In May 2021, a question from Darren Stewart, a data scientist at the United Network for Organ Sharing (UNOS), revealed the moral complexity hidden inside a seemingly trivial technical detail: how many decimal places should be used when calculating a kidney transplant allocation score. The score determines which patient receives priority for a donated organ. UNOS systems could compute scores to 16 decimal places, but Stewart and colleagues questioned whether extreme precision was appropriate when differences at the 14th decimal place no longer reflected meaningful medical distinctions. “We want to make our decisions as much as possible based on clinical criteria and not flipping a coin,” Stewart said. Yet if two patients are clinically indistinguishable, refusing to randomize may itself be unfair.
This tension—between technical precision and moral responsibility—mirrors challenges across domains where algorithms influence human lives. Resume-screening software must define relevant and fair hiring criteria. Risk assessment tools in courtrooms require thresholds for labeling defendants as dangerous. School assignment algorithms must quantify what makes a “good” student. Often, decision-makers defer these moral judgments to technical experts, who then face ethical dilemmas disguised as neutral calculations.
Organ transplantation stands out for its openness about such dilemmas. Surgeons and data scientists in this field have a tradition of inviting broader participation in ethical decision-making. This culture traces back to the late 1950s, when Belding Scribner’s innovation—using Teflon tubing to enable long-term dialysis—transformed kidney failure from a fatal condition into a chronic one. Dialysis machines were scarce, and Scribner faced the agonizing question of whom to save. His team delegated non-medical allocation decisions to a committee of seven laypeople, selected by the local medical society. Life magazine’s 1962 coverage, headlined “They Decide Who Lives, Who Dies,” described the committee’s biases, which favored male breadwinners and likely white applicants. The model was short-lived, ending when Congress funded dialysis for all eligible patients in 1972, but it demonstrated moral modesty and a willingness to democratize value judgments in high-stakes medical systems.
Today, about 100,000 people in the United States await kidney transplants. Allocation is governed by an algorithm that blends medical data, logistics, and moral considerations. The system incorporates public input through meetings and open comments, explains rules in plain language, forecasts the impact of changes, and undergoes annual third-party audits. This governance approach reflects lessons from Scribner’s era, embedding inclusivity and transparency into algorithm design.
The algorithm’s evolution illustrates how public participation reshapes technical systems. In 2007, an initial proposal to maximize total life-years saved appealed to medical logic but risked worsening racial disparities and disadvantaging older patients. Clive Grawe, a 55-year-old patient, argued against age-based bias at a public meeting in Dallas. Critics noted that wealthier candidates, with better health and access to care, would benefit disproportionately. The plan was abandoned in favor of a balanced approach: the healthiest kidneys go to those likely to live longest, but other patients retain reasonable chances at receiving organs. This compromise improved both total life-years saved and racial equity.
The redesign process spanned a decade, marked by contentious debate and incremental progress. Persistent barriers remain—low-income patients often struggle to join the transplant list, and hospitals may exclude candidates lacking transportation or family support. Structural inequalities still influence outcomes, and while allocation logic is public, UNOS treats the source code as proprietary. Nevertheless, governance techniques such as participation, transparency, forecasting, and auditing have altered the moral framework of kidney allocation, spotlighting certain issues while leaving others less examined.
Public engagement in this context did more than capture opinions; it transformed perceptions of fairness. The infrastructure supporting these discussions enabled stakeholders to grasp trade-offs and implications. Numbers, often seen as neutral, can obscure human consequences, but when presented with clarity and context, they can illuminate the ethical stakes of algorithmic decisions.
As for Stewart’s original question on decimal places, the committees and data scientists at UNOS continue to deliberate—an enduring reminder that even the smallest technical choices can carry profound moral weight.