> ## Content Index > Fetch the complete content index at: https://ai-public-health.com/llms.txt > Use this file to discover other available public pages before exploring further. # The Dignity Audit: Who Holds the Authority When AI Protects You? - URL: https://ai-public-health.com/the-dignity-audit-who-holds-the-authority-when-ai-protects-you/ - Published: 2026-06-15T14:00:19.000Z - Updated: 2026-08-15T16:14:20.000Z - Author: Min Wu - Tags: Public health, AI, AI Ethics, Health Technology, Digital Health **Case of the Week,** Min Wu, PhD · ai-public-health.com *When my child started college last fall, I kept the location sharing turned on. We never said anything about it. We both knew it was there.* *For years, tracking the phone felt like parenting. My child was in high school. I needed to know they got home safe. But now they’re eighteen, living in a dorm, making their own choices — and I’m still watching a blue dot move across campus at 11 PM on a Tuesday.* *One evening I caught myself checking. They were at the library. I felt relieved. Then I felt something else: uncomfortable. Not because the technology failed, but because I realized I was using a tool designed for protection in a way that had quietly become control.* *I turned it off the next morning. They probably never noticed. But I noticed.* Experiences like this feel private — a parent’s small decision at a kitchen table. But the same tension is about to become very public, because the tools we use to protect people are getting much more powerful. If you work in public health, you have already felt this. Maybe you helped design a contact tracing system during COVID and watched people stop answering the phone. Maybe you run a community health program and know the difference between a wellness check that reassures and one that feels like surveillance. Maybe you are a student learning about digital health interventions and wondering: how many reminder texts is too many? These are not hypothetical questions. They are workflow decisions that land on someone’s desk every week. ## The Real Problem: Protection That Drifts into Control As AI systems move deeper into health and daily life — tracking medications, monitoring fall risk, nudging people toward cooling centers during heat waves — the same tension scales up. The institution wants to protect. The person wants to be left alone. Technology sits in the middle. I have watched this tension move through three eras. In the paper era, protection meant a public health nurse at the door — high effort, low reach, the visit itself signaling care. In the digital era, registries and reminder calls scaled the reach but flattened the signal. In the AI era, the reach is near-total and the signal is ambient — a text, a nudge, a follow-up that arrives whether the person wanted it or not. Each shift made protection easier to deliver and refusal harder to register. In public health, we are building AI agents that can identify a high-risk resident during a heat emergency, offer a ride to a cooling center, and follow up if the person declines. That capability is genuinely lifesaving. But what happens when the resident says no — and the system keeps calling? This is the central design question for every AI system that serves individuals on behalf of institutions. And it turns on a word most technologists skip past too quickly: dignity. ## Two Dignities That Pull in Opposite Directions When I tracked my child’s location, I was exercising *dignity-as-protection*. I cared about their safety. That care was real. When they silently tolerated it but probably wished I’d stop, they were asserting *dignity-as-autonomy*. Their right to move through the world without being watched. That claim was also real. Most AI ethics conversations treat dignity as a single thing. It is not. The parent and the teenager feel this tension in their bones. So, does every public health system that must decide: do we call the high-risk resident again, or do we respect the refusal? Now scale this up. Imagine a health department during a heat emergency. An AI agent has identified two hundred elderly residents at high risk. It has offered each one a ride to a cooling center. Forty said no. The system has the capability to keep calling. But should it? Who decides the limit — the engineer who built the system, the health officer running the response, or the resident who already said no? ## When Both Sides Are Right, the Architecture Decides **Pure user authority** — where the individual controls everything — abandons the institution’s protective mission. A resident who opts out entirely during a heat emergency may die. We cannot design a public health system that simply walks away when someone says no. **Pure institutional authority** — where the system decides how much contact is appropriate — becomes unchecked monitoring. It tracks, it nudges, it follows up, and the user’s only option is compliance or refusal, with no power to shape the terms. This is what my location tracking had quietly become: all protection, no authorship. Neither pattern works because neither lets both sides be right at the same time. The question is not which side wins, but whether the system’s architecture can hold both. ## The Observer’s Insight I turned off my child’s location sharing not because tracking technology is bad, but because the architecture of our arrangement had drifted. What began as mutual had become one-sided. I held all the authority. They held none of the authorship. That drift is exactly what happens when AI systems designed for protection operate without user-authored boundaries. Technology is not the problem. The authorization architecture is. In my current research, I am developing a governance pattern where the user holds primary authority over what the institution may do, while the institution retains a bounded protective role that operates only through pathways the user has specified in advance. The structure is still being tested across cases, and I will share more — including a name — as the work matures. But the core principle is already clear: the protection must be real, and it must flow through a channel the user has sanctioned. **The dignity audit for any AI system starts with one question: whose authority are we operating under? If the answer is only the institution’s, something has already gone wrong.** I hold one uncertainty openly: this framing works well for competent adults who can specify their own boundaries. Whether it extends to cases involving minors, individuals with advanced cognitive decline, or acute psychiatric contexts is an open question I have not yet resolved. Naming what you do not know is part of the work. *Have you ever been on either side of this tension — as the protector or the protected? What would it have meant to have a say in how far the monitoring could go? Reply or leave a comment — I’d love to hear your story.* --- *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](https://tidd.ly/4mH9389?ref=ai-public-health.com)