Jul 2026
PhD Researcher, Computer Science
- Designed and ran a 746-participant mixed-methods study showing human accuracy at spotting AI images sits at chance (47–55.5%), with measurable impostor and automation bias.
- Co-built WILD (20,000 images, 20 generators) and SynthForensics (6,815 videos, 5 text-to-video models), benchmarks where state-of-the-art detectors lose up to 29 AUC points.
- Led a scoping review of 49 studies on wearables, apps and ML for smoking detection and cessation (IEEE JBHI).
- SAMOTHRACE project: AI for wearables and cognitive bias in forensics.