The Pole Does Not Lift You: A Physics Lesson for AI in Public Health

In ninth grade, I signed up for the pole vault at our middle school track meet. I had never done it before. But I was signed up, so I went to prepare. There was no YouTube back then — no tutorial videos, no slow-motion breakdowns, no coaches on my phone. I borrowed a bamboo pole and practiced on a sand pile at a construction site near my home. My father, worried, came out to be my coach.

The key, he told me, was to get the pole vertical — to plant it and drive it upright, not let it collapse back toward me. Grip too high without enough speed, and the pole will never straighten. It will fall back on you.

I fell off the sand pile more times than I could count. The night before the meet, I replayed it in my head on a loop: approach, plant, vault. I did not sleep a minute. I was fourteen.

At the meet, six of us had entered. The opening height was 1.60 meters. One by one, the others ran up and crashed down in every imaginable posture. Only one person cleared that bar. I had gripped low — safely low — but I had dared to run, and I came in sideways, a kind of improvised fisherman’s cast, and went over on the first try. The organizers lowered the bar to 1.55 meters to let the others compete for silver and bronze. The falls continued.

The Real Problem: Horizontal Motion

Years later, in college, I majored in physics — and only then did I have language for what had actually happened on that sand pile.

The vault is an energy conversion. Your run-up is kinetic energy (½mv², the energy of motion). The bar is potential energy (mgh, the energy of height). The pole is just the mechanism that converts one into the other — and only if the planting technique is right. Without it, all your kinetic energy stays horizontal. You run fast. You go nowhere up. The pole falls back on you.

Technique is what converts motion into elevation.

From the Sand Pile to the Public Health Team

Now scale this up.

I have sat with public health teams using AI across every step of a community health needs assessment — summarizing focus group transcripts, drafting survey items, synthesizing county-level data, generating stakeholder briefings — and watched them produce twice the volume in half the time, only to arrive at the same recommendations they would have written without AI at all. The activity was everywhere. The elevation was nowhere. The pole kept falling back on them, and they kept running faster.

This is what I want to call it: horizontal motion. Faster activity, same altitude. The instinct, when a team notices this, is to use AI more, deeper, faster. That instinct is wrong, for the same physics reason it was wrong on my sand pile. The problem is not speed. The problem is conversion.

Public health AI does not lack momentum. It lacks the technique to turn momentum into elevation.

The Question Worth Asking

The question I want to ask about AI in public health is not how fast we are running, or how tall our pole is. Those are the easier, noisier questions. The real question is the conversion one:

What is the planting technique for AI? Inside a real public health workflow, what is the specific move that converts all that prompting, querying, and summarizing into a better decision, a clearer insight, a sounder judgment?

I don’t think the planting technique is a single move. In a vault, planting has a moment, an angle, and a depth — all three have to be right together, or the energy doesn’t convert. I suspect the AI version is similar: not one trick, but a small coordinated set of moves performed at the right step of a workflow. Which step, and how deep AI goes at that step, matter more than how fast the team is running overall.

That is the direction I want to take in next week’s Case. For now, the claim I am willing to stand behind is the smaller one: the question worth asking about AI in public health is not how fast we are running. It is where, in the workflow, our kinetic energy is converting into elevation — and where it is not.

The Observer’s Insight

At fourteen, I could not have told you any of this. I just knew the pole kept falling back on me, and I was afraid, and I went anyway. Decades later, physics gave me the vocabulary: I was not failing to run fast enough. I was failing to convert. The energy was there. The conversion wasn’t.

A lot of public health teams are in that exact position with AI right now. The energy is there. The conversion isn’t. And the instinct, as it was mine on the sand pile, is to run faster — when the real move is somewhere else entirely.

The sideways fisherman’s vault was not the right form. But it was the form I had, and I ran at the bar with it, and I cleared 1.60 meters while everyone with better intentions fell. That may be the honest picture of where most of us stand with AI in public health right now — gripping low, running anyway, clearing the bar while others are still tumbling off. The next question is whether we can learn the actual technique, now that we have stopped being afraid of the pole.

Pick one workflow in your own practice where you have been using AI. Where in that workflow is the energy actually converting into elevation? And where are you just running faster without getting any higher? Reply or leave a comment — I would love to hear your story. I will pick this up in next week’s Case.

— Min Wu, PhD, Associate Professor,
Zilber College of Public Health, University of Wisconsin-Milwaukee

As a member of the Springer Nature Author Affiliate Program, I may earn a small commission from purchases made through this affiliated link to my book. Support the author by checking out my textbook, Artificial Intelligence in Public Health: https://tidd.ly/4mH9389