Keane, Megan (2026) Fake It Till You Make It: Confidence, Emotions and Distinguishing AI from Real Faces. Bachelor thesis, Psychology.
![[img]](https://gmwpublic.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
ThesisM.Keanes469751030.1.pdf Restricted to Repository staff only Download (351kB) |
Abstract
Artificial Intelligence has been shown to create highly realistic human faces, yet many observers often struggle to tell fake from real images apart, posing risks for misinformation and deception. This study examines whether brief inductive training can improve AI-face detection and whether detection accuracy varies by emotional expressions of the faces and how it relates to confidence calibration. In an online mixed design, participants were assigned to a training condition (interleaved, labelled examples) or a control condition and classified faces as AI-generated or real across angry, happy and fearful expressions. After applying time-based exclusions, the final sample comprised of 273 participants (125 training, 148 control). Inductive training substantially improved detection accuracy (training: M=.87, SD=.09; control: M=.74, SD=.15, d=1.07). Accuracy differed by emotion, with fearful faces classified most accurately and angry faces least accurately. However, training benefits were consistent across emotions (no emotion × condition interaction). Post-test confidence was positively associated with accuracy, providing no evidence for an overconfidence effect. Overall, the findings suggest that perceptual judgements of face authenticity are adaptable and can be improved through short training interventions, although generalisability may be limited. Keywords: artificial intelligence, face perception, inductive learning, confidence calibration, emotion
| Item Type: | Thesis (Bachelor) |
|---|---|
| Supervisor name: | Gutzkow, B. |
| Degree programme: | Psychology |
| Differentiation route: | None [Bachelor Psychology] |
| Date Deposited: | 25 Feb 2026 07:03 |
| Last Modified: | 25 Feb 2026 07:03 |
| URI: | http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/6266 |
Actions (login required)
 |
View Item |