Affective Signal Intelligence
Multimodal emotion inference that keeps every learner in the flow zone
Why this name
“Affective signal intelligence” is the field's own language: affect is the scientific term for emotional state, and Aura's job is to infer it from the signals a learner naturally gives off, tone of voice, response timing, hesitation, and turn that inference into better teaching.
The problem
Learning is emotional. An anxious, bored, or overwhelmed student learns very little, however good the material. Human tutors constantly read the cues, a sigh, a long pause, a flat “…yeah, I get it”, and respond by slowing down, encouraging, or switching tack. AI tutors are blind to all of it, so they push on while the learner silently disengages.
What we're building
An affective layer for tutoring systems. From prosody, tempo, and interaction patterns, Aura estimates where a learner sits between boredom and anxiety (the classic valence to arousal map) and steers the session back toward flow: easing off when stress rises, adding challenge when a student is coasting, encouraging at exactly the right moment.
Key capabilities
Multimodal sensing: reads tone, pace, pauses, and behaviour, affect = f(prosody, tempo, pauses).
Flow-state pacing: keeps difficulty in the channel between “too easy” and “too hard”.
Timely encouragement: notices struggle early and responds with support, not more content.
Wellbeing-aware: flags sustained frustration so learning stays healthy and motivating.
Privacy-first: emotional signals help the learner, on the learner's terms, never scored, never sold.
Research focus
Multimodal emotion and engagement recognition in real learning conversations.
Mapping affective state to the right pedagogical response, not just a label.
Doing this ethically: transparent, consent-based, and protective of student wellbeing.
Where it fits at ILM AI
The empathy layer across Ilmino and the voice tutor, the most literal expression of ILM AI's mission of human-aligned AI.
Aura: learning that meets you where you are.