First principles · Small decisions
Nobody Voted for That
Large outcomes are decided by the sum of small choices, and the large choice is not on the ballot. AI is about to make the small choices for us, faster than any habit could, and to give the large one its first ballot.
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
- ✓
✓ + ✓ + ✓ + … = ✗
You walk past the bookshop on the corner and the windows are papered over. A handwritten sign thanks the neighbourhood for thirty-one years. Something drops in your chest. Then you try to remember the last book you bought there, and you cannot.
The ballot that never comes
In 1966 an economist at Cornell, Alfred Kahn, wrote about the passenger train that had served Ithaca, New York. It had stopped running. Nobody in Ithaca had decided to end it. Each traveller, on each trip, had simply picked whatever was quickest that day: a car, a bus, a plane. Each choice was sensible. Added together they starved the railroad, and the town lost the one way out that worked in a snowstorm, the one it would have paid to keep had the question ever been put to it.
Kahn called this the tyranny of small decisions. The principle is simple. A large outcome is decided by the sum of many small choices, and in each small choice the large outcome is assumed to stay where it is. You drive to the city this once because the train will still be there next time. Everyone reasons the same way, and next time it is gone. The large question, train or no train, is the one question nobody is ever asked.
The ecologist William Odum later used the same idea to explain how a wetland disappears without anyone voting to drain it: one permit, one ditch, one lot at a time. It explains the bookshop too. You did not choose its closing. You chose, forty times, to have a book arrive tomorrow.
The large decision is made by the small ones. It just never appears on the ballot.
What AI changes
AI changes who makes the small choices. A delegate that shops, books and orders for you makes them faster than you could, at a scale you never would, and with a clean eye for the stated goal: cheapest, nearest, soonest, best reviewed.
The part people miss is what that clean eye removes. For most of history the small choices were made badly. People went to the same pharmacy because it was on the way home. They kept buying from the hardware shop because the owner knew their name. They paid a little more out of loyalty, habit, guilt or laziness. Economists treat those as frictions. They were also a quiet subsidy. Human inefficiency was paying for a great many large outcomes that no one had ever voted for either, among them the corner shop, the local paper and the restaurant you think of as yours.
A delegate does not have that inefficiency unless you give it some. So the first effect is acceleration. The tyranny of small decisions, run at machine speed and machine indifference, will close in a few years what habit would have kept open for another twenty.
The second effect is the one I find hopeful, and it is new. A delegate can carry a standing instruction into every small choice it makes for you. A person cannot. You might decide on a Sunday night that you want the bookshop to survive, and by Wednesday you have forgotten and tapped the button that says tomorrow. A delegate does not forget. Write the large decision down once, in plain words, and it is applied to every small decision that follows.
That is the first ballot the large question has ever had. Kahn’s tyranny was not about bad people. It was about a missing place to say the large thing, and that place is now here. Whether it gets used is up to each of us.
Your delegate shows you the month: forty-one small orders, all handled. One line in the summary is yours, written in January: if the shop on the corner carries it, buy it there, even at a few dollars more. Eleven of the forty-one went to the corner. The sign in its window says it is hiring.
Five predictions
Standing instructions become the most important setting a delegate has.
Within three years, the personal AI products people use for errands will ship with a plain-language page of standing preferences, and it will be the screen people edit most after the first month. The large decisions will live there.
A wave of habit businesses closes first.
The shops that lived on being on the way home will close before the ones people choose on purpose. Their owners will say no customer left them. They will be right. The customers stayed; the errands left.
Small businesses ask to be named, not bought from.
A sign by the till will ask you to put the shop in your delegate’s standing instructions. Being named once will be worth more to a local business than any single sale, because one line written in January decides a year of errands.
Defaults become a policy fight.
The standing instruction a delegate ships with is a large decision made for millions of people at once. Expect it to be argued over the way organ-donor defaults and pension enrolment defaults were, with cities and states asking for a local option.
Towns publish suggested instructions.
Neighbourhood associations and main-street groups will publish short recommended lines for residents to paste into their delegates, the way a voter guide suggests a ballot. The places that do this will keep shops that the places that do not will lose.
Where I stand
I have told my own AI to decide a great many small things for me, because I had become the bottleneck in my own life. That makes me exactly the person this principle is about. I have handed over the small choices. What I have not always done is write down the large ones, and until I do, my delegate is running Kahn’s tyranny on my behalf with perfect efficiency. The small choices were not where my values lived. They were where my values leaked.
The claim ledger
- Already true
- Large outcomes, from rail lines to wetlands to main streets, are routinely decided by the sum of small choices in which the large outcome was assumed to stay put. Delegates that shop and book for people are already in use and spreading. They optimise the stated goal of each errand.
- What has to happen
- Delegates have to carry standing instructions reliably into every small choice, and people have to write them. Businesses have to learn that the thing to win is a line in a standing instruction, not a single sale.
- Where I am probably wrong
- If people leave the defaults alone, the hopeful half fails and only the acceleration remains. And if delegates turn out to learn our old habits so faithfully that they keep going to the same pharmacy out of imitation, the acceleration fails too. I will know I was wrong if, by 2030, neighbourhoods where delegates run most errands keep their small shops at the same rate as neighbourhoods where people still run their own.
The windows are still papered over. You cannot reopen the shop, and you cannot take back the forty taps. You can open your delegate’s settings and write one sentence about the next shop on the next corner.
Write the large decision down. It is the only one nobody else will ask you for.
Background
- Alfred E. Kahn, “The Tyranny of Small Decisions: Market Failures, Imperfections, and the Limits of Economics,” Kyklos, 1966.
- William E. Odum, “Environmental Degradation and the Tyranny of Small Decisions,” BioScience, 1982.
- Who Gets Paid for Your Indifference? That piece is about who sells the default a delegate picks. This one is about what the sum of those picks decides.
- The Wintering Rule, on shared resources. The tyranny of small decisions needs no shared resource; it works inside a single person.
- The Immutable Past