Kaiser Nurses Say AI Is Now Grading Their Empathy — and Patient Care Is Paying for It

If you had to name the most human job in a hospital, it would probably be nursing. Holding a patient’s hand, reading their face, talking them down from panic. So what happens when an AI starts scoring that work on a dashboard? Nurses at Kaiser Permanente, the largest nonprofit healthcare system in the US, went public with exactly that complaint. The story climbed to 221 points on Hacker News with 140 comments, and the argument underneath it is sharp enough to deserve a close look.

An Algorithm That Grades Compassion

The core issue here isn’t ordinary automation. It’s that an AI has started evaluating the emotional labor of nursing.

According to the nurses, the system tracks call duration. Spend too long with a patient, and it can count against your performance review. One nurse put it plainly: they’re afraid a long call will show up as a bad mark on their record.

See the problem? Some patients need five minutes. Others need thirty before they’ll relax. Explaining a serious diagnosis, or calming a terrified family member, takes time. But the system reduces all of that to a single crude equation: short call equals efficient.

One Hacker News comment nailed it. The gist: anyone who thinks it’s a good idea to have a machine measure how well a human expresses empathy has already disqualified themselves from holding power over other people. Blunt, but it clearly struck a nerve.

What Gets Measured Survives — What Matters Disappears

Management culture has an old maxim: what gets measured gets managed. The dark side of that line is playing out on Kaiser’s floors right now.

An AI monitoring system only sees what turns into numbers. Call length, case volume, handling speed. Those slot neatly into a dashboard. The relief a patient feels, an accurate read of their condition, the clinical intuition to catch a warning sign before it becomes an emergency — none of that shows up as a number.

The result is predictable. Nurses naturally optimize for the behavior that scores well. Wrap up the call, move to the next case, act in whatever way looks good in the data. Real care gets pushed to the back of the line.

Economists have a name for this: surrogation — when an easy-to-measure proxy quietly replaces the actual goal. When that happens in a hospital, the ones who pay the price are the patients.

A Watched Worker Can’t Do Care Work

Go one level deeper. This is bigger than a badly designed performance metric.

The mere sense of being watched corrodes care work itself. Psychology has understood this for decades: monitored people turn defensive. Creative, flexible judgment gets harder. Every moment becomes a calculation of how this will look on the record.

Nursing is, at its heart, a stream of improvised judgment calls. You read a subtle change in a patient, you respond to situations no protocol covers. A surveillance system pushes in the opposite direction — toward doing things by the book, in a way that logs cleanly, that avoids risk. Defensive medicine, one call at a time.

One HN commenter dryly flagged the whole thing as a “mandatory dystopia notification.” A joke, sure, but there’s a bone in it. We’ve seen this world in science fiction plenty of times. The only difference is that this one isn’t fiction — it’s a real hospital.

Why Hospitals, and Why Now

There’s a reason this trend shows up so sharply in healthcare.

Medicine runs on a chronic labor shortage. Nurse staffing is perpetually thin, and payroll is the single largest cost in running a hospital. From an executive’s chair, an AI monitoring tool looks tempting: extract more output from fewer people, and prove that output in numbers.

But there’s a fundamental contradiction here. Care doesn’t improve by going faster and doing more. If anything, time and slack are what determine quality. Trying to graft manufacturing’s efficiency logic onto a hospital ward — that’s the collision happening at Kaiser right now.

Note who’s driving the pushback: the frontline workers themselves. The people closest to the technology are the ones saying this is wrong. Their testimony is a warning worth reading for any industry weighing an AI rollout, not just healthcare.

The Line Worth Holding

Efficiency isn’t the villain here. The question is what you choose to make efficient. The Kaiser nurses are forcing a basic one: can a machine really put a score on work whose whole point is human judgment and human empathy?

Something being unmeasurable doesn’t make it unimportant. Usually it’s the opposite — the things that matter most live outside the numbers. Odds are your own workplace has moments where the work-for-the-score and the work-that-matters quietly pull apart. As AI surveillance spreads, that gap is exactly the line worth deciding not to cross.