Anticipatory Governance

This is a primer on Anticipatory Governance.

Anticipatory Governance is a broad-based societal capacity to steer emerging, knowledge-based technologies while such steering is still possible. Developed by David Guston and colleagues in the mid-2000s, it rests on three capacities: foresight (exploring plausible futures), engagement (bringing publics and stakeholders into the conversation early), and integration (weaving social and ethical reflection into research itself). The ambition is to move governance upstream, into the phase where technological paths remain open, rather than reacting after markets, infrastructures, and interests have locked in.

Table of Contents

• Origins
• What is Anticipatory Governance?
• Situating Anticipatory Governance
• The Three Core Capacities
• Evolution of the Concept
• Anticipatory Governance and Artificial Intelligence
• Democratic Theory and Citizen Participation
• Critical Perspectives
• Open Questions
• Sources

Origins

Anticipatory Governance was born from a policy embarrassment. In 2000, the United States launched the National Nanotechnology Initiative (NNI), a multi-billion-dollar research program built on soaring promises: cleaner energy, targeted medicine, materials stronger than steel. Questions about risks, social consequences, and public consent came later, if at all. The pattern was familiar. Governance arrived after the science, as cleanup rather than as design.¹

To interrupt that pattern, the 21st Century Nanotechnology Research and Development Act of 2003 required attention to the societal dimensions of the field. Out of this mandate came the Center for Nanotechnology in Society at Arizona State University (CNS-ASU), funded by the National Science Foundation in 2005. CNS-ASU was less a conventional research center than a laboratory for new forms of technology steering. Its task was to study the social dimensions of nanotechnology and, at the same time, to build capacities that might let society engage the technology while it was still forming.²

Anticipatory Governance emerged there as the programmatic frame for that work. Guston and colleagues drew a deliberate contrast with the reactive posture of conventional technology policy, which regulates only after a market breakthrough or after the first documented harms. For a field as broad and early as nanotechnology, that posture guarantees lateness. The alternative was to shift governance forward, into research and early innovation, and to distribute the capacity for it across many actors rather than concentrating it in a regulatory agency.¹

The term itself has a shallow but pointed history. Guston notes that “anticipatory governance” barely appears as a term of art before 2001. Its lineage runs back to Alvin Toffler’s idea of anticipatory democracy, introduced in Future Shock (1970). Toffler argued that the antidote to disorientation in the face of accelerating change was not more expert control but a distributed, participatory attention to the consequences of technology: ordinary people helping to imagine and choose among futures.³ Anticipatory Governance inherits that democratic impulse and couples it to the machinery of research funding, universities, and science policy. The Collingridge Dilemma, the observation that technologies are easy to change when little is known about them and hard to change once their effects are clear, supplies the underlying justification. If control is only ever easy while ignorance is greatest, then governance has to learn to act early and under uncertainty, or it will always act too late.

What is Anticipatory Governance?

Guston’s most cited formulation defines Anticipatory Governance as a broad-based capacity extended throughout society that can act on a variety of inputs to manage emerging knowledge-based technologies while such management is still possible.¹ Each part of that sentence carries weight.

A capacity, not an instrument. Anticipatory Governance is not a single policy tool, a law, or an office. It is an ability that a society either has or lacks: the collective competence to see technological change coming and to do something about it. This framing lowers the stakes on prediction and raises them on preparedness.

Extended throughout society. The capacity is meant to be diffuse. Scientists in the lab, funders, regulators, journalists, civil society organizations, and lay publics all hold a piece of it. Governance is not delegated upward to a technocratic elite but built horizontally.

A variety of inputs. Scenarios, expert assessments, public deliberation, ethical reflection, and market signals all feed the system. No single source of knowledge is privileged, least of all forecasting.

Emerging, knowledge-based technologies. The targets are science-intensive fields whose futures are genuinely open: nanotechnology first, then synthetic biology, neurotechnology, geoengineering, and now artificial intelligence.

While such management is still possible. This is the temporal heart of the concept. Anticipatory Governance concentrates on the early phase, before path dependencies, lock-ins, and vested interests have hardened. It is a bet placed against the clock.

