Type: Report
Authors: Gitcoin Research
Sources:
- Gordon Brander — Noosphere, a Protocol for Thought
- GreenPill Episode 124 — A Protocol for Thinking Together with Gordon Brander & Michael Garfield
- GreenPill Episode 65 — Better Collective Intelligence to Better Collective Resource Allocation with Griff Green
- Pol.is
- Loomio
The Sensemaking Crisis
Before a community can allocate capital well, it needs to think well. Before a funding round can identify the most impactful projects, voters need to understand what impact looks like. Before a DAO can govern itself, its members need shared context, shared language, and shared methods for resolving disagreement.
This is the sensemaking problem, and it precedes every question of mechanism design. You can build the most elegant quadratic funding mechanism in the world, but if the community using it cannot distinguish signal from noise, cannot surface relevant information, cannot converge on reasonable assessments of value — the mechanism will produce garbage outputs from garbage inputs.
The public goods funding ecosystem has invested heavily in allocation mechanisms: quadratic funding, retroactive funding, conviction voting, participatory budgeting. It has invested far less in the cognitive infrastructure that makes those mechanisms work. This report examines the emerging field of collective intelligence infrastructure — the tools, protocols, and design patterns that help groups think better together.
What Collective Intelligence Means
Collective intelligence is not a metaphor. It is a measurable phenomenon: the capacity of a group to solve problems, make decisions, and generate knowledge that exceeds the capacity of any individual member.
Research in collective intelligence identifies three core ingredients. Collective memory is the group's ability to store, retrieve, and share information across its members. Collective attention is the group's ability to synchronize focus on the most relevant problems and information. Collective reasoning is the group's ability to process information, evaluate options, and reach decisions that account for diverse perspectives and knowledge.
When these three functions work well, groups exhibit a kind of distributed cognition that is qualitatively different from individual thinking. The classic example is the "wisdom of crowds" — the finding that the median estimate of a large group is often more accurate than the estimate of any individual expert. But collective intelligence goes beyond statistical aggregation. At its best, it produces emergent understanding that no individual could have reached alone.
When these functions break down, the result is collective stupidity: groupthink, information cascades, polarization, and the kind of mob dynamics that produce bank runs, witch hunts, and bad governance decisions. The difference between collective intelligence and collective stupidity is not the quality of the individuals involved — it is the quality of the infrastructure they use to think together.
The Landscape of Collective Intelligence Tools
A growing ecosystem of tools and protocols is being built to support collective intelligence. These range from simple collaboration platforms to ambitious new protocols for decentralized knowledge.
Pol.is: Mapping Agreement at Scale
Pol.is is one of the most proven collective intelligence tools in existence. Developed by the Computational Democracy Project, it is a wiki-survey platform designed for large-group deliberation. Participants respond to short statements by voting agree, disagree, or pass. Crucially, they can also submit new statements. Pol.is does not allow replies to other people's statements, which eliminates the trolling and flame wars that plague traditional comment sections.
The real innovation is in the visualization. Pol.is clusters participants into opinion groups using dimensionality reduction algorithms, then highlights the statements that bridge groups — the areas where people who disagree on most things nonetheless agree. This surfaces consensus that would be invisible in a traditional debate format.
Pol.is gained international attention through its deployment in Taiwan's vTaiwan initiative. Beginning in 2014, vTaiwan used Pol.is to facilitate citizen deliberation on technology regulation issues. Of the 26 national issues discussed on the platform, 80% led to government action. The platform was instrumental in developing Taiwan's regulations for ride-sharing services and the regulation of online alcohol sales, among other issues.
The lesson for public goods funding is direct. Before a community decides how to allocate a matching pool, it needs to understand what its members actually value. Pol.is-style tools can surface that understanding at scale, producing a legible map of community priorities that funding mechanisms can then act on.
Loomio: Structured Decision-Making for Distributed Groups
Where Pol.is excels at opinion mapping, Loomio focuses on structured decision-making. Founded in 2012 by members of the Occupy movement in New Zealand, Loomio provides a platform for asynchronous deliberation and voting. Groups can create proposals, discuss them over a defined period, and reach decisions through various voting methods (consent, consensus, majority, etc.).
Loomio has been adopted by a range of organizations, from cooperatives and nonprofits to DAOs and Web3 communities. Its strength is in making the invisible labor of decision-making visible — tracking who has participated, who has not, where objections lie, and whether the group has actually reached a decision or merely run out of patience.
For DAOs and funding communities, Loomio addresses a common failure mode: the illusion of consensus produced by token-weighted voting. When a small number of token holders determine the outcome of every vote, the group is not exhibiting collective intelligence — it is exhibiting plutocracy with democratic aesthetics. Loomio's deliberative structure can complement on-chain voting by ensuring that discussion, objection, and revision precede the final vote.
Colony: Reputation-Weighted Collaboration
Colony takes a different approach, building decentralized organizational infrastructure that uses reputation rather than token holdings as the basis for governance power. In Colony, reputation is earned through contributions — completing tasks, delivering work, participating in governance. This reputation decays over time, ensuring that governance power reflects ongoing participation rather than historical token purchases.
