Synergy-first information decomposition frameworks

Develop scalable, synergy-first information decomposition methods that directly quantify synergistic information in multivariate systems without defining synergy as residual after subtracting redundant components, complementing coarse summary metrics like O-information and redundancy-first partial information decomposition.

Background

Most available frameworks for multivariate information decomposition are redundancy-first, typically defining synergy as what remains after redundant information is accounted for. The O-information provides only a global redundancy–synergy balance and does not directly quantify synergy itself.

The paper identifies the lack of mature synergy-first approaches as a key gap and points to recent initial attempts, calling attention to the need for methods that foreground synergy in a scalable and interpretable manner.