Project summary

Advanced AI is creating existential governance challenges that are technically complex, marked by deep uncertainty yet regulated by lagging institutions. 
While progress has been made in AI safety research and public engagement, very few initiatives integrate:

• Rigorous expert analysis;
• computational stress-testing; and
• representative democratic deliberation.
  **The Solution **
  The Odyssean Process: The Odyssean Institute’s integrated governance model for navigating the complexities of AI governance.
  Please read the following blog post to find out more.
  Methodology: the Odyssean Process combines:
  • Structured expert elicitations on risk scenarios and interventions against them, with:
    (through an IDEA Protocol AI Horizon and Solution Scan and the Horizon Summit methodology).
  • Decision-Making Under Deep Uncertainty (DMDU) modelling; and
  • A nationally representative Citizens’ Assembly (in the United Kingdom this autumn).

What are this project's goals? How will you achieve them?

Value Proposition

Rather than producing another AI policy report, standalone discursive interventions, or a single deliberative event, the Odyssean Process generates governance recommendations that are:

• Epistemically robust;
• Stress-tested under uncertainty; and
• Democratically generated.

**Open-source outputs **

We will publish:

• Methodologies (extant, and newly refined);
• Research outputs (open access); and
• Implementation resources (Odyssean Process White Paper 2.0).

These outputs enable governments, researchers and civil society organisations to adapt the Process either modularly, or as a full pipeline, to future AI governance challenges.

Goal 1 - Build a rigorous evidence base for AI governance

Through the Odyssean Process applying horizon scanning, foresight and public deliberation, we will:

• conduct structured expert elicitation that identifies AI risks, governance challenges, and potential policy interventions; and
• synthesise perspectives from AI safety, AI governance, economics, law, philosophy, public policy and other relevant disciplines.

These create a broad and deep evidence base from which public interest research and policymaking can derive.

Outcome: A shared, interdisciplinary evidence base for understanding AI governance challenges and potential responses.

Goal 2 - Identify governance interventions that remain robust under uncertainty

We will:

• translate expert insights into computational DMDU models;
• explore thousands of plausible futures;
• stress-test proposed governance interventions against these futures; and
• identify policies that remain effective across a wide range of conditions.

Outcome: Governance interventions that are not only desirable, but demonstrably more robust to uncertainty.

Goal 3 - Support informed democratic deliberation

The outputs of the:

• Horizon and Solution Scan;
• expert elicitation; and
• DMDU modelling

Will provide the evidence base for a nationally representative Citizens’ Assembly.

Participants in the Citizens' Assembly will:

• engage with balanced, high-quality evidence;
• deliberate on alternative governance options; and
• develop recommendations for policymakers.

Outcome: Recommendations that are both technically informed and democratically legitimate.

Goal 4 - Build long-term governance capacity

Beyond the immediate project, we will publish:

• open-source methodologies;
• governance resources;
• implementation guidance; and
• research outputs.

These resources will enable governments, researchers and civil society organisations to replicate and adapt the Odyssean Process for future AI governance challenges.

Outcome: A reusable governance framework that can strengthen decision-making on transformative technologies beyond the initial UK project.

Goal 5 - Raise societal awareness, systemic policy capacity, and mobilise the public

To contribute to an ecosystem of social mobilisation and movement building, media artefacts like the documentary will dovetail with grassroot and protest mobilisations through our consortium. This ensures both political insiders and activist outsiders can be utilised for maximum robustness of strategy under conditions of rapid change and deep uncertainty.

Community & Media Outreach

We will generate comprehensive media and community engagement, as outputs for societal mobilisation and capacity building. This also increases socialisation with political actors.

Media and community outreach products include:

• A documentary (main trailer here), and also a work in progress treatment) to shift awareness and wider discourse, and spread the Odyssean Process’ recipe accessibly for movement building
• 10 grassroots community assemblies, as fractal instantiations of the Odyssean Processes in more agile form, in constituencies where data centres are being built without consultation
• A 2027 global assembly, feeding the outputs of the national assembly into a world-first global assembly, the Citizens Track on AI Governance.

How will this funding be used?

Funding will support delivery of the full Odyssean Process, enabling us to move from evidence generation through to democratic deliberation and real-world policy impact for AI regulation reform.

The project, through consortium, has raised £400,000 plus, with the Odyssean Institute directly raising £170,000, but there are remaining delivery outputs for which this funding would provide completion, addressing some of the technical and scaling dimensions required.