The definition is at once descriptive and normative. It says something about how governance could work and insists that this is how it should work. Crucially, Guston separates anticipation from prediction. To anticipate is not to forecast a single future correctly but to build readiness across a range of plausible ones. That distinction does most of the conceptual work, and it is also the concept’s most contested move: critics question whether anticipation can really be separated from prediction at all, an argument taken up in the critical perspectives below.

Situating Anticipatory Governance

Anticipatory Governance rarely travels alone. It overlaps with a cluster of adjacent fields that share its worry about the mismatch between technological and institutional time, but differ in focus, time horizon, and the actors they mobilize. Within this primer it serves as the organizing lens. It is the broadest of the steering ambitions, and more specialized practices slot into it.

Two relationships deserve care. The first is with Responsible Research and Innovation. Anticipatory Governance is the earlier concept; when the RRI framework crystallized around 2013, anticipation became one of its four dimensions, standing alongside reflexivity, inclusion, and responsiveness.⁴ So the two are not nested but interleaved. Anticipatory Governance supplied much of the vocabulary that RRI later systematized, and RRI in turn gave anticipation an institutional home in European research funding.

The second is with Legal Foresight, which reads as a specialized application of Anticipatory Governance aimed squarely at law-making, adjudication, and regulatory design. Where Anticipatory Governance asks how a whole society can steer a technology, Legal Foresight asks how legal systems in particular can prepare for futures that have not yet arrived.

The German-language tradition of Technology Assessment provides a further point of reference, and a critical one. Scholars such as Armin Grunwald have long insisted that assessing a technology means interpreting the present debates around it rather than divining its future, a hermeneutic stance that sits uneasily with any governance model built on anticipation. That tension surfaces again in the critiques below.

The Three Core Capacities

The operational core of Anticipatory Governance is a triad of capacities set out by Daniel Barben, Erik Fisher, Cynthia Selin, and David Guston in their 2008 chapter for the Handbook of Science and Technology Studies: foresight, engagement, and integration.² The point was never to run these as separate exercises but to let them reinforce one another.

Foresight

Foresight is the capacity to explore plausible futures systematically, without claiming to predict them. It maps possible paths, risks, and opportunities so that decisions taken today can account for a spread of tomorrows. The methods are familiar from Futures Studies: scenarios, Delphi surveys, technology assessment, and stakeholder visioning.² What matters is the stance. Guston insists that foresight here must be open-ended and exploratory rather than predictive.³ Its job is to feed strategic imagination into governance.

Engagement

Engagement is the early, substantive involvement of publics and stakeholders in evaluating and shaping emerging technologies. Guston is emphatic that it must be pursued in a less instrumental and more communicative spirit.³ Engagement that exists only to manufacture acceptance, to smooth the path for a technology already decided upon, misses the point. Done well, it produces mutual learning: publics articulate values and concerns that experts had not registered, and problem definitions themselves shift. The formats range from citizen dialogues and consensus conferences to participatory technology assessment and scenario workshops. At CNS-ASU this took concrete shape in the National Citizens‘ Technology Forum, which convened lay panels across several US cities to deliberate on nanotechnology and human enhancement.³

Integration

Integration is the coupling of natural-scientific research with social-scientific, humanistic, and ethical reflection, inside the research process rather than bolted on afterward. Social scientists and ethicists are not called in to clean up once a technology exists; they work alongside the engineers who are building it. CNS-ASU pursued this through “embedded” social scientists and humanists placed within nanoscience laboratories, probing the assumptions and choices of research as it happened.² The aim is reflexivity at the bench: a lab that keeps the question of whether and why it should build something live, not only whether it can.

A fourth element appears in Guston’s earlier writing. He calls it ensemble-ization: the deliberate combination of foresight, engagement, and integration into a mutually reinforcing whole.⁵ Participatory scenarios (foresight plus engagement) run inside transdisciplinary research collaborations (integration) are the paradigm case. In later, standard formulations the three capacities take center stage and ensemble-ization recedes into an implicit process, the connective tissue that keeps the triad from fragmenting into three unrelated activities.