Colony's relevance to collective intelligence is in its approach to signal quality. Not all opinions are equally informed. In a technical funding decision, the assessment of an experienced developer should carry more weight than the assessment of someone who joined the community yesterday. Colony's reputation system provides a mechanism for weighting contributions by demonstrated competence, without resorting to centralized credentialing.
Prediction Markets: Aggregating Distributed Knowledge
Prediction markets are perhaps the purest mechanism for collective intelligence. By allowing participants to trade on the probability of future events, they aggregate distributed information into a single price signal that reflects the collective's best estimate of what will happen.
Polymarket, the leading decentralized prediction market, has demonstrated remarkable accuracy. In 2025, Polymarket correctly predicted outcomes — from election results to Nobel Prize winners — hours or days before official announcements, consistently outperforming polls and expert forecasts. The mechanism works because participants have financial incentives to be accurate, diverse information is aggregated through market prices, and those with better information or analysis are rewarded with greater profits (and thus greater influence on future prices).
The connection to public goods funding is through futarchy — Robin Hanson's proposal to use prediction markets for governance. In a futarchy system, decisions are made by betting on which policy will produce better outcomes according to a predefined metric. Applied to public goods funding, futarchy could allow communities to bet on which projects will generate the most impact, with funding allocated based on market-aggregated predictions rather than subjective voting.
Quadratic Voting: Democratic Signal Aggregation
Quadratic voting, developed by Glen Weyl, is itself a collective intelligence mechanism. By allowing voters to express the intensity of their preferences (not just their direction), it captures more information per vote than simple majority rule. The quadratic cost structure — where the price of additional votes increases quadratically — prevents wealthy participants from dominating while still allowing people to express strong preferences on issues they care most about.
In the context of public goods funding, quadratic voting has been deployed both as a direct allocation mechanism (through quadratic funding) and as a governance tool for deciding which projects receive support. Its strength as a collective intelligence tool lies in its ability to aggregate preference intensity, not just preference direction.
Gordon Brander's Noosphere: A Protocol for Thought
One of the most ambitious projects in the collective intelligence space was Noosphere, created by Gordon Brander and Chris Joel. Noosphere aimed to be nothing less than a protocol for thought — a decentralized, credibly neutral infrastructure for shared knowledge.
The vision was built on several key principles, as Brander articulated in his Squishy Computer newsletter and on GreenPill Episode 124:
Self-sovereign data. Users own their notes, ideas, and connections through public-key cryptography. No platform can lock them in or lock them out.
A shared thought graph. Every user maintains a personal knowledge graph. By "following" other users, their graphs merge — creating a massively multiplayer network of interconnected ideas. Each follow adds new synapses to a collective second brain.
Credible neutrality. Noosphere was designed as a protocol, not a platform. No single entity controls the network, the data, or the rules of engagement. This is essential for a tool that mediates collective thought — the infrastructure of thinking should not be owned by any party with an interest in what people think.
Subconscious, the app built on top of Noosphere, was designed as a "tool for thinking together." It used a slipbox (Zettelkasten) format, where notes are atomic, linked, and composable — more like neurons in a brain than pages in a book.
The project shut down in 2024, but the ideas it explored remain deeply relevant. The core insight — that collective intelligence requires shared infrastructure for knowledge, not just shared infrastructure for voting — points to a gap in the current public goods funding ecosystem. We have sophisticated mechanisms for allocating capital. We have far less sophisticated mechanisms for building the shared understanding that should inform those allocations.
The Sensemaking-to-Allocation Pipeline
The relationship between collective intelligence and capital allocation is not abstract. It is a concrete pipeline:
Step 1: Information gathering. The community surfaces relevant information about the problem space, the available projects, and the criteria for evaluation. Tools: Pol.is, Noosphere-style knowledge graphs, prediction markets.
Step 2: Sensemaking. The community develops shared understanding of the information — identifying patterns, resolving contradictions, building models of impact. Tools: deliberation platforms (Loomio), structured dialogue processes, AI-assisted synthesis.
Step 3: Preference expression. The community expresses its preferences about what should be funded, with what priority, and at what scale. Tools: quadratic voting, conviction voting, participatory budgeting.
Step 4: Capital allocation. The expressed preferences are translated into actual funding flows through a mechanism. Tools: quadratic funding, retroactive funding, direct grants, streaming payments.
Most public goods funding systems start at Step 3 or Step 4. They ask communities to vote or donate without first investing in the collective intelligence infrastructure needed for Steps 1 and 2. The result is predictable: funding decisions driven by name recognition, social capital, and narrative skill rather than by genuine assessment of impact and need.
Improving the sensemaking pipeline does not require abandoning existing mechanisms. It requires layering collective intelligence tools on top of them — ensuring that communities have the infrastructure to think well before they are asked to decide.
AI as an Amplifier of Collective Intelligence
The emergence of large language models and other AI systems creates both opportunities and risks for collective intelligence.
The Opportunity
AI can augment each component of collective intelligence:
Collective memory: AI can summarize, organize, and retrieve information across vast corpora of community knowledge. Instead of expecting every voter to read every project application, an AI system can synthesize key themes, highlight relevant history, and surface connections that humans would miss.