1. Research and evidence generation

• Evidence synthesised from across the full Odyssean Process for insights on future foresight, modelling, and deliberations - iterating on leading fields for anticipatory and resilience policy making.
• Prioritised risks, and their accompanying policy interventions.
• Scenarios of risk and the interventions that can address them, derived from multiple futures exercises (CLA, Backcasting, Domain and Stakeholder Mapping, XLRM etc.)

2. Modelling and stress-testing

• Decision-Making Under Deep Uncertainty modelling to develop:
  • Computational analysis of governance interventions
  • Stress-testing proposed policies across plausible futures
  • Developing an open-source toolkit to rapidly develop causal maps from large-scale citizen engagement
  • Leveraging this toolkit for technical AI safety risk awareness and mitigation in specialist research and policy circles

3. Democratic deliberation

• Development and delivery of the nationally representative Citizens’ Assembly, with:
  • Popular (often 75-90%+ consensus) policies for more tractable AI safety implementation
  • Preparation of education materials, made open access for similar exercises
  • Research findings on transcripts for learning how to ensure outputs are representative, nuanced and epistemic.
  • Stakeholder coordination and engagement through with 10 community assemblies across the UK.

We will use the following indices to assess the effectiveness of our methods:

• Discourse Quality Index (DQI) - a quantitative research tool used to measure the level of rational deliberation and democratic quality in political or civic speech acts. This analyses outputs in terms of participants providing extensive and complex justifications for positions. Orienting towards the common good, this shows respect for other demands and counterarguments, in addition to empathy for the needy, and offering a mediating proposal of how to resolve the dispute.
• Listening Quality Index (LQI) - Analyses the quality of listening, first from the responses of the listeners, starting with passive acknowledgement, then argumentatively substantive responses, and then responses that are well-received by the original speaker. This ensures each layer builds up to fully deliberative outputs that acknowledge the standpoints and evidence more comprehensively.
• Deliberative Reason Index (DRI) - involves measuring mutual constructions of understanding (or shared ‘logic’) of what is at stake, establishing the contours of ‘reasonableness’. Individual positions may diverge, but on terms that are mutually agreed; such terms include higher-level agreement on relevant interests, values, and beliefs (metaconsensus), as well as coherent, mutually shared ‘representation’ of the issues, guiding how considerations inform preferences.

The indices facilitate reasoning under ideal deliberative conditions of openness, reciprocity, and sincerity. These induce ‘enlarged thinking’, which enhances ‘integrative reasoning’, a more demanding version of the psychological construct of integrative complexity, which measures reasoning in terms of differentiation and integration of perspectives (Tetlock).

Read more: Tetlock, Philip E. 1986. ‘A Value Pluralism Model of Ideological Reasoning’. Journal of Personality and Social Psychology 50: 819–827.

4. Communications and policy translation

• Promotion and dissemination of research findings via traditional and contemporary media formats, including news interviews, online news articles and social media outreach.
• Engagement with policymakers and relevant institutions, including MPs and Senior Civil Servants.
• Translation of Citizens’ Assembly recommendations into practical policy options, to socialise with peak bodies, think tanks and public institutions.
• Supporting the uptake of findings into real-world governance.

5. Open-source capacity building

• Publication of methodologies
• Open-source research outputs
• Governance resources and implementation guidance
• Development of materials that can support international replication

Who is on your team? What's your track record on similar projects?

The team covers the full range of expertise required to deliver the Odyssean Process, from foresight and AI safety through to computational modelling, democratic deliberation and institutional engagement.

Giuseppe Dal Prá - Founder & CEO

• Co-creator of the Horizon Summit methodology with James Balzer.
• Leads institutional partnerships and programme leadership.
• Leads the development and implementation of the Odyssean Process.

James Balzer - Foresight Lead

• Co-creator of the Horizon Summit methodology with Giuseppe Dal Prá.
• Leads strategic foresight, horizon scanning and systems analysis across the Odyssean Process.
• Leads the development and application of foresight methodologies to complex governance challenges.

Jacob Haimes - AI Safety Lead

• Leads technical AI safety.
• Leads structured expert elicitation.
• Coordinates analysis of AI risks and governance interventions.

Dr Swaptik Chowdhury - DMDU Lead

• Develops computational models for Decision-Making Under Deep Uncertainty.
• Translates qualitative expert insights into quantitative policy analysis.
• Leads computational stress-testing of governance interventions.

Renske van Vroonhoven - Operations Lead

• Leads programme management.
• Coordinates stakeholders and delivery.
• Supports delivery of the Citizens’ Assembly and associated activities.

