Analytical tool supporting human-machine teaming in intelligence analysis

Ox Intel was founded out of the Hacking 4 Ministry of Defence (H4MoD) program at King’s College London, where our team was tasked by our government sponsor, the Royal Air Force, with tackling a problem familiar to many in the US intelligence community (IC): subjective analysis of information in the intelligence cycle. Subjective analysis involves intuitive decision making, which, unlike more rational decision-making, lacks repeatability and consistency, and presents barriers to incorporation of machine learning. To solve this problem, our team developed Ox, a structured analytical tool to quantify and collect judgments that factor into qualitative source analysis, designed for users with no modeling or data science skills and aimed at improving human-led analysis at scale. Our solution is the product of an extensive beneficiary discovery process that solicited feedback from more than 100 intelligence officers, data scientists, and other stakeholders across the US and UK ICs.   Our solution takes what is irreplaceable by machines from human judgment processes, and reinforces it with objective indicators of reliability, building on existing methodologies for multiple-criteria decision analysis in an intuitive, interactive interface. The resulting output optimizes human-led analysis and helps train machine learning algorithms to improve intelligence-led decision-making at scale.