No, I can only demonstrate that it approximates the results shown so far.
I never said it was an incontrovertible fact.
No. My claim is that the expected number of connections due to chance increases with more selections. Expected values are a well established aspect of probability theory.
It's restating the null hypothesis, which is a critical aspect of any scientific evaluation or testing of any claim.
Good think I didn't say my probability estimations are incontrovertible. What is (and what I said is) incontrovertible is that the number of opportunities for chance connections between selections, and hence the expected number of such connections, increases with more selections.
I'm using "expected" in the sense of probability theory. Potatoes vary in weight and we might not know how the exact weights of potatoes are distributed, but we can still rationally expect more potatoes to weigh more than fewer potatoes.
Expected Value
The null hypothesis is that there was no "it" done because what was observed happened by chance alone.
It is an untested null hypothesis, until someone tests it.
You claim below that analysis or tests can refute the null hypothesis, so the null hypothesis is falsifiable.
My argument relies on no assumptions besides the axioms of combinatoric math, and the near-certain likelihood that the entries in a large curated library selected based on a consistent cloud of topics of interest will sometimes share common references. More selections mean more distinct pairs of selections and therefore more chance (if there's a nonzero chance to begin with) that pairs sharing a common reference will occur. Since what you call "structure" consists of such common references, it follows that lengthier interactions with the system that call up a greater number of selections will tend on average to exhibit more of said structure.
The null hypothesis here is not "no structure" (as you have operationally defined "structure" as occurrences of identifiable common references between pairs of selections), it's that the structure observed is consistent with chance expectations.
As for further analysis, look, I'm better at this kind of lateral thinking poetry criticism than anyone I know. My English teachers used to gush about things like how I found a hidden reference to the four Classical elements in Poe's "
Sonnet—To Science." Here's a trick: instead of looking for segment A references theme 1 and segment B also references theme 1 (easy to spot but a bit rare), you can look for segment A references theme 1 and segment B references theme 2, but theme 1 and theme 2 both relate to something else. Like, segment A references obsession and segment B references industrial labor, but
Moby Dick references both of those things, so, bam!, a "hidden" connection between A and B, plus now you have a big white whale of a novel with dozens of motifs of its own to make more "connections" with! Those are harder to spot but the additional cross-connection increases the world of possibilities so astronomically that with enough effort you could probably find hundreds of items of "structure" from a score of selections, instead of a dozen. That's the kind of thinking modern art critics use to find all kinds of significant meaning in a new installation, until they find it out's actually the janitor's cart. Even an AI can do it, as you've observed.
