Bubble Forecast Intelligence™
The Bubble Research Concept: human-led, AI-enhanced. Bubble Forecast Intelligence is a conceptual model for research and forecasting that is led by people. It combines the rigor of lived experience, intuition, and years of expertise with AI tools that people direct and check. Four bubbles of experts take each question from raw signals to a decision-ready brief. or .
Human-led, AI-enhanced research and forecasting.
Bubble Forecast Intelligence™ is a model for research and forecasting in which experienced researchers lead the work and AI helps them move faster. People review and sign off on everything the AI does.
The idea
AI can now summarize what has been published in minutes. But much of what matters in a real decision was never written down: how a county council will actually vote, why a hospital is really closing, what people on the ground are saying.
That knowledge lives with people. In 2025, more than 500 staff lost their jobs or were told not to return at just three U.S. think tanks. Many of them spent their careers answering exactly these kinds of questions.
Bubble Forecast Intelligence puts that experience back to work. Small teams of researchers, called bubbles, take each question in stages. AI speeds up the reading, sorting, and checking. People make the calls and put their names on them.
How a question moves
A bubble can only close, or pop, after the one before it. Each one records its sources, how confident it is, and exactly how AI was used.
Where we are
This is a concept. The demo on this site walks a sample question through all four bubbles so you can see how it would work.
We are raising funds to build the platform, recruit and train the first teams, and run a pilot on real questions, with results scored in public against what actually happened.
If you fund research, run a forecasting or prediction market platform, or have spent a career answering hard questions, we would like to talk.
People
Bubble Forecast Intelligence is built around people with real research experience. This page explains who does the work, how people join, and what each role requires.
Who does the work
Experienced researchers from think tanks
500+
More than 500 staff at just three U.S. think tanks lost their jobs or were told not to return in 2025. RAND cut 11% of its workforce (192 people), including policy analysts, researchers, and scientists. The U.S. Institute of Peace dismissed more than 200 U.S. staff, and about 130 Wilson Center employees were told not to return.
Many more researchers left federal agencies, which lost about 9% of their workforce by March 2026. This is a deep pool of trained, experienced people ready for meaningful work.
Research students and early-career researchers
Graduate students and recent graduates in policy, economics, data science, and the social sciences. They join bubbles alongside experienced researchers, learn the method on real questions, and build a track record.
Students gain hands-on experience. Teams gain energy, fresh skills, and extra capacity.
Distinguished experts
Senior fellows, former agency leaders, retired officials, and recognized specialists in their fields. They serve as senior advisors and red team members, bringing deep judgment and credibility to the most important questions.
Their involvement signals to clients and funders that every conclusion has been tested by people who know the field.
Sources: RAND layoffs, The Lookout, Oct. 24, 2025 · U.S. Institute of Peace, NPR, Mar. 30, 2025 · Wilson Center, Buenos Aires Herald, Apr. 5, 2025 · Federal workforce cuts, Wikipedia
How participants join
- ApplyShare your background, areas of expertise, and the roles that interest you.
- VettingCredentials and references are checked. Everyone discloses conflicts of interest, including any prediction market positions.
- Role placementYou are placed in the role that fits your experience, from research student to senior advisor.
- OnboardingTraining on the four-bubble method, the rules for using AI, and the public-information-only policy.
- AssignmentYou are matched to bubbles on questions where your expertise fits.
- GrowthYour track record, confidence accuracy, and peer reviews open the way to larger roles.
Roles and what each requires
Proposed requirements for this concept. Final requirements will be set as the model is tested.
| Role | Typical background | Required for the position | Where they work |
|---|---|---|---|
| Research student | Graduate student or recent graduate | Research methods training, a supervising team lead, and signed ethics rules | Swarm and Scrape; supports other bubbles |
| Researcher | Policy or data analyst, often from a think tank or agency | At least 3 years of research experience and a writing sample | Swarm and Scrape; Synthesis and Context |
| Subject matter expert | Specialist in a field, such as health, energy, or security | Recognized expertise, usually 8 or more years in the field | Synthesis and Context; Stress Testing |
| Peer auditor | Senior researcher with methods or evaluation experience | Independence from the team being audited, and strong methods skills | Signs off on any bubble |
| Team lead | Experienced project or research manager | At least 5 years leading research projects | Runs a question across all four bubbles |
| Senior advisor | Former official, senior fellow, or distinguished expert | A distinguished record and no conflicts on the question | Stress Testing; Rapid Translation and Action |
Your roster
Add researchers, subject matter experts, and peer auditors along with their areas of expertise. Then assign them to each bubble of a question from the Research tab. People whose expertise matches the question are suggested automatically.
Add a person
Roster
Demo: watch a question move through the model
Press Play to start the demo
It takes about a minute. You can pause or restart at any time.
