Integrated Information Decomposition (ΦID)

A mathematical framework for decomposing information flow into causal constituents, used here to quantify causal emergence from latent dynamics.

Neighborhood — ranked by edge-count

thinker

Fernando E. Rosas
introduces
Developed the φID framework for causal emergence and downward causation.

method

ΦID-based estimation of causal emergence in RL latent dynamics
implements
The specific procedure: train RL agents, extract latent representations over time, and compute causal emergence using the Integrated Information Decomposition framework.

framework

Rosas's Causal Emergence via φID
implements
Quantitative emergence framework using partial information decomposition and integrated information decomposition.
Partial Information Decomposition (PID)
extends
Williams and Beer's decomposition of joint mutual information into unique, redundant, and synergistic components.

artifact

Yuan 2023 Emergence and Causality in Complex Systems: A Survey
mentions
This review paper surveys quantitative theories of causal emergence and their connections to machine learning.

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

ΦID (Integrated Information Decomposition)concept0.943
Recently proposed method for estimating causal emergence; applied to RL agent latent-space representations.
integrated information (ΦID)concept0.898
Measure of self-coherence; proposed to test if memories increase composite system coherence.
Information Integrationmethod0.716
Integration of signals across components to compute holistic states, a hallmark of organismic individuality and basal cognition
Integrated Information Theoryframework0.697
Tononi et al. framework quantifying consciousness via integration; provides mathematical tools for measuring agent complexity.
VPD (adVersarial Parameter Decomposition)concept0.695
Core methodological framework introduced in this paper; decomposes weight matrices into rank-one interpretable subcomponents using adversarial ablations.
Parameter Decomposition (vs Activation)concept0.693
Activation decompositionconcept0.690
The conventional approach (e.g., SAEs, transcoders) of decomposing activations into interpretable features.
Organised structure of relationships between parts, sufficient to exhibit information integration and collective action, can be produced via fully-distributed unsupervised learning.claim0.690
Central claim from connectionist models: complex coordination emerges without centralized control or external teacher.