Zijlstra, Machteld (2026) Comparing AI Detection Between Poetry and Paintings Among Varying Levels of Human-AI Co-Creation. Bachelor thesis, Psychology.
|
Text
Bachelor-Thesis-M.G.-Zijlstra-S5554128.pdf Download (1MB) | Preview |
Abstract
This study examined whether AI detection accuracy differs between poetry and paintings, and how detection accuracy varies across three different levels of human-AI co-creation. 133 participants in a Non-Transparent condition repeatedly evaluated their perceived degree of AI for either poems or paintings, on three AI Levels. The AI Levels consisted of: Level 0 (human-made art), Level 1 (AI-generated art following a single prompt), and Level 3 (AIgenerated art following continuous prompting until a satisfactory result was reached). Findings indicated a medium effect between poetry and paintings, where AI was more accurately detected in paintings than in poetry, across all three levels. Furthermore, for poetry and paintings separately, detection accuracy on Level 1 and 3 were compared, to explore whether increasing human input (Level 3) makes AI detection more complicated. No significant result was found for poetry, but for paintings a small effect was found. These findings suggest that visual AI-generated art is easier to recognize compared to AI poetry, and that including an extensive co-creation level has a negative influence on detection accuracy of visual art. Further implications about the use of co-creative AI in poetry and visual art are discussed.
| Item Type: | Thesis (Bachelor) |
|---|---|
| Supervisor name: | Cox, R.F.A. |
| Degree programme: | Psychology |
| Differentiation route: | None [Bachelor Psychology] |
| Date Deposited: | 26 Feb 2026 07:04 |
| Last Modified: | 26 Feb 2026 07:04 |
| URI: | http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/6274 |
Actions (login required)
 |
View Item |