Evolution of the Concept

Anticipatory Governance quickly outgrew its nanotechnology origins. The frame proved portable, and it has since been applied to synthetic biology, neurotechnology, geoengineering, and genome editing, wherever a science-intensive technology arrives faster than the institutions meant to govern it.¹

Its most consequential migration was into the broader agenda of RRI. As the framework took hold in European science policy, particularly through the European Commission’s Framework Programmes, the anticipatory frame was absorbed into that wider movement.⁴ Anticipatory Governance thereby gained reach and lost some distinctiveness. It became one strand in a movement rather than a standalone program.

International organizations have since codified the idea. The OECD’s 2024 Framework for Anticipatory Governance of Emerging Technologies cites Guston directly and translates the concept into five elements: guiding values, strategic intelligence, stakeholder engagement, agile regulation, and international co-operation.⁶ The mapping is legible. Foresight reappears as “strategic intelligence,” engagement keeps its name, and the framework binds the whole to liberal-democratic values and cross-border coordination. The translation also domesticates the concept. What began as a critical intervention in science policy becomes, in the OECD’s hands, a checklist for well-run states.

National institutionalization has followed. Germany’s Federal Environment Agency (Umweltbundesamt) drew explicitly on Guston in a 2025 report on Anticipatory Governance within the environmental portfolio, framing it as a system of institutions, rules, and norms that uses strategic foresight to reduce risk and to strengthen the state’s ability to respond early to emerging events.⁷ Here the concept leaves the seminar room entirely and enters the routines of policy planning and crisis prevention. The through-line holds across all of these adaptations: early, learning-capable steering of emerging technologies through foresight, engagement, and integration. What shifts is the surrounding language, which increasingly speaks of values, agility, and resilience.

Anticipatory Governance and Artificial Intelligence

Artificial intelligence is the sternest test the framework has faced. If Anticipatory Governance can steer any technology, it should be able to steer this one. If it cannot keep pace with AI, the concept’s core promise, to act while steering is still possible, comes under strain.

The EU AI Act, approved in 2024, is the most ambitious attempt to govern AI comprehensively, and it wears several anticipatory features on its sleeve. Its risk-based architecture sorts systems into unacceptable, high, limited, and minimal risk, and it defines “AI system” by function rather than by specific technique, a move meant to keep the law relevant as methods change. EU officials have described the design as “future-proof.”⁸ The Act also builds in adaptive machinery: delegated acts, reliance on evolving harmonized standards, fundamental-rights impact assessments, and new institutions such as the European AI Office. Read generously, these map onto Guston’s triad. Foresight shows up in the attempt to legislate for capabilities not yet deployed, engagement in the consultative and scientific-advisory structures, integration in the alignment of law, standards, and oversight.⁹

Read critically, the picture darkens. The Act’s provisions for general-purpose AI and foundation models were added late in the process, after large generative models had already reshaped the field in 2022 and 2023. Scholars and civil-society analysts point to this sequence as evidence that governance was overtaken by events: foresight and early drafting did not anticipate the scale and speed of generative AI, so the regulation risked being dated before it was even adopted.⁹¹⁰ This is anticipation failing on its own terms. The “future-proof” law had to be retrofitted to a present it had not seen coming.

Regulatory sandboxes are the Act’s most explicitly experimental instrument. Article 57 requires every member state to establish at least one AI regulatory sandbox, operational by 2 August 2026, either alone or jointly with other states. A sandbox is a controlled environment in which providers develop, train, test, and validate innovative systems under supervision before placing them on the market, with real-world testing permitted and mandatory reporting of lessons learned back to the AI Office.¹¹ Scholars read sandboxes as textbook Anticipatory Governance: learning-centric and evidence-generating. They also warn that the same design can be captured, tilting toward well-resourced incumbents and normalizing risky practice under the banner of innovation, which raises the old question of whose futures the sandbox actually anticipates.¹¹

The concept is now being applied to AI by name. The OECD’s work on AI and strategic foresight extends its Anticipatory Governance framework directly to the field, and a 2025 special issue of Policy and Society on the governance of generative AI stages a debate between technocratic and participatory versions of Anticipatory Governance, arguing that foresight over generative systems cannot be left to experts and regulators alone.¹²

The United States, by contrast, offers a cautionary tale about durability. The NIST AI Risk Management Framework (2023) is a voluntary, adaptable structure built around the functions of governing, mapping, measuring, and managing risk, and firms increasingly treat alignment with it as anticipatory regulatory compliance.¹³ But federal AI policy has proven brittle. President Biden’s 2023 executive order on safe, secure, and trustworthy AI was rescinded within days of the change of administration in January 2025.¹⁴ Anticipatory governance depends on sustained institutional commitment. A directive that can be revoked at the stroke of a pen signals to agencies, industry, and international partners that the anticipatory frame may not survive the next election.

