Known-and-Unknown Risk Map
Compare visible risks while reserving space for harms not yet detectable.
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 98%
Hari borrows the known-known, known-unknown, and unknown-unknown distinction to structure uncertainty around newer weight-loss drugs. Known knowns include measured benefits and documented adverse events. Known unknowns are questions already visible but unresolved, such as effects in pregnancy, tolerance, or long-term weight and metabolism after stopping. Unknown unknowns are harms that may require decades of exposure before anyone knows to look for them. The framework does not allow uncertainty to settle the decision automatically against treatment. Hari repeatedly compares drug uncertainty with the well-established long-term risks of obesity and explains that his family history shifted his personal judgment. A sound risk map therefore records evidence, gaps, time horizons, and the untreated baseline, then defines what new evidence would trigger reassessment. Speculation stays labeled rather than being presented as incidence data.
Origin
Hari applies Donald Rumsfeld's uncertainty categories after discussing long-latency effects with researcher Greg Stanwood. Extracted from The Diary of a CEO.
Core principles
- 01Known harms, unanswered questions, and unforeseeable harms are different risk classes.
- 02Long-term absence of evidence cannot be manufactured from short follow-up.
- 03The risk of an intervention must be compared with the risk of doing nothing.
- 04A decision under uncertainty should remain revisable as evidence changes.
How to run it
- 1
Record known knowns
List benefits and harms supported by the best evidence currently available, including their absolute scale where known.
Pro tip Separate a relative-risk increase from the underlying absolute risk.
Watch out A reported association is not automatically proof that the intervention caused the outcome.
- 2
Name known unknowns
Write the questions researchers already recognize but cannot yet answer and identify what study or follow-up would reduce the uncertainty.
Pro tip Include populations poorly represented in existing evidence.
- 3
Reserve for surprise
Identify plausible long-latency domains without inventing a specific outcome or pretending to know its probability.
Pro tip Use comparable historical cases to set monitoring horizons, not to claim the same harm will recur.
Watch out Unknown unknowns justify humility, not unsupported medical claims.
- 4
Map the untreated baseline
Estimate the likely harm of doing nothing or delaying action for the relevant person or population.
Pro tip Apply the same evidential standard to action and inaction.
- 5
Personalize the balance
Account for factors that materially change expected benefit or risk, using qualified professional guidance for high-stakes decisions.
Pro tip Explain why the same evidence can support different decisions for different people.
Watch out The transcript is not a substitute for an individualized medical assessment.
- 6
Set review triggers
Define new evidence, symptoms, life events, or regulatory findings that would cause the decision to be revisited.
Pro tip A decision under uncertainty should include its own update rule.
In the wild
Hari says the long-term risks of the drugs are not fully known, while the obesity-related heart risk in his family is visible and immediate to him. He therefore chose the drug risk over continued obesity for himself, while emphasizing that another person could review the same evidence and reasonably decide differently.
→ The framework produces a conditional, revisable judgment rather than a universal verdict.
Common mistakes
Treating unknown as proven harm
A plausible concern must stay labeled as uncertainty until evidence establishes an outcome and its frequency.
Ignoring the cost of inaction
Focusing only on intervention risk can hide serious established risks in the untreated baseline.
Making the decision permanent
New safety, efficacy, or personal evidence should be able to change the conclusion.
Is it for you?
Best for
High-stakes choices involving new treatments or technologies with meaningful known benefits and incomplete long-term evidence.
Not ideal for
Turning every theoretical possibility into an equal-probability danger or making individualized medical decisions without professional care.
From the transcript
“there are unknown unknowns things we don't know we don't know”
“we don't know the long-term risks of these drugs”
“we do know the long-term risks of obesity”
From the episode
Ozempic Expert: Ozempic Increases This Disease By A Factor Of 9! They're Lying To You About Ozempic Side Effects & What It's Doing To Our Brains! - Johann Hari