Signals from the frontier. Projects happening now that illuminate — from different angles and at different scales — what becomes possible when genuinely different ways of knowing are brought into sustained encounter with hard problems.
The historical examples in The Argument show that synthesis across different ways of knowing has happened before. What this section documents is something different: what is happening now, in our time, at the frontier of collaboration. These are not proof that WhatIfWe will succeed. They are signals — each one illuminating a specific dimension of what becomes possible, and each one honest about the ceiling it reached.
Read together, they trace a spectrum — from collaboration across nations within one tradition, through collaboration across disciplines, to the rarer and more demanding collaboration across genuinely different ways of knowing. That arc makes legible both how far humanity has come, and the gap that WhatIfWe is designed to close.
Each case study identifies what it demonstrates — and acknowledges its limit. The limit is not a criticism. It is the map of where the next attempt must begin.
What shared purpose can achieve across the deepest national competition — and what it cannot.
Scientists from countries that cannot agree on almost anything have repeatedly assembled the most sophisticated instruments in history. The governance model that made this possible — holding scientific independence and political authority in deliberate tension — is one of the closest analogies to institutional design that WhatIfWe will need.
The first formal architecture for integrating indigenous and local knowledge alongside Western science in global policy.
IPBES didn't just include indigenous voices — it built procedures for indigenous knowledge to function as a parallel and co-equal epistemic system alongside peer-reviewed science. That structural choice, and the resistance it generated, is one of the most instructive case studies in what genuine cross-epistemological encounter actually requires.
A religious community that ended a sixteen-year civil war — achieving what professional diplomacy could not by operating across domain boundaries.
Sant'Egidio's success in Mozambique — and subsequently in Guatemala, Kosovo, Burundi and Congo — came precisely from its refusal to be a single-domain actor. It operated simultaneously as a faith community, a neutral mediator, a humanitarian organisation, and a long-term relationship builder. The synthesis of those roles is what made it credible where states were not.
The most deliberately engineered large-scale collaboration framework in existence — and what its design reveals about the architecture WhatIfWe needs.
Unlike CERN or the ISS, which emerged from geopolitical circumstance, Horizon was designed from the start as a system for producing cross-boundary collaboration. Its mechanisms — for forming consortia, distributing funding, measuring outputs, iterating across programme generations — are the closest available model for thinking about what WhatIfWe's own infrastructure needs to do.
WhatIfWe does not exist in a vacuum. A significant body of research and a growing number of initiatives are working on adjacent problems — from AI-assisted deliberation to collective intelligence systems to knowledge synthesis across domains. Understanding precisely where that frontier is, and what it has not yet reached, is what makes WhatIfWe's specific contribution legible.
The most sophisticated existing attempt to use AI for large-scale collective intelligence is the vTaiwan / Polis ecosystem, developed in Taiwan since 2014 under Digital Minister Audrey Tang. Polis uses machine learning to cluster opinions across large populations — surfacing latent consensus that polarised discourse makes invisible. By 2025 it had engaged over 200,000 participants and contributed to 26 pieces of legislation. Google DeepMind's Habermas Machine (2024) extended this further: tested across 5,700 UK participants, its LLM-based system produced consensus statements rated by participants as clearer and less biased than those generated by human mediators.
Both systems are designed to surface agreement within a shared framework of values and categories. They are architecturally optimised for consensus — not for sustaining productive difference across incommensurable ways of knowing. The Habermas Machine's own researchers acknowledge it lacks the capacity to handle 'mediation-relevant' aspects of real-world deliberation. More fundamentally: it operates within a single epistemological register. When participants share the same basic assumptions about what constitutes evidence, progress and reasonable argument, Polis and the Habermas Machine work extraordinarily well. When those assumptions are themselves what differs — as they do between a Quechua knowledge holder, a systems scientist, and a Sufi philosopher — the architecture has no mechanism for the encounter.
A substantial body of peer-reviewed research — anchored by Woolley and Gupta's 2024 framework and the Patterns journal's 2024 review of AI-enhanced collective intelligence — establishes that collective intelligence emerges from three interdependent ingredients: collective memory, collective attention, and collective reasoning. The research shows that AI can support each of these: helping groups leverage distributed knowledge, synchronise attention, and amplify diverse thinking styles. The UK's AI for Collective Intelligence initiative (AI4CI), spanning seven universities, and the Collective Intelligence Project in the US are the most developed institutional expressions of this research agenda.
