Can machine learning turn crowdsourced footage of airstrikes into legal evidence on human rights abuses in Yemen?
This experiment will use machine learning to classify crowdsourced footage of airstrike images. The goal is to build a database for the preservation, organisation, and secure transfer and use of evidence for war crimes investigations. The experiment will test whether the classification and filtering of digital evidence will help official investigators to place more reliance on it, and whether it enables courts to admit it for consideration. This will shed light on whether a collective intelligence approach to processing this kind of evidence increases the chances of successful war crimes prosecutions.
Technologies like smartphones, social media, and participatory mapping tools enable witnesses to human rights violations to capture and share footage of war crimes in real time. These images create a wealth of potential evidence for legal investigations. But legal practitioners have been hesitant to use open-source and crowd-sourced evidence in court cases due to the sheer volume of the material, the effort related to verifying the content, and the vulnerability of civilian witnesses on the ground. Developing a machine learning tool to locate specific weapons, landmarks, or other characteristics that can identify and verify this digital evidence could help fully realise the potential this material offers.
The results of this experiment will help us understand how machine learning can be combined with human intelligence to lessen practitioners' burden in verifying content and identifying evidence that can be used in lawsuits. The findings will have wider applicability to contexts where open source and digital evidence provides an evidence base, such as investigative journalism, crisis response, and academia.
Follow the project lead on Twitter: @ProfYvo