Health ยท Conversation
Start with a conversation.
For patients, doctors, nurses, and people building tools in care. Questions you can ask. A simple way to talk as a group.
With the Institute for AI Translation. We help fit AI to the people who will live with it.
First
How to have this conversation
This talk goes wrong when it becomes a debate to win, or a lecture. It goes well when people try to understand the same thing, from more than one role.
01
Seek understanding, not victory
You are here to understand each other. If someone wins, the group lost.
02
Listen until you could restate
Before you answer, say back what you heard, in words they would accept.
03
Offer a thought. Then let it go.
Once you say it, it belongs to the group. Drop it if a better idea shows up.
04
Different views help
A second view is not an interruption. Youth, elders, staff, and patients each hold a piece.
05
Leave with a question, not a verdict
Do not end in certainty. Take one question into the week. Come back with what it showed.
In health
Come in as who you are.
Patient or family
You left a visit with questions.
Take these to a doctor visit, a portal message, or a family meeting. You do not need a degree to ask.
BeginDoctors, nurses, clinicians
A wrong score can harm a patient.
Questions for a huddle, a vendor meeting, or a board. Ask to understand, not to prosecute.
BeginBuilding in health
Ask these before you ship.
For teams putting a tool into a clinic, a record, or a daily workflow.
BeginHome and school
A school tool, a chatbot, a child at 2 a.m.
Questions for the kitchen table, a school meeting, or a device that talks back.
BeginHosting a talk
You are bringing people together.
Who is in the room, what to open with, and which talk to start with.
BeginHealth
Clinic, home, and community.
Named problems in care, and a talk for a team that wants to ask before it buys.
Health & wellnessAlready in care
What a family or a clinic is already living with
Triage algorithm bias
Under-triaging minority patients.
25NowAI documentation "ambient scribes"
Accuracy, hallucination, consent.
20NowAI companions and adolescent development
Sensitive-window disruption.
20NowPediatric AI safety
Adult-trained models mis-applied to kids.
20NowSuicide-prevention AI in schools
False positives and privacy.
20NowDiagnostic errors
Missed cancers, misread scans.
20Now
Named concerns
Problems you can point to
Cross-context prompt injection
Attacker exploits shared context between users or sessions.
25D3Many-shot jailbreaks
Exploiting long contexts to shift model behavior.
25D3Autonomous vulnerability discovery
Agents scanning codebases for exploitable flaws.
25D3Autonomous exploit development
LLMs writing working exploits from CVE descriptions.
25D3AI-augmented malware
Polymorphic, target-adaptive malware.
25D3Password and credential-stuffing acceleration
AI generating plausible password guesses.
25D3