"Perceived ownership steadily declines with each AI rewrite, dropping fastest for texting and slowest for code. Turns out, the ownership ratings fell off a cliff around the third or fourth rewrite for texting and essay writing, but held surprisingly steady in code. A spline model (R, lme4, three-knot natural spline) showed the steepest drop early on: texting collapsed fastest, essays followed, code barely flinched."
"By the numbers: Texting: sharpest drop of -1.4 points around iteration 3. Essay: a slower, smaller decline (-1.25 points near iteration 5). Code: almost no falloff (-0.3 points around iteration 4). When asked when their text felt "more AI than me" (below a 4 on the 1-7 scale), people hit that midpoint at iteration 3 for texting, iteration 7 for essays, and never for code."
Participants worked across three modes: casual messages, short essay paragraphs, and code snippets. Each person wrote original content and ran it through an AI writer ten consecutive times, rating ownership after each iteration on a 1–7 scale. Responses were anonymized and analyzed with a mixed-effects spline model in R to trace ownership decay. Perceived ownership declined steadily with repeated AI rewrites, dropping fastest for casual messages, slower for essays, and minimally for code. Ownership ratings fell sharply around iterations three to four for messages and essays; code remained largely stable. The midpoint "more AI than me" occurred earlier for messages than essays and never for code.
#ai-assisted-rewriting #authorship-perception #mode-comparison-messagesessayscode #mixed-effects-modeling
Read at Medium
Unable to calculate read time
Collection
[
|
...
]