Human-AI Ensembles Improve Deepfake Detection in Low-to-Medium Quality Videos

TL;DR

This study shows that combining human judgment with AI detectors significantly improves deepfake detection accuracy, especially in low-to-medium quality videos where AI alone struggles.

Contribution

It demonstrates the effectiveness of human-AI ensembles in deepfake detection, highlighting the importance of collaboration over sole reliance on AI.

Findings

Humans outperform AI detectors on both datasets.

AI accuracy drops significantly on low-quality videos.

Hybrid human-AI systems reduce high-confidence errors.

Abstract

Deepfake detection is widely framed as a machine learning problem, yet how humans and AI detectors compare under realistic conditions remains poorly understood. We evaluate 200 human participants and 95 state-of-the-art AI detectors across two datasets: DF40, a standard benchmark, and CharadesDF, a novel dataset of videos of everyday activities. CharadesDF was recorded using mobile phones leading to low/moderate quality videos compared to the more professionally captured DF40. Humans outperform AI detectors on both datasets, with the gap widening in the case of CharadesDF where AI accuracy collapses to near chance (0.537) while humans maintain robust performance (0.784). Human and AI errors are complementary: humans miss high-quality deepfakes while AI detectors flag authentic videos as fake, and hybrid human-AI ensembles reduce high-confidence errors. These findings suggest that effective real-world deepfake detection, especially in non-professionally produced videos, requires human-AI collaboration rather than AI algorithms alone.