No. 16 Workshop
Say how sure you are.
A hedge is not a confidence level. Quantify instead of hedging, sort claims into bins, and find the step in your argument most likely to break.
If you don’t say how sure you are, they will assume you are surer than you are.
“Likely” means different numbers in different heads
Before you look
A report says a supplier delay is “likely”. What probability do you hear?
60%
The dashed span shows, roughly, how widely people read the word. Intelligence analysts in the 1960s read the same estimative words as wildly different probabilities. A large public survey in 2018 found the same spread. The solid band is the fix the climate scientists adopted: in IPCC reports, “likely” means 66–100%.
In one risk register (an invented case, but a familiar one), the project lead read “likely” as about 60%. The finance lead read it as near certain and froze a payment. Writing “about 60%” would have cost three words.
Numbers barely dent trust. Vague caveats cost more.
Five experiments, including a field test on the BBC News website, 5,780 people in all. Stating uncertainty as a numeric range caused only a small drop in trust in the number and its source. Verbal caveats did more damage.
The fear that admitting uncertainty costs credibility is mostly misplaced. Vagueness costs more than honesty.
× Hedged
“The programme may potentially have had some positive impact on attendance.”
- Every word a hedge
- Nothing a reader can act on
✓ Calibrated
“Attendance rose 3 to 5 points in the six pilot schools.”
- Emerging: one term, no comparison group
- A second term with a comparison would settle it
The second version admits more doubt than the first, and it is the one a reader can trust.
Sort every claim into one of three bins
1Established
Replicated and reviewed.
Holds across settings. Safe to build on.
2Emerging
Early and promising.
Limited samples. Watch it and test it.
3Extrapolated
Reasoned from elsewhere.
Untested here. Treat it as a bet.
Say in the text which is which. Readers can then disagree with your bets without throwing out your science.
Most claims aren’t false. They are inflated, one adjective at a time.
“Improved” becomes “transformed”. “In two schools” becomes “in schools”. A small, conditional finding gets sold as a large, general rule. Numbers over adjectives is the cheapest defence.
Underline the warrant: it is the part that has to travel
Stephen Toulmin split an argument into parts: the claim, the grounds for it, and the warrant that links the two, plus a qualifier and the conditions under which it fails. Most arguments at work state the grounds and the claim and leave the warrant unsaid, because the writer finds it obvious. It is also where claims break when they move to a new context.
Predict before you look · an invented vendor
“Schools using our platform raised maths scores by 12%.” What is the first question to ask?
Grounds40 schools that chose the platform raised scores.
WarrantWhat worked where schools chose it will work where they didn’t.
RebuttalUnless the schools that chose it were already improving.
If the vendor can’t answer, the claim is still a hypothesis.
One more distinction saves a lot of arguments. A knowledge claim (“abandonment peaks at the rubric step”) is settled with evidence. A value claim (“teacher time matters more than feature breadth”) is settled by deliberation. It is a choice, and disagreement about it is legitimate. Papers that blend the two invite the wrong fight: people attack the evidence when what they object to is the values.
Take this with you
- Quantify, don’t hedge. A range costs less trust than “may potentially”.
- Bin every claim: established, emerging or extrapolated. Say which in the text.
- Find the unstated warrant in your argument. That is where it will break.
Say what you are sure of, how sure, and what would change your mind.
Try this week
Calibrate one document.
- Mark three claims in a draft as established, emerging or extrapolated.
- Replace every probability word with a number or a range.
- Lay out your weakest claim as grounds, warrant and rebuttal. Underline the warrant. Add the sentence that would change your mind.
To mull over
- When did you last hedge to protect yourself rather than to inform the reader?
- Which disagreement on your team is really about values but is being argued as evidence?
- What is the warrant under your own team’s favourite success story?
Sources
- Kent, S. (1964). Words of estimative probability. Studies in Intelligence, 8(4). Mauboussin, A., & Mauboussin, M. J. (2018). If you say something is “likely,” how likely do people think it is? Harvard Business Review.
- Mastrandrea, M. D., et al. (2010). Guidance note for lead authors of the IPCC Fifth Assessment Report on consistent treatment of uncertainties.
- van der Bles, A. M., van der Linden, S., Freeman, A. L. J., & Spiegelhalter, D. J. (2020). The effects of communicating uncertainty on public trust in facts and numbers. PNAS, 117(14).
- Toulmin, S. (1958). The Uses of Argument. Novak, J. D., & Gowin, D. B. (1984). Learning How to Learn.