How can technology strengthen fact-checking?
As the technology to create realistic fake images, video and audio becomes more sophisticated, fact-checkers and journalists need similarly advanced tools to counter this threat. Assembler is an experimental platform from Jigsaw and Google Research that hopes to make it easier to judge manipulated media and help prevent the spread of disinformation.
Fact-checking images can be a time-consuming process. Image manipulation detectors — technology that can help speed up the process by identifying if and where an image has been manipulated — are not always accessible to fact-checkers and journalists. Assembler aims to make it easier and faster for journalists and fact-checkers to get the information they need to judge an image’s authenticity by bringing together multiple image manipulation detectors from academics into one tool. To specifically address the threat of deepfakes, we developed a synthetic media detector for StyleGAN-type image deepfakes.
Individual detectors are often unable to accurately detect different types of image manipulation. For example, detectors designed to identify images manipulated through copy and paste are usually unable to detect manipulations to image brightness — and vice versa. To address this, we built an experimental detector, the “ensemble model,” that is trained using signals from multiple detectors. Because the ensemble model can identify multiple image manipulation types, the results are, on average, more accurate than any individual detector.
Fact-checkers and journalists need to understand if an image is manipulated and how much trust to put on that conclusion. Assembler aims to provide clear explanations on individual detector performance and analysis so fact-checkers and journalists can feel empowered communicating their evaluation of an image’s authenticity.