How I designed the UX strategy for AI systems that recognize emotions and behavior in healthcare triage — in 2017, before the generative AI wave.
Dr. Sintomas was a healthtech startup using AI to identify medical conditions and guide treatment. The differentiator: AI systems that recognize emotions and human behavior to make triage more accurate and humane.
In 2017, this was academic research, not a commercial product. My role: define the end-to-end UX strategy, build flows for different user segments, and ensure the system respected patient privacy in a pre-LGPD context.
Testing with patients revealed a fundamental tension: users in healthcare contexts became uncomfortable when the system "knew" too much too soon. The AI's accuracy was real, but the perception of surveillance triggered abandonment.
The solution: progressive triage. Simple questions first, emotion and voice recognition as a secondary confirmation layer. The system learned gradually, and users felt in control at every step.
In parallel, we operated with privacy-by-design as a product constraint from day one: every UX decision considered which data was necessary, how it was communicated, and what would never be collected. This was 2017, pre-LGPD, post-GDPR.
"The system understood the patient's emotional state before they finished describing their symptoms. Privacy had to be as careful as precision."
· Users in healthcare contexts are more sensitive to the perception of surveillance. Design must communicate control at every step, not just in the privacy policy.
· Progressive AI: gradually revealing what the system perceives reduces discomfort without losing precision. The timing of "I know" matters as much as the data itself.
· Operating in AI before the hype means navigating without established references. Which makes the work harder and, also, more original.