Beneath all of this sits the question the Collingridge Dilemma first posed, now sharpened. Foundation models are retrained, fine-tuned, and repurposed faster than any legislative cycle can turn, and their emergent capabilities often become visible only after wide deployment. Foresight can sketch scenarios, but it cannot enumerate the adversarial and unintended uses of a general-purpose system. The pacing problem, the structural lag between exponential technical change and incremental institutional change, presses hardest exactly where Anticipatory Governance stakes its claim. Whether the framework answers the AI challenge or merely dramatizes it remains genuinely open.

Democratic Theory and Citizen Participation

The democratic ambition was there from the start. By tracing his concept back to Toffler’s anticipatory democracy, Guston framed technology steering as a matter for the many rather than the few.³ The engagement capacity is where that ambition lives. It insists that publics enter the conversation about emerging technologies early, before design choices harden, and that they do so as genuine participants rather than as an audience to be persuaded.

The theoretical lineage is twofold. Anticipatory Governance draws on deliberative democracy in its stress on reasoned, reflective exchange over values and options, and on participatory democracy in its insistence that broad publics, not only elected representatives, take part. It stops short of replacing representative institutions. The goal is to raise the steering capacity of state and society, not to dissolve the state into permanent assembly.

Guston does not prescribe specific mechanisms, but his engagement concept fits naturally with the deliberative instruments that have since gained ground. Citizens’ assemblies convene a cross-section of the public to deliberate on a contested question and offer recommendations, and they map cleanly onto the demand for a less instrumental, more communicative form of participation. Sortition, selection by lot, is the quiet radical here. Random selection can amplify what Guston, borrowing the phrase, calls the “still, small voices” of groups usually shut out of technology debates.¹ Where open participation tends to over-recruit the confident and the credentialed, a lottery reaches the disengaged, the unaffiliated, the ordinary. As a design choice for the engagement capacity, sortition is a way of making anticipation fairer than the publics that normally show up.

The connection runs in both directions. Deliberative bodies supply Anticipatory Governance with a legitimate venue for engagement and foresight. Anticipatory Governance supplies deliberative bodies with a rationale for looking forward rather than only adjudicating present disputes. A citizens‘ assembly on synthetic biology or frontier AI, embedded in a policy cycle and flanked by scientific integration, is close to the ideal the concept gestures at. Whether such bodies exert real influence or serve as decorative consultation is a separate question, and one the critics press hard.

Critical Perspectives

Anticipatory Governance has attracted serious objection, much of it from within the same science-and-technology-studies community that produced it. Guston addressed several of these lines directly, which makes the concept unusual: it comes with its own catalog of doubts.

Speculative Ethics: Anticipation May Not Be Necessary

The sharpest critique belongs to Alfred Nordmann. In his account of speculative ethics, anticipation reifies uncertain futures. It takes hypothetical scenarios, some of them wildly improbable, and treats them as objects solid enough to claim scarce ethical and political attention. The result is a double distortion: the present loses attention to an imagined future, and an illusion of control settles over developments no one can actually foresee. Nordmann’s conclusion is deliberately provocative. Anticipation, he argues, is “not necessary” for good governance, and the reflex to anticipate everything can crowd out the harder work of attending to what is already in front of us.¹⁵ The hermeneutic strand of Technology Assessment, associated with Armin Grunwald, reaches a related verdict from a different direction: what looks like knowledge of the future is usually a projection of present hopes and fears, better read as a text about the now than as a forecast.

Guston’s reply is that this assumes a narrow, predictivist idea of anticipation. Anticipatory Governance does not claim to foresee specific products or outcomes. It aims to build adaptive capacity and reflexivity under acknowledged uncertainty, which is precisely the condition Nordmann describes. The disagreement is real and unresolved, and it is the most important argument in the field.

The Democracy Deficit

Who decides which futures count as plausible enough to act on? In practice, often a small circle of experts, foresight professionals, and officials. Sortition can widen who sits at the table, but it cannot guarantee that the table has power. Engagement can still become a ritual that legitimates decisions already taken elsewhere. The panel deliberates, its recommendations are filed, and the binding choices are made before or after, in rooms the participants never enter. The performative worry compounds this. Scenarios that circulate widely can consolidate into shared imaginaries that shape behavior and foreclose alternatives, so the act of anticipating a future helps bring it about while presenting itself as neutral description.

The Illusion of Control and Complicity with Hubris

A structural objection holds that the very idea of anticipatorily steering complex innovation stabilizes an overblown sense of control. A softer version worries about complicity: by helping to optimize the governance of a technology, Anticipatory Governance may quietly concede that the technology should proceed, narrowing the question from “should we build this?” to “how do we build this well?” Critics in this vein see the framework as too comfortable with technoscientific progress, an accessory to hubris rather than a check on it. Guston answers that engagement and integration are meant to smuggle exactly these critical questions back into the lab and the policy process, and that reflexivity, honestly pursued, keeps the “should we” question live.

Normative Vagueness

The OECD’s “guiding values” and similar formulations leave a gap where the substance should be. Which values guide, and who chooses them? Without an answer, the framework risks being filled by whatever interests are already dominant, usually commercial or technocratic ones. Anticipatory Governance can specify a process without guaranteeing that the process serves the public rather than the powerful. Guston concedes that the concept is no panacea. His modest claim is that it might bend the long arc of technoscience a little more toward humane ends, which is a long way from the sweeping control its critics fear and its enthusiasts sometimes imply.

Open Questions

The field leaves several questions unresolved, and they cut deeper than matters of technique.

Methodological. How can foresight that includes potentially affected parties avoid capture by organized interests? What standards should govern which scenarios get taken seriously and which are dismissed as noise?

Democratic. How can Anticipatory Governance be held accountable when its anticipations prove wrong? Who answers for a policy shaped by a future that never arrived, and how would anyone even measure success when the counterfactual is unknowable?

Temporal. Can anticipatory logic keep pace with technologies, AI above all, that move faster than the institutions meant to govern them? If the pacing problem is structural rather than incidental, anticipation may be permanently one step behind, and the concept’s central promise would need rewriting.

Normative. Whose values guide the steering, and by what right? A framework that can describe a good process without naming a good end leaves its most political question open.

The most fundamental issue is performative. Anticipatory Governance does not merely observe possible futures. It helps produce them. The scenarios a society rehearses shape what it prepares for, and sometimes what it builds. That loop returns to the Collingridge Dilemma that motivates the whole enterprise. If anticipation shapes the trajectory it claims only to foresee, then early intervention might dissolve part of the information problem by steering development itself. In its place stands a new and harder problem: accountability for the futures we prevented, or caused, without ever being able to prove which.

Sources

Further Reading

Priority 1: Foundations

• Guston, David H. (2014). “Understanding ’anticipatory governance‘.” Social Studies of Science, 44(2). The essential starting point.
• Barben, D., Fisher, E., Selin, C., & Guston, D. H. (2008). “Anticipatory Governance of Nanotechnology: Foresight, Engagement, and Integration.” Handbook of Science and Technology Studies.

Priority 2: Methods and Evolution

• Stilgoe, J., Owen, R., & Macnaghten, P. (2013). “Developing a framework for responsible innovation.” Research Policy, 42(9).
• OECD (2024). Framework for Anticipatory Governance of Emerging Technologies.
• OECD (2025). AI in Strategic Foresight: Reshaping Anticipatory Governance.

Priority 3: Critical Perspectives

• Nordmann, Alfred (2007). “If and Then: A Critique of Speculative NanoEthics.” NanoEthics, 1(1).
• Nordmann, A. & Rip, A. (2009). “Mind the gap revisited.” Nature Nanotechnology, 4.