Case Study: UCL improves deepfake detection research with Prolific

UCL uses Prolific to benchmark deepfake detection across 2 language experiments

A PhD candidate at UCL needed high-quality human data for research into how well people can detect AI-generated deepfake audio clips. The challenge was finding a platform to recruit engaged and trustworthy participants who would provide detailed, high-quality written responses for two separate experiments targeting fluent English and Chinese speakers.

UCL used Prolific for its high data quality and participant engagement. The platform's comprehensive filtering allowed the researchers to easily target participants by language fluency. The results established a human performance benchmark for detecting speech inconsistencies in deepfakes, providing crucial data that can be used to improve machine learning detection models.

UCL