The Daily Ubuntu · dailyubuntu.com · Printable sheet
Thread 02 · Technology and Dehumanization · September 25, 2026 · 2 min read
Measuring illness in dollars
A U.S. health algorithm judged patients' needs by their past spending. Black patients, who received less care, therefore seemed less sick.

The Story Granary
It was in the waiting room of a clinic in Montréal-Nord.
Ginette, a patient care attendant, was chatting with the woman in the next seat. She hadn't seen a doctor in four years. Two jobs, no time, no family doctor.
"At least I don't cost the system much," she said with a laugh. Her neighbour didn't laugh.
(This scene is invented. What follows is not.)
Costing little is not the same thing as being well. Yet one algorithm confused those two things.
On October 25, 2019, the journal Science published the study by four researchers: Ziad Obermeyer, Brian Powers, Christine Vogeli and Sendhil Mullainathan.
They had examined a commercial algorithm widely used by U.S. health systems. Its job: identify patients with complex needs, to offer them extra follow-up.
According to the researchers, this kind of tool influences decisions for more than 100 million Americans.
They analyzed the records of 43,539 white patients and 6,079 Black patients at a large academic hospital.
Their finding: at the same risk score, Black patients were markedly sicker.
Why? The algorithm predicted future health spending. Not illness.
But because of unequal access to care, less is spent treating Black patients.
Lower past spending created the illusion of lower needs.
Picture measuring someone's thirst by the number of glasses they have already been served.
The person who was forgotten will always seem less thirsty.
Correcting this bias would raise the share of Black patients receiving extra help from 17.7% to 46.5%.
The researchers did not stop at the finding. They worked with the manufacturer, which confirmed their results.
Together, they showed that by taking actual health needs into account, racial bias could be reduced by 84%.
One of them, the economist Sendhil Mullainathan, put it this way:
“Algorithms by themselves are neither good nor bad.”
The machine was not malicious. It had learned from an unequal world.
A society's numbers tell how it has cared for its own. Ubuntu tells us that we exist through one another: when researchers and a company look at the problem together, correction becomes possible, for all of us.
A number does not only say what a person needs. It also says what they were given, or denied.
Today, somewhere, someone seems to "cost little." Maybe someone should ask them how they are.
And you?
In our own numbers, at work and in our public services, what truly measures a person's needs, and what only measures what they have already been given?
My answers:
Figures
Sources
- Z. Obermeyer, B. Powers, C. Vogeli and S. Mullainathan, "Dissecting racial bias in an algorithm used to manage the health of populations," Science, vol. 366, October 25, 2019. escholarship.org
- University of California, Berkeley, "Widely used health care prediction algorithm biased against black people," October 24, 2019. news.berkeley.edu
- Princeton University, RRAPP, "Algorithms Can Replicate or Remedy Racial Biases in Healthcare Resource Allocation." rrapp.spia.princeton.edu
Every figure on this page links to a verifiable public source.
Tags
Read also
Lend Me a Hand
In 2011, a South African carpenter lost fingers in an accident. With a stranger in the United States, he invented a mechanical hand, then gave it away to the whole world.
ReadPutting the Forgotten on the Map
In February 2016, in Paris, about sixty volunteers traced more than 4,000 buildings in Uganda in three hours. One evening at a computer can help bring care.
ReadThe Phone Connected to the Wind
In Japan, in a garden, an old phone is connected to no network. Since the 2011 tsunami, thousands of people have come to talk there to those they lost.
Read