No. 05 Essay
Take the option away.
Fixes that remove the possibility of failure hold up. Warnings, labels and reminders mostly fade. How to tell them apart.
Good design removes an option. Weak design asks people to be careful.
That one distinction does more sorting than any framework in the field. It explains why some famous safety measures worked and others didn’t. It also reflects badly on a lot of design work, which is worth saying out loud.
Two fixes for the same accident
Predict before you look
In an operating theatre, oxygen, nitrous oxide and other gases arrive through outlets that look broadly alike. Connect the wrong one and you can kill somebody. Separately, many drug names look alike on a label. Two fixes:
A The connector
Give every gas a differently shaped fitting
- The wrong cylinder physically cannot attach
- In the standards for decades
B The label
Capitalise the letters that differ
- predniSONE beside predniSOLONE
- A tired pharmacist might notice
Which one has a body of controlled studies behind it?
The connector has no trial behind it and never will. It doesn’t need one. A fitting that doesn’t fit can’t be attached to the wrong cylinder. That is logic, not statistics. The lettering has controlled reading studies, and thin, mixed evidence that it prevents real dispensing errors.
Hold on to that inversion. It comes back in the next piece.
One removes the possibility. The other improves the odds.
Same hospital, same accident, same profession designing the fix. Only one of them works whether or not anybody is paying attention.
The pattern holds across medicine, aviation, public administration and personal finance, which is unusual for anything in this field:
- Take an option away, or take effort away, and the effect holds up when it is tested properly.
- Add a warning, a label, a training course or an encouragement, and the effect mostly shrinks under scrutiny, and sometimes vanishes.
Four hundred and sixty crashes, filed as pilot error
In 1947, two research psychologists went through US Army Air Forces reports. Pilots kept pulling up the landing gear instead of the flaps on landing. The two levers sat side by side and felt identical. Somebody shape-coded them: a small wheel for the gear, a wedge for the flaps, so a gloved hand could tell them apart without looking. The convention is still in the standards. The pilots were never the problem.
Defaults move people. Incentives barely do.
When a large American employer switched its pension scheme from opt-in to automatic enrolment, participation more than doubled. Most people never changed the rate or fund they had been placed in. The default didn’t persuade anybody. It simply became what happened. A study of Denmark’s entire saving population found about 85% of people never actively decide anything about their pension. Cutting the tax break on saving, the standard economic lever, moved almost nobody.
Warnings train people to ignore them
20,802
surgical patients randomised to pulse oximetry or not. No difference in deaths or serious complications.
49–96%
of hospital prescribing alerts overridden by clinicians, across published studies
The alert figure is the more damning one. The warnings fire so often, and so often irrelevantly, that clicking past them becomes automatic. A safety feature has trained its users to ignore it.
Hawaii, January 2018: the alert software listed the internal test message and the live missile alert in the same dropdown, with the same content-free confirmation for both. The alert went to every phone in the state. It took 38 minutes to retract. An “Are you sure?” was there. It didn’t help.
And the honest size of a nudge
Two government teams, one American, one British, ran 126 randomised trials on redesigned letters, forms, reminders and defaults, reaching more than 23 million people. The effects were roughly six times smaller than those in academic journals. If 40 people in 100 were doing the thing before, about 41 did it after. Journals publish the experiments that worked, and small experiments produce wild numbers.
Sort these fixes
Each one is a real kind of fix from a learning platform. Does it remove the possibility of the failure, or ask somebody to be careful?
Remove, or ask to be careful?
- Autosave a teacher’s work every few seconds Losing work by forgetting to save is no longer possible.
- A banner reminding teachers to save before they leave It improves the odds that someone remembers. By week three, nobody reads it.
- An “Are you sure?” before deleting a class Confirmations get clicked through the same way alerts do. Hawaii had one.
- Keep deleted classes restorable for 30 days Permanent loss by accident becomes impossible, whoever clicks what.
- A training video on how to tag resources properly Training adds information. It helps whoever watches it, for as long as they remember it.
- Fill class lists from the timetable instead of asking teachers to enrol students A default. The forgotten-student failure can’t happen because the step is gone.
What did you make impossible?
This is the question to ask of any fix, any designer and any year of your own work. A strong answer names a specific failure that used to occur, describes the change that removed the possibility of it, and can say how anybody knows. A weak answer describes something that looks better than it did before.
It takes about a minute to ask. Unlike a portfolio, the answer can’t be assembled the night before.
Take this with you
- Fixes that remove an option, or remove effort, hold up. Fixes that add information mostly fade.
- Warnings and confirmations that fire too often train people to ignore them.
- When someone proposes a fix, ask whether it removes the possibility of the failure or only improves the chance someone spots it. The second kind isn’t worthless. It should never be the only thing on the table.
Good design doesn’t ask people to be careful. It makes being careless harmless.
Try this week
Audit your last five fixes.
- List the last five problems your team fixed, and how.
- Mark each fix R (removes the possibility) or C (asks someone to be careful).
- For every C, sketch an R that would have done the same job. Even if you can’t ship it, you now know what it would cost.
To mull over
- When is removing an option the wrong move: when does it take away something legitimate people need?
- How many of your product’s warnings would anyone notice if you deleted them tomorrow?
- Most of what design teams produce (personas, journey maps, workshops) produces information and alignment, not constraints. Where does that work earn its keep?
Sources
- Fitts, P. M., & Jones, R. E. (1947). Analysis of factors contributing to 460 “pilot-error” experiences. Aero Medical Laboratory.
- Madrian, B. C., & Shea, D. F. (2001). The power of suggestion. Quarterly Journal of Economics, 116(4). Chetty, R., et al. (2014). Active vs. passive decisions. QJE, 129(3).
- Møller, J. T., et al. (1993). Randomized evaluation of pulse oximetry in 20,802 patients. Anesthesiology, 78(3). van der Sijs, H., et al. (2006). Overriding of drug safety alerts. JAMIA, 13(2).
- Federal Communications Commission (2018). Report on the false emergency alert issued in Hawaii.
- DellaVigna, S., & Linos, E. (2022). RCTs to scale: Comprehensive evidence from two nudge units. Econometrica, 90(1).