Rethinking how machines read human emotion.
A four-year research initiative investigating why current Emotion AI fails — and how a personalized, culturally-adaptive alternative could work. Two peer-reviewed papers, one Master's thesis, and a working prototype that questions the foundational assumptions of the field.
What if we've been training machines to read emotion the wrong way?
Most Emotion AI today is built on Paul Ekman's 1978 model — assuming seven universal emotions visible through facial expressions, generalized across all humans.
But what happens when these systems meet a country of 1.4 billion people, 22 official languages, and thousands of cultural display rules? What happens when context — the cognitive layer that gives emotion its meaning — is completely absent from the dataset?
This research argues for a fundamental shift: from universal classification to personalized calibration.
Four stages of inquiry — from theoretical critique to working prototype.
Three structural failures of current Emotion AI.
Emotion AI in India: Why Western models fail in a culturally diverse nation.
This paper ran two studies to test the limits of facial-only emotion recognition.
Culturally-Adaptive Emotion AI: An ethical framework for individualistic models.
Four studies using IDEO's AI Ethics Cards as design principles.
AffectLab: From universal labels to individual models.
Start with one individual, train deeply, then look for transferable patterns.
The System
Built in Python using OpenCV, MediaPipe Face Landmarker, scikit-learn, and Random Forest. Captures dense facial landmarks paired with self-labelled affect tags.
Six Affect-Family Buckets
Results
Three ideas from four years of research.
A system that asks how you feel could become an affective surveillance layer.
Four non-negotiables: Transparency at every extraction point. Consent-based collection. Minimum viable data. Individual ownership of personal models.
Peer-reviewed work and ongoing research.
The future of Emotion AI should be honest, personalized, and culturally aware.
If you're researching affective computing or thinking about the ethics of these systems — I'd love to talk.
Start a conversation →