Collective attention: AI can help groups focus on what matters by filtering signal from noise, identifying the most contested or consequential decisions, and flagging information gaps that need to be filled before a decision can be made.
Collective reasoning: AI can serve as a "reasoning prosthetic" — helping groups explore the implications of different decisions, model potential outcomes, and identify blind spots in their analysis. Research from the University of Michigan suggests that AI can enhance group problem-solving by contributing speed, scale, and pattern recognition that complement human judgment.
The Harvard Business Review has reported on how AI can help tackle collective decision-making by creating structured deliberation environments, synthesizing diverse viewpoints, and ensuring that all relevant perspectives are represented in the decision-making process.
The Risk
The risk is that AI replaces collective intelligence rather than amplifying it. If communities delegate sensemaking entirely to AI systems, they lose the distributed cognition that makes collective intelligence valuable. The strength of collective intelligence is that it aggregates diverse perspectives, local knowledge, and lived experience that no centralized system — whether human or artificial — can replicate.
There is also the risk of AI-mediated manipulation. If an AI system summarizes project applications for a funding round, the way it frames those summaries shapes how the community evaluates them. Whoever controls the AI controls the frame, and whoever controls the frame influences the outcome.
The design principle should be: AI as infrastructure for human collective intelligence, not AI as a replacement for it. AI should make it easier for humans to share information, build understanding, and express preferences. It should not make decisions on their behalf.
Why This Matters for Public Goods Funding
The public goods funding ecosystem faces a scaling problem. As the number of projects, the amount of capital, and the diversity of stakeholders increase, the cognitive demands on participants grow faster than their capacity to meet them. Gitcoin Grants rounds now feature hundreds of projects. Optimism RetroPGF rounds have included over 500 recipients. No individual voter can meaningfully evaluate every project.
Without collective intelligence infrastructure, the result is predictable: voters rely on heuristics (fund projects they recognize), social signals (fund projects endorsed by trusted people), and satisficing (fund the first few projects that seem reasonable and stop). These are not irrational strategies — they are adaptive responses to cognitive overload. But they produce systematically biased outcomes that favor established projects over newcomers, popular narratives over important-but-boring infrastructure, and visible work over invisible contributions.
Better collective intelligence tools can address this directly:
- Pre-round sensemaking using Pol.is-style tools to identify community priorities before funding decisions are made.
- Structured deliberation using Loomio-style platforms to discuss and debate project evaluations before voting begins.
- AI-assisted synthesis to help voters navigate large project pools without losing the nuance of individual applications.
- Reputation-weighted evaluation to ensure that domain experts have proportional influence on funding decisions in their areas of expertise.
- Prediction markets to aggregate distributed knowledge about which projects are most likely to deliver impact.
None of these replace the funding mechanisms themselves. They improve the quality of the inputs those mechanisms receive, which directly improves the quality of the outputs they produce.
Design Principles for Collective Intelligence Infrastructure
Based on the tools and experiments reviewed above, several design principles emerge:
1. Separate sensemaking from decision-making. Pol.is succeeds because it maps opinions before asking for decisions. Combining the two — as most voting systems do — produces worse outcomes for both.
2. Reduce the cost of participation. The biggest bottleneck in collective intelligence is not the quality of individual thinking but the quantity of participation. Tools that make it easy to contribute — through quick voting, lightweight feedback, or passive data collection — generate better collective intelligence than tools that require deep engagement from every participant.
3. Surface bridging opinions. Pol.is's most powerful feature is its ability to identify areas of agreement between people who disagree on most things. This is where collective intelligence lives — not in the things everyone agrees on, but in the unexpected consensus that emerges from structured dialogue.
4. Weight contributions by competence. Not all opinions are equally informed. Collective intelligence systems should incorporate mechanisms for recognizing expertise without creating gatekeepers. Reputation systems, prediction market track records, and peer evaluation all offer approaches.
5. Make the infrastructure credibly neutral. The tools for collective thinking must not be controlled by any party with a stake in the outcome. This is the strongest argument for protocol-level infrastructure (like Noosphere) over platform-level products (like any single app).
6. Iterate and adapt. Collective intelligence is not a static property. It emerges from repeated interaction, feedback, and adaptation. The best tools are those that learn from each cycle and improve the next one.
Conclusion
The public goods funding ecosystem has made remarkable progress on the question of how to allocate capital. Quadratic funding, retroactive funding, conviction voting, and participatory budgeting represent genuine innovations in mechanism design. But these mechanisms are only as good as the collective intelligence that feeds them.
The next frontier is not better allocation mechanisms — it is better thinking infrastructure. Tools that help communities gather information, build shared understanding, and express informed preferences will do more to improve funding outcomes than any tweak to the mathematical formulas that distribute matching pools.
As Gordon Brander put it on GreenPill, the goal is to build "a protocol for thinking together." The crypto ecosystem has built remarkable protocols for transacting together, governing together, and funding together. The missing layer is the protocol for understanding together — the collective intelligence infrastructure that makes all the other coordination possible.
Building that infrastructure is, itself, a public good. And it may be the most important one.
