Researchers at MIT have developed a new algorithm designed to confront a fundamental challenge in risk assessment: planning for catastrophes that have never happened. Traditional predictive models rely heavily on historical data, which by definition lacks information on truly unprecedented events, leaving critical systems exposed to black swan scenarios.
The new approach sidesteps that limitation by generating plausible, high-impact scenarios itself, without needing prior examples of similar extremes. Rather than waiting for disaster to supply the data, the algorithm learns to anticipate the conditions that would most severely stress a system, whether a power grid or a global supply chain.
For planners in government and industry, the method offers a way to probe vulnerabilities they might not even know exist. By surfacing the unprecedented scenarios that critical infrastructure and supply chains are least prepared for, the tool aims to help decision-makers stress-test and fortify systems before the truly extreme arrives.