Geopolitical Forecasting and Actionable Intelligence

Abstract

In geopolitics, even rough probabilistic forecasts are difficult to make. Scenario design, table-top simulations and structured discussions are heuristic activities that focus the attention of analysts and increase policymaker satisfaction with intelligence products, but their usefulness is not based on confidence that forecasts are accurate. Traumatised by 9/11 and encouraged by advances in social science, computerised data collection and artificial intelligence, the US government has greatly increased investments in reliable geopolitical forecasting. Despite progress by using computerised data processing, sophisticated statistical methods and machine learning, the results have so far been of limited practical assistance. Forecasters must not only provide credible probability judgements but also show the causal pathways underlying them to enable policymakers to mitigate harms and exploit opportunities. Building that capacity means enhancing the role of social scientists in the design and testing of computer simulation and other forecasting techniques.