HyPhi: Information-Geometric Hyperscanning for Inter-Brain Alignment

Abstract

We introduce and further develop HyPhi(Φ), a scalable hyperscanning analysis pipeline that characterizes inter-brain coupling using discrete Ricci curvature and curvature entropy. Instead of treating inter-brain synchrony as a scalar quantity, HyPhi models time-resolved inter-brain networks as evolving geometric objects. Edge-centric curvature metrics (Forman–Ricci Curvature and its augmented variant) quantify how coupling reorganizes during social interaction. Curvature entropy captures regime transitions in inter-brain geometry.

Hypothesis

Distinct interaction regimes correspond both to differences in average synchrony strength and to shifts in geometric heterogeneity of inter-brain networks.

Specifically:

• Rule-constrained coordination → stable, low-entropy geometric regimes
• Unconstrained interaction → higher dispersion and transient recruitment of negatively curved (bridge-like) edges
• Regime transitions → detectable entropy modulations at sub-trial resolution

This moves beyond node – or pairwise synchrony toward network-level geometric reconfiguration.

Data and equipment

We will work with dual-EEG/fNIRS hyperscanning data from structured dyadic coordination tasks and complementary connectome-informed simulations.

No new acquisition is required. Analysis will use:

• Curvature computation
• Sliding-window inter-brain connectivity graphs (PLV / CCorr)
• Standard laptops; optional cloud compute for scaling

Brainhack Objectives

1. Further expand the open-source toolkit HyPhi(Φ) into additional modules
2. Implement additional curvature computations for windowed graphs
3. Benchmark curvature entropy against further synchrony metrics

Deliverable by end of Brainhack

• Public GitHub repository module (documented pipeline)
• Reproducible hyperscanning example analysis
• Benchmark comparison figures
• Draft extended abstract suitable for proceedings