If you want to know if someone is actually "connecting" with an AI, you have to check their heart rate, not what they’re saying.
April 10, 2026
Original Paper
Jointly Tracing Coupling Mode Transitions in Human-AI Interaction: Semantic and Somatic Signals
SSRN · 6439347
The Takeaway
AI is getting so good at mimicking human speech that words alone can't prove a real bond is forming. Researchers found that our bodies react physically to deep AI interactions, showing a level of biological 'syncing' that text logs can't capture.
From the abstract
We present an instrumentation framework for cross-substrate coupling analysis during human-AI dialogue, capturing semantic and somatic signals simultaneously. Earthian-BioSense acquires cardiac autonomic data via chest-strap heart rate monitor; Semantic Climate computes real-time semantic trajectory metrics, building on Morgoulis's (2025) 4D coupling framework. Three naturalistic sessions with large language models reveal distinct coupling signatures. During epistemic rupture, where a model repe