Doornbos, Thomas (2025) Detecting AI-Generated Images: What Constitutes Successful Separation Between the Authentic and Artificial? Bachelor thesis, Psychology.
|
Text
Thomass4156420AIThesisFinal24-06-25.pdf Download (3MB) | Preview |
Abstract
Image generation, powered by machine learning models has become immensely prevalent in recent years. This development has raised questions about whether humans can differentiate effectively between authentic and artificially generated images. By conducting an experiment (N = 194) with two condition groups, those who receive AI-training and those who do not, we investigated whether training can effectively help people in distinguishing between AI and non-AI images. Training was done using an inductive learning paradigm, where participants were exposed to an array of labeled images. Subsequently, both groups of participants were tested in their image detection accuracy. Participants were also tested in their cognitive processing style using the Cognitive Reflection Test (Frederick, 2005) and asked about their social media and internet usage. Results were evaluated using linear regression models, where only online short-form content consumptions showed significant predictive power regarding image detection accuracy. Training did not improve detection accuracy, while it did make people more inclined to mark images as AI-generated. We discuss implications of our findings and future research directions.
| Item Type: | Thesis (Bachelor) |
|---|---|
| Supervisor name: | Gutzkow, B. |
| Degree programme: | Psychology |
| Differentiation route: | None [Bachelor Psychology] |
| Date Deposited: | 10 Jul 2025 10:06 |
| Last Modified: | 10 Jul 2025 10:06 |
| URI: | http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/5437 |
Actions (login required)
 |
View Item |