Every system in this research tradition — however sophisticated — has been developed for groups operating within a shared epistemological culture: typically Western, academic, and scientifically trained. The Doshi and Hauser experiment (Science Advances, 2024) revealed a structural tension that this tradition has not resolved: AI assistance enhances individual creativity while reducing collective diversity by measurable degrees. When different AI tools are used by multiple participants, their outputs converge — a phenomenon now termed algorithmic monoculture in the research literature. The antidote proposed in the latest research (Wan and Kalman, 2025) is deliberate epistemic differentiation: AI systems with genuinely distinct orientations that preserve, rather than collapse, difference. WhatIfWe's multi-agent, tradition-oriented architecture is the first attempt to build this at the level of genuinely different ways of knowing — not different personas within a single epistemological framework.
Nora Bateson's Warm Data framework — developed at the International Bateson Institute and now applied across policy, healthcare, and organisational design — names a structural problem that technical collective intelligence systems have not solved: knowledge that exists only in the relationship between things is destroyed when extracted from its context. Warm Data is 'transcontextual information about and within the interrelationships that integrate elements of a complex system.' The Warm Data Lab methodology offers structured processes for working with exactly this kind of knowing — the knowing that tells you not what a system contains, but how it is alive.
The Warm Data framework diagnoses the problem with extraordinary precision. What it does not yet provide is a scalable infrastructure for what it points toward — a persistent environment where transcontextual knowledge can accumulate across sessions, be held by an AI layer that learns the relational texture of what is emerging, and connect people from genuinely different traditions around problems hard enough to require it. Bateson's work names the territory. WhatIfWe is an attempt to build infrastructure for it.
The Plurality project — developed by Glen Weyl and Audrey Tang and published in 2024 — represents the most ambitious current attempt to articulate a political and technological philosophy for harnessing diversity rather than erasing it. Plurality argues that digital tools should be designed to channel the 'potential energy in social diversity that can erupt in conflict instead for progress, growth and beauty.' Drawing on Taiwan's decade of digital democracy experimentation, the book proposes a range of technologies — from augmented deliberation tools to new models for digital identity — that recognise and strengthen collaboration across difference.
Plurality's diversity is primarily political, cultural and demographic. It addresses the problem of how communities with different values and interests can collaborate on shared governance. What it does not address is epistemological diversity — the deeper difference between traditions that have developed genuinely distinct instruments for perceiving what is real. A Daoist philosopher and a systems ecologist may share democratic values and civic goodwill. What differs between them is not their politics but their instruments of perception. That gap requires something Plurality is not designed to create: sustained encounter at the frontier of what each tradition can see, held long enough for something genuinely new to emerge.
The most technically adventurous current research in this space is Contemplative AI (Laukkonen et al., arXiv 2025): a framework that attempts to instil contemplative principles — mindfulness, non-duality, emptiness, boundless care — directly into AI architectures and constitutions. The research demonstrates that prompting AI to reflect on these principles measurably improves performance on ethical benchmarks and increases cooperative behaviour. Separately, the Abundant Intelligences initiative (Lewis, Whaanga et al., 2024) argues that AI's current trajectory suffers from fundamental epistemological shortcomings, systematically excluding non-Western ways of knowing, and proposes rebuilding AI's epistemological foundations on Indigenous knowledge systems.
Both initiatives are working within the AI system itself — trying to make AI more contemplative, more epistemically diverse, more aligned with non-Western knowledge. This is important work. But it addresses a different problem than WhatIfWe. The question WhatIfWe is asking is not how to make AI wiser, but how to create the conditions under which human wisdom — already present in living traditions — can encounter other human wisdom in a way that generates something neither could reach alone. AI is the facilitator of that encounter, not its subject. The contemplative and Indigenous AI research is developing better facilitators. WhatIfWe is designing the encounter those facilitators need to serve.
The technology world has made remarkable progress on four adjacent problems: surfacing latent consensus within shared frameworks (Polis, Habermas Machine); augmenting collective intelligence within epistemologically homogeneous groups (AI4CI, Collective Intelligence Project); diagnosing what happens to knowledge when it is extracted from its relational context (Warm Data); and beginning to address AI's epistemological monoculture (Contemplative AI, Abundant Intelligences). Each initiative is working at the ceiling of what its approach can reach — and naming, at that ceiling, the gap it cannot cross.
The gap they are each naming — from different directions, in different vocabularies — is the same gap. It is the gap between intelligence that operates within a single way of knowing, however sophisticated, and the intelligence that can only emerge from sustained genuine encounter between genuinely different ways of knowing. That gap is what WhatIfWe is specifically designed to address. Not by building better AI. Not by scaling deliberation. But by creating the conditions — the human depth, the structural design, the AI facilitation layer — under which something can emerge that none of the existing approaches, combined, can yet produce.
These four case studies are a starting point, not a definitive map. This space is larger than any single curator can see. If you are aware of a project — in any domain, at any scale — that illuminates what becomes possible at the frontier of cross-tradition collaboration, we want to know about it. In a later phase, community members will be able to contribute case studies directly. For now, write to us.
hector@whatifwe.community