What we do not know
Improving only what is already known makes the known smaller and the confidence larger. So the loop also asks what is missing, what would show each entry wrong, and what nobody has looked at, and publishes the answers beside each entry.
Where it stands
- 122open known unknowns and challenges
- 0closed, each with a recorded reason
- 11%of entries red-teamed
- 122net new questions in the last run
| Region | Open | Closed |
|---|---|---|
| Known unknowns | 95 | 0 |
| Challenges from the red team | 26 | 0 |
| How much has not been examined | 1 | 0 |
Accounting: 122 gaps entered, 122 open, 0 closed with a reason. The books balance: nothing has dropped out unrecorded. Last run 2026-11-17 15:33:46. As JSON
New questions the loop generated from the gaps
Each of these came from a gap, not from a list written by hand. They are leads for the loop to work on, never findings.
- A question: The studies behind "Clean-fuel stoves (the largest trial found no clear pneumonia benefit)" disagree. What explains it (population, dose, delivery, period)?
- A question: The studies behind "School-based deworming (evidence is contested)" disagree. What explains it (population, dose, delivery, period)?
- A question: The studies behind "Subsidise grid connections for rural households (benefits are smaller than hoped)" disagree. What explains it (population, dose, delivery, period)?
- A problem to look into: No reviewed solution is listed for "Children whose birth is never registered". Which interventions have trial or review-level evidence?
- A problem to look into: No reviewed solution is listed for "People offline". Which interventions have trial or review-level evidence?
- A problem to look into: No reviewed solution is listed for "Mothers dying in childbirth". Which interventions have trial or review-level evidence?
- A problem to look into: No reviewed solution is listed for "Unsafe sanitation". Which interventions have trial or review-level evidence?
- A problem to look into: No reviewed solution is listed for "Tuberculosis". Which interventions have trial or review-level evidence?
- A question: For problem air, A regulated pollution market with continuous monitoring work(s) while Clean-fuel stoves (the largest trial found no clear pneumonia benefit) show(s) mixed or no effect. Is the difference real or an artefact of context?
- A question: For problem education, Teaching at the Right Level: group by skill, not grade work(s) while School-based deworming (evidence is contested) show(s) mixed or no effect. Is the difference real or an artefact of context?
- A problem to look into: Every solution listed for "Adults without a financial account" rests on weak designs. What does the strongest evidence say?
- A problem to look into: Every solution listed for "Early death from heart disease, cancer, diabetes and lung disease" rests on weak designs. What does the strongest evidence say?
How it works
Three kinds of ignorance need three responses. Known unknowns (a missing source, an untested assumption, a disagreement between studies, an estimate with no uncertainty range, a problem with no solution listed) are worked on directly. Unknown unknowns are probed for: a red-team pass challenges each entry, an inversion asks what would be true if the opposite held, and a pre-mortem imagines the first test failing. Ignorance about ignorance is measured: the share of entries nobody has challenged yet is itself a gap.
Every challenge names what would change our mind. A gap is closed only with a recorded reason and evidence, and it re-opens if the condition comes back. If several runs in a row find nothing new, the next run widens, not narrows.
Today the challenges come from a fixed set of rules written from the audit questions below. A language model may add more, from a different model family than the author, and anything it adds is marked as unverified.
The audit questions
The method is the Universal Ignorance Audit (Quni-Gudzinas, Universal Ignorance Audit, doi 10.5281/zenodo.21901984): five phases, fifteen questions. "We do not know" is an acceptable answer; skipping is not.
- Surface the structure What are the hidden assumptions and scaffolds holding this situation or belief up?
- Surface the structure Where might the map be mistaken for the territory?
- Surface the structure What is the wobble? Where is the tension, anomaly, or thing that does not fit?
- Stress-test the frame What if the opposite were true?
- Stress-test the frame What would a world look like in which this was false?
- Stress-test the frame If I changed the base, origin, center, or frame, what would remain the same?
- Multiply perspectives How would this look to a radically different observer?
- Multiply perspectives Who knows something about this that I do not, and what would they say is my biggest blind spot?
- Uncover the hidden forces Who benefits from the current framing? Who or what is silenced, excluded, or harmed?
- Uncover the hidden forces What is the most dangerous question I could ask about this, the one that threatens my certainties?
- Uncover the hidden forces What does this confusion or ignorance feel like, and what does the feeling itself know?
- Act What do I already know but am pretending not to know?
- Act What can I do with this uncertainty right now, without needing to resolve it?
- Act What does the unknown want from me?
- Act What question am I not asking that I should be asking, given all of the above?