Post

Christian Guckelsberger
@creativeendvs.bsky.social

2/ We ran a preregistered experiment (N = 273) in which participants wrote short creative stories either (A) entirely without AI, or with (B) AI using single-prompt vs iterative prompting, crossed with (C) whether they could directly edit the generated text afterwards. We evaluate A+BxC.

Christian Guckelsberger
@creativeendvs.bsky.social

3/ Key finding #1 = an interaction: editing strongly increased ownership after single-prompt generation (Δ = 0.89), but much less after iterative prompting (Δ = 0.27). Iterative prompting thus supported psychological ownership mainly when editing was unavailable.

Christian Guckelsberger
@creativeendvs.bsky.social

4/ This suggests prompting and editing can provide partially substitutable routes to ownership: both let users invest themselves in and exert control over the text, but at different stages of the writing process. More interaction was not simply additive!

0 reposts0 likes

Christian Guckelsberger
@creativeendvs.bsky.social

5/ Key finding #2: every AI-assisted workflow still produced significantly & substantially lower ownership than writing without AI, even when participants could iteratively prompt and edit. With AI writing tools, one-shot delegation can be particularly harmful when ownership matters.

Christian Guckelsberger
@creativeendvs.bsky.social

6/6 Bonus finding: participants who were more intrinsically motivated also felt substantially more ownership over their stories. Exploratorily, the prompting × editing effect was strongest at lower intrinsic motivation and attenuated as intrinsic motivation increased.