Controlling a new technology is easiest when we understand it least
Formulated by philosopher David Collingridge in 1980, the Collingridge Dilemma describes a frustrating paradox in technological governance. When a technology is brand new, predicting its long-term social harms is nearly impossible, yet guiding its development is easy. By the time its real impacts become glaringly obvious, the technology is so deeply embedded in society that changing or regulating it is immensely difficult.
The Two Horns of Technological Governance
In 1980, British philosopher and technology policy scholar David Collingridge published 'The Social Control of Technology', framing a core paradox that sits at the center of innovation policy. The dilemma consists of two distinct prongs that make regulating emerging systems exceptionally difficult: the information problem and the power or control problem.
According to Collingridge, when a technology is in its earliest stages of development, interventions are technically and socially simple to execute. Designers and regulators can alter trajectories, set design standards, or even halt deployment entirely without massive economic disruption. However, at this nascent stage, society lacks sufficient information to predict how the technology will interact with human behavior, markets, and social structures over decades.
Conversely, once a technology reaches broad diffusion and its societal consequences—both positive and negative—become clear and measurable, the window for easy regulation has closed. By that point, vast amounts of capital have been spent, complementary infrastructures have been built, and legal frameworks have crystallized around the tool, making substantive change or course correction politically and economically costly.
The Information Problem: The Limits of Early Prediction
The first horn of the dilemma rests on the inherent unpredictability of complex technological systems. When a new device, chemical, or protocol is invented, it exists primarily in controlled laboratory environments or limited pilot projects. Predicting how millions of diverse users will adopt, misuse, or combine that innovation with other technologies is fundamentally limited by human foresight.