Bayesian Meets Politics: When Evidence Stops Constraining Power

March 22 2026

The Ruling

On March 16, 2026, a federal judge did more than block a set of vaccine recommendations.

He exposed the flaws in logic of how public-health decisions are made.

In ruling that actions taken by Health and Human Services Secretary Robert F. Kennedy Jr. were likely unlawful and “arbitrary and capricious,” the court invalidated a newly appointed Advisory Committee on Immunization Practices (ACIP). It canceled its upcoming meeting, and erased its recent votes, including changes to routine childhood immunizations such as hepatitis B and combination measles, mumps, and rubella schedules.

This episode is a case study in what happens when the relationship between evidence and decision-making begins to break down.

The sequence of events was stunning. In June 2025, Kennedy dismissed all 17 members of the standing ACIP and replaced them with new appointees. By January 2026, the reconstituted committee had endorsed revisions to roughly one-third of the national vaccine schedule, ending the universal birth dose recommendation for hepatitis B, altering combination-vaccine guidance, and revisiting core elements of routine childhood immunization. These were significant changes. ACIP recommendations determine insurance coverage, shape clinical practice, and often define what is required for children to attend school.

Medical organizations, including the American Academy of Pediatrics, challenged these decisions and the process behind them. Their argument was the system designed to evaluate evidence had been bypassed. The court agreed, concluding that the restructuring of the advisory process likely violated the Federal Advisory Committee Act, a law meant to ensure that federal advisory bodies are balanced, transparent, and grounded in expertise.

Bayesian Meets Politics

To see why this matters, it helps to understand what that system is supposed to do.

ACIP is an institutional mechanism for slowing decisions down until they are justified. Its members review randomized trials, post-market safety signals, including data from the Vaccine Adverse Event Reporting System, and large observational studies. They meet in public, debate uncertainties, record votes, and explain their reasoning.

This layered structure reflects a deeper logical process, one that statisticians formalize as Bayesian updating.

The idea is simple, but demanding. Belief has to yield to evidence, and it has to do so in proportion to how strong that evidence actually is. Using Bayesian reasoning, policy should begin with what is already known: a “prior” built from accumulated evidence. New information, the “likelihood”, should then be weighed carefully, and policy should change only in proportion to the strength of that new evidence. The result, the “posterior”, is an adjustment anchored in both past knowledge and present data.

In vaccine policy, the "prior" is unusually strong. It rests on decades of clinical trials, continuous safety monitoring, and real-world evidence gathered across millions of administrations. The "likelihood", by contrast, must clear a high bar. Not every new study, signal, or hypothesis is enough to overturn what is already well established. The application of Bayesian updating assumes that evidence must be consistent, replicable, and contextualized.

From this follows a basic Bayesian implication: the larger the policy change, the stronger and more transparent the evidence must be.

Measured against that standard, the recent changes are difficult to justify on evidentiary grounds alone. The 'prior' did not collapse, and the underlying evidence base remained intact. There was no clearly documented surge of new, large-scale evidence, no decisive study or series of findings that would normally justify such a sweeping revision. Yet the policy shifted rapidly and extensively, reshaping a substantial portion of the vaccine schedule in a matter of months. More revealing is how the change occurred. By dismissing the entire advisory committee and installing a new one, the government altered the mechanism that evaluates evidence itself. When that happened, the link between evidence and outcome became unstable.

The Court Reveals The Broken Chain

There is the deeper significance of the court’s ruling. The decision did not hinge on identifying a decisive new scientific finding. It turned on process, on whether the system designed to ensure that evidence constrains policy had been compromised. By calling the actions “arbitrary and capricious” and invalidating the committee and its decisions, the court effectively concluded that the logical chain connecting knowledge to action had been broken.

Supporters of the changes argue that established institutions can become insular, that public trust has eroded, and that alternative perspectives deserve consideration. These claims speak to real tensions in modern public health. But reform requires discipline. Without a stable method for evaluating evidence, policy risks becoming untethered from the data it is meant to reflect. At stake is whether power remains bound by evidence or begins to move independently of it.

Once evidence no longer constrains decisions, it becomes impossible to tell whether policy is following facts or simply following power. The difference has broad and profound implications.