Track record

Together, the team has conducted several groundbreaking foresight exercises (EEJ) via both Horizon Scans (and some joint Solution Scans blended) and the novel, internally developed Horizon Summit methodology across:

• global catastrophic risks and their critical tipping points;
• nuclear resilience;
• urban resilience;
• extreme heat; and
• AI governance.

This includes a world-first horizon and solution scan of the second-order impacts of nuclear conflict (funded by Longview Philanthropy), which is now informing international resilience work, through another collaboration on extreme resilience with Nagasaki University and Seoul National University, to produce a peer-reviewed input to the UN independent scientific panel on nuclear war impacts.

The Horizon Summit methodology has also supported work with international and city-level partners. These include cities such as New York City Government, Auckland City Council, City of Melbourne, City of Cape Town and City of Recife. Real world policy applications included urban resilience and AI policy futures for cities.

The Institute has both independently, and leading broader consortia built capacity around the project, including in-team experience and external partners such as:

For EEJ:

• James Balzer and Matthew Ball have robust experience in foresight and policymaking across Australian and British governments, in addition to international organisations. James also has policy and foresight experience across several state governments.
  • Experience from this led to us exceeding the outputs and rigour often used in national governments’ foresight, to ensure we continue to push the art of the possible.
  • The methodology for the elicitation / horizon scanning process was more rigourous and detailed than several studies that our team members implemented in government, learning directly from the limitations that slower moving institutions face. It included more deliberative rounds and iteration between stages, a formal debiasing protocol, and mathematical aggregation methods (rather than editorial drafted synthesis of comments from workshops which GO-Sci does). Both the solution scan and DMDU are also differentiators.
• Historically on AI, and for this project currently, experts we’ve worked with include:
  • Matthijs Maas
  • Ortwin Renn
  • Luke Kemp
  • Sam Stockwell
  • David Thorstad
  • David Manheim
  • Omer Bilgin

For DMDU:

• Leading practitioners and professors including
  • President of the DMDU Society, Jan Kwakkel
  • Imperial College London and Advisor to British Government, Mark Workman
• For the Assembly:
  • Involve UK (the leading delivery and consultancy organisation for deliberation in the United Kingdom) and
  • The Sortition Foundation (the leading representative sampling organisation across Europe and Oceania for democratic deliberation) supporting the Citizens’ Assembly and its implementation.

What are the most likely causes and outcomes if this project fails?

Primary risk: insufficient funding

The principal risk is insufficient funding to deliver the Odyssean Process at its intended scale.

If only partial funding is secured, we will prioritise:

1. Strategic policy translation, engaging decision makers and social movements in Britain and Europe, including with the documentary
2. Further extension of Fellows to assist with the automated causal loop, robust possible future modelling. Funding here can produce a replicable toolkit more quickly for other risk management projects.
3. Operational runway to alleviate any bottlenecks in finalising this, and beginning scaling operations in 2027.

We would then reduce investment in scaling:

• Communications; and
• international dissemination.

Consequences

A partially funded project would still generate useful research and deliberative outputs, but would have:

• less computational stress-testing of proposed interventions;
• reduced capacity to translate findings into real-world policy by socialising findings as actively;
• less extensive communication and dissemination;
• weaker international replication capacity; and
• fewer resources for developing the open-source infrastructure required to extend the Odyssean Process.

Mitigation

Even under partial funding, we will ensure that:

• major research outputs are documented;
• methodological insights are openly recorded; and
• the core findings remain available for future use and replication.

How much money have you raised in the last 12 months, and from where?

How much money have you raised in the last 12 months, and from where?

The Odyssean Institute has secured support from a growing network of philanthropic partners, including:

• King’s College, University of Cambridge £780
• Survival and Flourishing Fund - $70,000 USD
• Nagasaki University Research Center for the Abolition of Nuclear Arms (RECNA) - £15,000
• Anonymous Donor - £62,000 (Forthcoming, September 2026)

We have also fundraised in periods just over 12 months, as the lead time to develop the correct experienced partners, and the part time bottlenecks in team capacity, meant some of the funding for this project (and the Institute) was raised earlier than September 2025:

• Greg Colbourn (February 2025) - £10,000
• Kristian Rönn (April 2025) - $25,000 USD

These grants have enabled us to develop and validate core components of the Odyssean Process while building partnerships across:

• AI governance;
• public deliberation; and
• computational policy analysis.