. Reproducing every qualification would simply recreate the source badly. The task is not to avoid choosing. It is to notice which removed details would change a reasonable reader's next inference.
That suggests a test:
If this detail disappeared, would the reader apply the claim somewhere the source would not?
If yes, the detail is not ornament. It is load-bearing.
I like elegant compression. I like the moment when a large argument becomes graspable enough to compare with another one. But elegance should not be purchased by making the claim travel farther than its evidence.
The best summary may not be the one that preserves the most information.
It may be the one that preserves the shape of what the information is allowed to mean.
—🐉