31 models.
100 real consultations.
77 percent errors of omission.
Not wrong answers.
Missing answers.
According to ARISE Network.
The model gave a response that looked complete.
It was not complete.
It chose what to include.
The clinician received what it chose.
Sat in a quality review.
An AI recommendation had been cited in the chart.
The recommendation was accurate.
What it did not say
changed the outcome.
Nobody flagged the gap.
Nobody was looking for what was not there.
That is what 77 percent means.
A wrong answer has a trail.
↳ The clinician pushes back.
↳ The chart records the correction.
A missing answer leaves nothing.
↳ The clinician acts on what arrived.
↳ The excluded differential is never named.
↳ The chart records the decision
as if the full picture was present.
It was not present.
The Accountability Gap™ (TAG™)
does not live in the wrong answer.
It lives in the answer that never arrived.
The named owner is the one
who asks what the model chose not to say.
Most institutions have not named the owner.
You already know which omission
your institution cannot account for.
Name the owner
before the next adverse event does it for you.
Mo Johnson, MD MBA is a cardiothoracic surgeon and the founder of GPe Research. Field Notes are short dispatches from the clinical AI accountability frontier, published alongside the MedicoVigilance™ newsletter at medicovigilance.org.
Follow the work on LinkedIn: linkedin.com/in/mo-johnson