Market price vs. bubble confidence
Sample market for illustration only. This is research, not trading advice.
What the client sees
Updates appear as the teams work.
- No updates yet.
Latest bubble package
Each bubble's package appears here when it pops.
About Bubble Forecast Intelligence
A conceptual model for human-led, AI-enhanced research and forecasting, designed around displaced RAND researchers and other experienced analysts.
Who we are
- Parent organizationBubble Research Group™
- Research unitBubble Forecast Intelligence LabResearching forecasting systems, prediction markets, human-AI intelligence, information flow, and decision support.
- TechnologyBubble Forecast Intelligence™The research and forecasting system shown in this concept, which moves each question through four bubbles of human experts.
In short
- Each research question passes through four bubbles of human experts.
- All four teams can work at the same time, but the bubbles close in order.
- A bubble pops only when it has recorded its findings and they have been checked.
- People lead, and AI assists. AI helps under human direction, and every use of AI is disclosed.
1Why this model exists
AI can now summarize published research in minutes. But speed is not the same as understanding. AI works from what has already been written down, and much of what matters in a real decision never is.
Bubble Forecast Intelligence is built on a simple idea: the best research and forecasting come from people with real experience, made faster and sharper by AI they control.
2Human-led, AI-enhanced
People lead every step. AI makes their work faster and more thorough, but it never has the final say.
What people bring
- Lived experience
- Knowing a community, a region, or an institution from the inside, not just from documents.
- Intuition
- Sensing when something does not add up, even before the data can show it.
- Expertise and judgment
- Years of research practice, and knowing which evidence to trust.
- Trust and access
- People talk to people. Interviews and fieldwork reach what is never published.
What AI adds, under human direction
- Speed
- Sorting large amounts of data, notes, and documents in minutes.
- Breadth
- Searching far more sources than a team could read on its own.
- A second check
- Looking for evidence that contradicts the team's conclusions.
- Drafting help
- Turning findings into clear first drafts that people then edit.
People stay in charge. They choose the questions, direct the AI, check its output against the original sources, and sign every conclusion. Each bubble records whether AI was used, what it did, and whether a person checked it.
3How a question moves through the model
Every question moves through the same four bubbles, from top to bottom.
4What every bubble records before it pops
So that nothing is lost between bubbles, each one passes on a short summary with four parts:
- Current consensus
- What the bubble agrees is true right now.
- Unresolved anomalies
- What does not fit or is not yet explained.
- Confidence rating
- How confident the bubble is in its findings.
- Next required skill set
- The expertise the next bubble will need.
In this concept version, each bubble also lists its sources, can attach images, says whether AI was used, and needs sign-off from named auditors.
5Why it beats AI-only research
- It is grounded in the real world
- AI cannot interview people, judge who can be trusted, or pick up on local issues that never appear online. People in the field can.
- It brings different minds together
- Passing a problem through separate, specialized bubbles avoids both a single analyst's blind spots and the flat, average answers of AI models.
- It is fast and accountable
- Like a medical shift change, each bubble passes on a structured summary, so speed does not come at the cost of accuracy.
6Built on proven ideas
Bubble Forecast Intelligence is new in how it combines these pieces, but each piece is already well established.
- Requests for Information (RFIs)
- Government agencies regularly publish RFIs to gather outside expertise before they make decisions. For example, in 2025 the White House Office of Science and Technology Policy issued an RFI to shape the national AI research and development plan (details). Bubble teams can prepare fast, evidence-backed RFI responses, and the lab can issue its own RFIs to find partners, funders, and researchers.
- Prediction markets such as Polymarket
- Prediction markets already turn many people's judgments into live probabilities. Bubble Forecast Intelligence works alongside them: a market price is one input, and the bubbles explain why the odds are where they are.
- Human forecasters assisted by AI
- Research on forecasting shows that people perform better with AI help. One study of 991 forecasters found a 24 to 28% accuracy gain when people had an AI assistant, and top human forecasters still beat AI systems on their own as of May 2026 (Metaculus research summary).
7Where the idea came from
The model grew out of Thought Relay, a game in which one person proposes a bold idea, a second connects it to a real-world system, and a third tests it against real limits. Bubble Forecast Intelligence turns that relay of different minds into a disciplined research process.
8About this concept version
This is an early concept, not a finished product. The demo on this site shows how the model would work. A simple prototype is used for internal testing, and the full platform will be built once funding is in place. Features and wording may change as the concept develops.
Contact us
Interested in research collaboration or funding, have a research question, or want to join the lab as a researcher or auditor? Send us a message and we will get back to you.
Research Built to Move
Have a Question?
Tell us about your project and we'll respond within one business day.