Human-Centered AI

Human-centered AI designs systems around the person using them: the human decides, the system supports. For emotion-aware software this is not a slogan — it determines who can see the emotional estimate, who can delete it, and who makes the final call in a health context.

What it means in practice

Three commitments run through the product: consent is explicit and per modality, the derived state is inspectable and deletable by the user, and any health-adjacent output is framed as material for a qualified person to review rather than as a conclusion.

  • Per-modality consent — camera and microphone are separate decisions
  • Inspectable memory — a user can see and delete what was remembered
  • Human-in-the-loop — health outputs are reviewed by a clinician, not acted on autonomously

Human-AI interaction

Emotional context is what lets an interaction adapt: pacing a study session to engagement, softening a delivery when distress is detected, or surfacing that a person hesitated when reporting a symptom. The point is a better exchange between a person and a system, not a more autonomous system.

Human-in-the-loop systems

In the multi-avatar consilium feature, specialist perspectives are shown including where they disagree, so the human reading the output sees the uncertainty instead of a synthesised single answer. Decision support means the decision stays with a person.