The march of “anticipatory governance”

Originally posted on The Horizons Tracker.

Governments around the world are turning to generative AI to help them look ahead. By predicting possible futures, these systems allow policymakers to act sooner and—at least in theory—more effectively. New research¹ from Trinity College Dublin examines how this form of “anticipatory governance” is gaining ground, especially in China, but also in parts of Europe and, increasingly, Ireland.

At the heart of this trend are “city brains”—large-scale AI systems embedded in digital urban infrastructure. These AIs, often powered by large language models, process vast amounts of data to generate visions of a city’s future. Policymakers then use these projections to craft new regulations and urban plans. China alone has over 500 cities developing such systems.

One example is in Beijing’s Haidian district, home to around three million people. There, a city brain collects real-time data from over 14,000 CCTV cameras and 20,000 environmental sensors. It focuses on managing traffic, public safety, and environmental risks. If a flood is forecast, for example, new infrastructure may be built to shore up riverbanks. If a crime is predicted, police may be dispatched to intervene before it happens.

Major risks

But this kind of predictive governance comes with major concerns. The first is transparency. Large language models often operate in ways even their designers struggle to explain. If a system suggests action, but no one can account for how or why, public trust—and accountability—may suffer. Second, as these systems grow in power, they risk sidelining citizens and human decision-makers. In effect, governance may become less democratic and more “posthuman.”

Accuracy is another issue. The Haidian system’s predictions range from 60% to 90% accurate. That might seem impressive, but it leaves a wide margin for error—particularly in sensitive areas like predictive policing. When an AI wrongly forecasts a crime, real people can be wrongly targeted. There’s no easy fix: the complexity of these models often makes it hard to challenge or even understand their decisions.

As governments become more reliant on such tools, the line between helpful foresight and opaque control becomes harder to draw. Generative AI may offer a smarter way to run cities—but only if its judgments remain subject to human review and democratic oversight.

Article source: The March Of “Anticipatory Governance”.

Header image: The dashboard of an existing city brain system in China. Source: Dr Ying Xu.

Reference: