Frequently asked questions

What are (representative) deliberative democratic processes?

Representative deliberative processes (which we also refer to as ‘deliberative processes’ or ‘representative deliberations’) aim to make the process of devising and deciding on policy solutions democratically representative.

This is achieved by creating a representative microcosm (or minipublic) of the populace being governed and giving it the time, information, and structure needed to make decisions wisely. Deliberators are selected from the population through what is known as sortition or a democratic lottery—such that every person has a roughly equal chance of being selected . Thus, far fewer people (from tens to thousands) are involved than are in a referendum or election. Limiting the number of participants makes it possible for conveners to invest more resources per deliberator, so that those chosen can often be compensated for their time (which may be over forty hours for more intense processes) spent grappling with the issue, in facilitated dialogue with each other, experts, and stakeholders.

In a high-quality deliberative-democracy process, sortition (which removes many of the perverse incentives of electoral politics) is coupled with significant investment to ensure that those selected 1) can participate, by providing appropriate compensation, childcare, eldercare, and the like, which helps to reduce self-selection; 2) have sufficient context, by providing briefing information about the decision at hand and access to experts and stakeholders; and 3) can deliberate effectively, through structured discussions and activities that ultimately result in wise decisions.

The democratic legitimacy of this process comes from the representative makeup of the assemblies—far more representative than one finds in a standard elected body. Moreover, the best representative deliberations effectively communicate the “deliberative journey” to the rest of the concerned population through mass media. In this way, the broader public can see people similar to themselves learning about the issues, learning from one another, and coming to a set of conclusions that might initially have been counterintuitive. The best processes also include a mechanism for collecting public feedback and opinions, which are then shared with the deliberators along with the more traditional multistakeholder and expert input. This approach of bringing the entire population along on the deliberative journey (parascaling) is particularly helpful for maintaining democratic legitimacy.

Representative deliberations have already been used by governments and organizations around the world at every level, from small towns and utilities all the way up to the EU and UN-endorsed global pilots. Sometimes called citizen assemblies, citizen juries, citizen panels, or deliberative polls (albeit with significant differences across different approaches), representative deliberations are usually convened by a government to answer a specific question, often one that involves difficult tradeoffs or value dilemmas, for example: “How can we lower climate emissions to 40 percent of our 1990 level?” or “Should we continue building nuclear power plants?” Processes vary significantly in duration (from hours to months); whether they are offline, online, or hybrid; the number of people involved (typically tens to hundreds, but there are examples with thousands); and the structured workflows and facilitation structures that enable the participants to successfully learn about an issue and make wise decisions (process design).

A key ingredient that modern representative deliberations provide, at least in theory, is the ability to provide informed policy responses to any targeted question, with democratic legitimacy, for any population.

Whereas many busy voters may need to cast their ballots on gut instinct, participants in representative deliberations are given the (generally compensated) time and resources to make decisions based on extensive information and deliberation—ideally making the process more robust to AI-augmented advertising and manipulation. A representative deliberation also has an advantage over solely multistakeholder processes, because a representative body can act as a “democratic adjudicator,” thereby democratically weighting the voices of the different stakeholders.

*Note: This is an updated version of the explanation of representative deliberative processes from this essay.

References

Why Picking Citizens at Random Could Be the Best Way to Govern AI
Hélène Landemore, Andrew Sorota, and Audrey Tang, 2023. Fortune.

Why focus on deliberative processes?

We consider representative deliberative processes as particularly powerful instruments in the democratic toolkit. High-quality deliberative processes have a combination of properties that make them well-suited to complex decisions where it is critical that decisions are made in the public interest:

They are representative, through sortition — every member of a population has an equal likelihood of being selected; the decision-making body truly looks and feels like the population they are making decisions for.
Sortition removes the perverse incentives of electoral politics: participants have no donors, no re-election concerns, no partisan base to perform for.
They are informed: participants are given time to engage deeply with the issue, often compensated and supported by briefing materials, expert testimony, and stakeholder input.
They are deliberative: structure and facilitation help participants move beyond initial reactions to surface deeper values, find common ground even on polarized issues, and can provide substantive decisions.
These properties together help make them more robust to manipulation than most democratic alternatives — and more legitimate than most expert-driven ones.
They function as a kind of modular decision-making infrastructure — composable, deployable at many different scales, and across any jurisdiction.
This jurisdictional flexibility also means they can even be used by corporations and other organizations to delegate governance decisions to a democratic microcosm, as a complement to CEO, board, or shareholder decision-making (we’ve supported organizations like Meta and Open AI in piloting this).

Deliberative processes are uniquely valuable for issues where existing power holders, including politicians and CEO’s, should not be the decision-makers because there are dangerous conflicts of interest. This is especially critical for decisions which could be used to overcome checks on their power, or where they are disincentivized from taking necessary collective action.

This makes them particularly valuable in contexts like AI governance and alignment where decisions involve inevitable normative trade-offs, where affected publics are large and diverse, and where the stakes of getting it wrong are high.

The promise of deliberative processes also outpaces current practice. Our Democratic Capabilities Gap Map is an effort to take stock of exactly that gap, mapping the tools, research, and infrastructure that would need to exist for deliberative processes to live up to their potential. We see this as particularly urgent given the scale of transformational change that we expect will result from AI advances.

See Reimaging Democracy for AI for more context.

What types of decisions can deliberative processes handle?

Deliberative processes can handle most kinds of decisions, including but not limited to:

binary “yes”/“no” or “go”/“no go” decisions;
selecting from a set of alternatives (which may be policy options or candidates);
budgeting; and
open-ended or free-form policy recommendations.

Who decides the remit for a deliberative process?

The remit for a deliberative process may be determined by:

some procedural trigger (e.g., when the previous step in a policy-making process concludes, at a regular cadence or whenever a decision of a certain type needs to be made);
an organization commissioning the process; or
an external agenda-setting process, such as civil society organizations convening a process with the aim of solving a specific problem or running a pre-process for identifying critical problems that need a deliberative process.

What is the difference between deliberative processes and a focus group or market research?

A high-quality deliberative process produces considered collective judgment from a representative microcosm of society. These outputs carry the legitimacy that comes from a process that is designed to be free from manipulation and substantive in its deliberation.

Focus groups and market research are designed to capture existing opinions and preferences, usually in service of a specific organizational or political goal or product strategy. They sample for perspectives, participants rarely have access to balanced information or expert input, and they are not given meaningful time to reason through complex tradeoffs. The output is a snapshot of what people think instead of a considered judgment about what they would conclude if properly informed.

That distinction matters when the questions being asked are morally complex, technically uncertain, or consequential enough that the answer needs to hold up to public scrutiny.

AI is a complex, technical policy area. Shouldn't experts make the decisions?

For the most part, deliberative processes are used to make normative or political decisions, not technical ones. Of course, it is not possible to completely disentangle the two. For example, most decisions involving technology, particularly at the scale of large AI systems or online platforms, have normative and political consequences because they impact how benefits and burdens are distributed. But to the extent possible, remits for deliberative processes typically focus more on normative or political questions like:

What kind of society do we want to live in?
How should benefits and burdens be distributed?
How should we proceed, given fundamental trade-offs?
How should we proceed, given conflicting preferences?
How should we adapt to achieve [outcome], given that the possible options distribute benefits and burdens differently?

Rather than technical or empirical questions like:

What options do we have to achieve [outcome]?
What will happen if we do [intervention]?

That said, deliberative processes typically include a learning component during which participants are brought up to speed, to the extent possible in the available time, with technical and empirical knowledge relevant to the domain. This includes hearing from and being able to question people with relevant expertise.

In the context of well-designed deliberative processes, randomly selected groups of people have a track record of making good decisions, even in highly technical domains. See, for example, the processes run by Sciencewise, a publicly funded body in the UK.

What do you mean by democracy?

We understand democracy as a set of social technologies, infrastructures, and methods for power-sharing. To be impactful, it needs essential building blocks:

legal, organizational, and potentially technical structures that bind decisions to legitimate processes
transparency mechanisms that enable accountability
the capacity to adapt when those structures risk being captured or made irrelevant

We understand democracy as both normative (some decisions should not be made unilaterally due to their consequential risks on society) and structural (we need systems capable of actually enforcing this normative principle).

Specifically, our approach focuses on representative deliberative democracy as a particularly powerful democratic system capable of working across organizations of various sizes, incorporating diverse perspectives, and generating common ground on polarizing questions.

