Same Technology, Different Master
AI as liberation technology, and a patient’s case for building tools on our side of the table.
Hugo Campos · Porto · July 8, 2026
The same technology can deepen compliance or widen agency.
Health care has a very old script for patients: leave it to the experts. In 1847, the American Medical Association could still describe patients as needing “condescension with authority” and “reasonable indulgence” for the “mental imbecility and caprices of the sick.”
Today’s AI systems may sound kinder than that. But if they are deployed by institutions, optimized for institutional goals, and pointed at patients, kindness is not enough. Patient-facing is not the same as patient-directed.
The question is not whether AI is accurate, safe, efficient, or empathic. The question underneath is: who is it serving, and for what purpose?
Patient-facing. Patient-aligned. Patient-directed.
It is aimed at you.
A portal message, chatbot, reminder, or risk score may be directed toward the patient while still serving the institution’s workflow.
It considers you.
The tool may account for your needs, preferences, and constraints, but you may still not control its goal or inspect its logic.
It works under your direction.
You choose the question, steer the tool, inspect the output, redirect it, refuse it, and bring the result back to care on your terms.
OpenKP is not a live demo. It is proof of a shift.
I did not build OpenKP because Kaiser gave me a better portal. I built it because the portal could not answer my questions. I pointed an AI coding assistant at my own records, my own goals, and my own need to see the pattern across institutional fragments.
The important part is not that I wrote Python. I did not. The important part is that a patient could direct an AI system to build a tool the institution had no reason to build for him.
Read through my visit notes from the last two years. Find every time I raised a concern, asked a question, or pushed back. Show me how each one was documented, and look for patterns in how my engagement gets characterized.
This is not asking AI to be a doctor. It is asking AI to help a patient read the system around his care.
A new health literacy for an AI-mediated system.
For decades, health literacy often meant understanding enough to comply. Critical AI Health Literacy asks for more: the capacity to use AI to understand power, data, incentives, and decision-making in health care, then act with more agency.
Where is my record?
What data exists, what is missing, who controls it, and how can I make it useful for my own questions?
What can I use?
Which AI tools are available to me now, and how do I judge their limits without surrendering their usefulness?
Who sets the goal?
Can I inspect, redirect, contest, refuse, and aim the system at outcomes that matter to me?
Design for the patient side of the table.
- Build tools patients can direct, not only tools that direct patients.
- Make institutional AI contestable, inspectable, and legible from the patient’s side.
- Treat cost as a design constraint. The most powerful patient-directed AI cannot become a luxury tool for the already resourced.
- Design for the person who is scared, tired, medically literate only because they had to become so, and still trying to ask a better question.
Paulo Freire wrote that liberation is praxis: reflection and action upon the world in order to transform it. Patient-directed AI belongs in that lineage: not AI that acts on patients, but AI patients can use to reflect, act, and change the conditions of their care.
Links from the talk and nearby work.
Critical AI Health Literacy paper
NAM Perspectives DOI.
CAIHL.org
Framework home.
OpenKP
Patient-directed access to the Kaiser record.
Josh Mandel’s Health-Skillz
Open health data skills for AI assistants.
Ryan Hughes / OpenRecord
Open-source patient-directed health record work from Fan Pier Labs.
Nature Medicine benchmark
Clinical AI agents benchmark cited in the talk.
Gallup / West Health survey
U.S. survey on AI use before or instead of health care visits.
AI Patients
Related patient-led AI work.
HugoScore
One of Hugo’s patient-directed projects.
myHeartData
Patient access and patient-generated health data work.
Hugotronic
Personal technology and cardiac-device work.
Endothelial
Patient-directed health project.
Your Healthcare Agent
Health AI agency project.
Thank you.
For questions, collaborations, or follow-up:
Contact me
ResearchPartner.org · Hugo Campos