JOVANA
Explore Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

From Pairs to Networks: Multi-Brain Systems

Connect three brains, or thirty. Multi-brain interfaces, collaborative decoding, and hyperscanning — plus the crucial line between an engineered channel and mere measured synchrony.

Networking the dyad

The two-person loop is the repeatable building block. A network is just many of these dyads, arranged into sender, receiver, and combiner roles.

BrainNet linked three people — two senders, one receiver — to cooperatively solve a task; the receiver could even learn to trust the more accurate sender. In animals, the Brainet wired several rat or monkey brains together via cortical microstimulation into a distributed 'organic computer' that performed pattern classification and, in monkeys, shared control of a single virtual arm.

Collaborative and shared BCI

You do not even have to write into brains to get most of the benefit. Combining several users' decoded signals — collaborative BCI, consensus / group decoding, shared control — reliably beats the best single user on detection and selection tasks such as group P300 or rapid visual target search. Reliability is the honest, reproducible payoff of going multi-brain.

Topology and weighting matter. Roles can be flat (a consensus of equals) or hierarchical (senders feeding a receiver). The optimal way to fuse them is not a plain majority but a reliability weighting: trust each contributor in proportion to how often it is right. BrainNet embedded exactly this — the receiver could weigh a trustworthy sender more heavily.

\hat{m} = \operatorname{sign}\!\left( \sum_{i=1}^{N} w_i\, s_i \right), \qquad w_i \propto \log\frac{P_i}{1-P_i}

Reliability-weighted fusion of N binary votes s_i: the optimal weight on each contributor is its log-odds of being correct (P_i its accuracy). A confident, accurate sender counts for more.

Hyperscanning is not a channel

Hyperscanning records two or more brains simultaneously and finds inter-brain synchrony during cooperation, conversation, or shared attention; inter-brain neurofeedback even trains people to increase it. But this is measurement of coupling, not an engineered information channel: no bits are written into anyone's brain. The correlation you observe between two heads is not, by itself, communication between them.

What the networks actually achieve

Honest scope, again. The tasks are toy — rotate or don't, a binary decision — the bandwidth is tiny, and most of the gain is the statistical benefit of averaging noisy estimates, not any new cognitive bandwidth between people. Yet that gain is real and reproducible: pooled brains make more reliable decisions, and that alone is a genuine multi-brain result worth taking seriously.