That said, deliberation is one instrument within a larger democratic infrastructure and not the whole of it. Our aim is to support those who run good enough* processes and ensure that the outcomes of those processes influence what happens in the world, that cooperation between powerful actors becomes more likely, and that no single entity, human or AI, is able to concentrate power in ways that disempower everyone else. In sum, then, by “democracy” we mean a governance architecture through which we try to keep that from happening.

*”Good enough” means that the outputs of these methods must satisfy key criteria across dimensions like quality, robustness, representativeness, or legitimacy. Success should be measured with respect to the current needs and the existing methods (and the ability to evolve to be better), not an unattainable ideal of perfect democracy and justice.

Who do you work with and how do you work with them?

Our work depends on a distributed ecosystem of actors. Some design and run the deliberative processes themselves. Others fund them, evaluate them, or help build the enabling infrastructure: the standards, guides, and community spaces that allow the field to grow. AIDF sits at the nexus of these other organizations, working to identify and prioritize gaps and opportunities to advance democratic governance.

Philanthropists and Funders

The gap between the promise of democratic representation in AI governance and current practice is, in significant part, a funding gap.

We act as a thought partner for funders navigating this space, helping them design funding strategies that build sustainable infrastructure rather than one-off pilots.

Our goal is to ensure they are well-informed about goings-on in the relevant ecosystems and have the information they need to channel funding towards organizations and projects that are underserved and vitally important.

AI companies

We help frontier AI labs and builders identify where public deliberation is most needed and connect them to practitioners capable of running rigorous processes. Over time, we aim to support labs in developing irrevocable governance commitments that include democratic participation as a trigger condition for high stakes decisions.

Builders

We also work with labs and builders on running democratic experiments, creating policies, and investing in technologies that:

enable new forms of democratic engagement (i.e. tools to support collective coordination)
make their own decision-making structure more democratic (i.e. engage larger groups of people in making key decisions)

Practitioners (Civil Society)

We actively engage, convene, and collaborate with different civil society organizations working around AI or democracy. These might include civic technologists who are developing new tools or processes to support democratic decision-making and want to explore new opportunities enabled by AI, or groups working on policy and theory for making democracies more resilient.

We work with these groups to share tools, methods, and findings across the ecosystem; to surface existing deliberative use cases that deserve wider recognition; and to help organizations build the internal capacity to commission, evaluate, or run participatory processes themselves.

See: DelibTech network

Regulators (governments, governance bodies and policy-makers)

Governments, as institutions accountable to the public, are natural homes for the kind of deliberative AI governance we’re working towards.

We support meaningful experimentation where it emerges, directly by suggesting or partnering with deliberative practitioners and organizations (many of whom have already worked at local, regional, national and transnational levels), and by supporting critical project preparation phases such as scoping, process suitability and connecting actors. We do not primarily design or implement representative deliberative processes ourselves. We pay close attention to where meaningful experimentation is happening, looking out for the use cases and key instances where democratic innovation is being applied to AI decisions. We’re drawing on those as both evidence and inspiration for the broader ecosystem.

We also maintain close relationships with policymakers, both inside and outside AI labs, who are grappling with how democratic innovation can be brought to bear on AI governance decisions. And we participate in multilateral convenings around AI and actively contribute to conversations about how democratic representation can support international AI coordination.

How does this relate to similar efforts?

Several research and policy areas work towards similar aims: building AI systems that serve the public interest and governance systems that can work better for people in a rapidly changing world. In particular:

pluralistic alignment is the project of ensuring model behavior is not monolithic and, in various ways, represent a population in which there are many different preferences or perspectives;
participatory AI is a broad research area that explores different ways in which stakeholders and affected parties can be involved in the design, development, and deployment of AI systems;
public AI focuses primarily on public ownership of compute infrastructure, public oversight of the development process, and public ownership of resulting models; and
AGI institutions (or “AGI-ready institutions,” “full stack alignment,” or “Post-AGI equilibria”) is an interdisciplinary research area focused on “the robust co-alignment of AI systems and institutions with what people value, from each individual’s pursuit of their vision of the good life to the collective achievement of shared values and ideals.”

Our work overlaps with all these, but with a particular focus on institutional design (how are decisions made?), with the practical goal of navigating the governance challenges posed by AI and ensuring that what happens in society stays broadly aligned with a democratic notion of the public interest in the long term.