--- title: Structured O-Information url: https://www.emergentmind.com/topics/structured-o-information type: topic --- # Structured O-Information Searching arXiv for the cited papers and closely related O-information work to ground the article in the literature. Structured O-information denotes a set of closely related refinements of O-information in which the redundancy–synergy balance is resolved relative to an explicit structure rather than reported only as a single system-level scalar. The reference quantity is the O-information, defined for a multivariate system as the difference between total correlation and dual total correlation, with positive sign indicating redundancy-dominated organization and negative sign indicating synergy-dominated organization. In the literature, this structural refinement appears in several forms: a grouped measure that isolates between-group dependence, a pointwise measure on individual patterns, finite-difference gradients attached to variables or pairs, conditioned lagged versions for dynamical circuits, and a more abstract information-geometric decomposition on partially ordered sets that isolates admissible higher-order interactions [1902.11239] [2507.08773]. ## 1. O-information as the reference quantity For a system $\mathbf{X}^n=(X_1,\dots,X_n)$, the classical O-information is \[ \Omega(\mathbf{X}^n)=TC(\mathbf{X}^n)-DTC(\mathbf{X}^n), \] with \[ TC(\mathbf{X}^n)=\sum_{i=1}^n H(X_i)-H(\mathbf{X}^n), \qquad DTC(\mathbf{X}^n)=H(\mathbf{X}^n)-\sum_{i=1}^n H(X_i\mid \mathbf{X}^n_{-i}). \] Equivalently, \[ \Omega(\mathbf{X}^n)=(n-2)H(\mathbf{X}^n)+\sum_{i=1}^{n}\big[H(X_i)-H(\mathbf{X}^n_{-i})\big]. \] Its sign gives the standard interpretation: $\Omega>0$ indicates a redundancy-dominated system, $\Omega<0$ indicates a synergy-dominated system, and $\Omega=0$ indicates either a balance or a case in which opposing structures cancel [1902.11239]. Several analytical properties explain why O-information became the basis for later structured variants. It vanishes identically for two variables, $\Omega(X_1,X_2)=0$, and for three variables it coincides with classical co-information or interaction information. It is symmetric in the variables, additive over independent subsystems, and attains extremal values on canonical redundant and synergistic distributions: for binary vectors, the copy system yields $\Omega=n-2$, whereas the XOR system yields $\Omega=2-n$. More generally, if all variables have alphabet size $m$, then \[ \max \Omega=(n-2)\log m,\qquad \min \Omega=(2-n)\log m. \] These properties make $\Omega$ a sign-sensitive high-order statistic rather than a mere measure of overall dependence strength [1902.11239]. The original formulation already contains a structured reading. The partition-lattice analysis decomposes total correlation, binding entropy, and hence O-information along paths in the lattice of partitions. In particular, along an assembly path, \[ \Omega(\mathbf{X}^n)=\sum_{k=2}^{n-1} I(X_k;\mathbf{X}^{k-1};\mathbf{X}_{k+1}^n), \] so the global scalar can be written as a sum of triple interaction-information terms across scales. This is the conceptual bridge from unstructured O-information to later localization schemes [1902.11239]. ## 2. Grouped structured O-information A specific measure named structured O-information was introduced for systems partitioned into non-overlapping groups. Let \[ X=\begin{bmatrix}X_1\\X_2\\\vdots\\X_K\end{bmatrix}, \] with covariance matrix partitioned accordingly as $S=(S_{ij})_{i,j=1}^K$, and let $\operatorname{blockdiag}(S)$ denote the block-diagonal matrix retaining the within-group blocks and setting all between-group blocks to zero. The same partition applies to the precision matrix $C=S^{-1}$ and to $\operatorname{blockdiag}(S^{-1})$ [2507.08773]. For complex Gaussian data, the grouped quantities are \[ \mathcal{O}TC(X)=\ln\det(\operatorname{blockdiag}S)-\ln\det S, \] \[ \mathcal{O}DTC(X)=\ln\det(\operatorname{blockdiag}(S^{-1}))-\ln\det(S^{-1}), \] and the structured O-information is \[ \mathcal{O}_2(X)=\mathcal{O}TC(X)-\mathcal{O}DTC(X). \] Hence \[ \mathcal{O}_2(X)=\ln\det(\operatorname{blockdiag}S)-\ln\det S-\ln\det(\operatorname{blockdiag}(S^{-1}))+\ln\det(S^{-1}). \] For real-valued Gaussian data, the same expression is used with a factor of $1/2$ on the right-hand side [2507.08773]. The interpretation parallels the ordinary O-information, but only at the group level. If $\mathcal{O}_2(X)>0$, between-group interactions are redundancy-dominated; if $\mathcal{O}_2(X)<0$, between-group interactions are synergy-dominated. The distinction from ordinary $\Omega(X)$ is essential: the grouped measure removes within-group effects by block-diagonalization, thereby isolating only the balance among groups. The motivating example is explicit: when three or more groups each contain strong within-group redundancy but interact synergistically across groups, ordinary O-information can remain positive because redundancy occurs more often in the full system, whereas structured O-information correctly reports predominant between-group synergy [2507.08773]. The same paper gives a statistical interpretation in terms of covariance and precision hypotheses. Structured total coherence is associated with the null \[ H_0:S=\operatorname{blockdiag}(S), \] and structured dual total coherence with \[ H_0:S^{-1}=\operatorname{blockdiag}(S^{-1}). \] The formulas extend beyond Gaussianity to elliptical distributions because the dispersion matrix plays the role of covariance and the relevant KL-based differences are invariant to additive entropy constants. A further localization is defined at the group level through connection contributions: \[ KTC(k)=\mathcal{O}TC(X)-\mathcal{O}TC_{\text{disconn}(k)}, \] \[ KDTC(k)=\mathcal{O}DTC(X)-\mathcal{O}DTC_{\text{disconn}(k)}, \] \[ K\Omega(k)=KTC(k)-KDTC(k), \] so negative $K\Omega(k)$ indicates that group $k$ contributes synergy to the between-group network, while positive $K\Omega(k)$ indicates redundancy [2507.08773]. ## 3. Local O-information and pattern-level structure A different use of structure arises in the local O-information, which assigns a signed value to each individual realization $\mathbf{x}^n$ rather than only to the distribution as a whole. The local total correlation and local dual total correlation are \[ \mathrm{tc}(\mathbf{x}^n)=\sum_{j=1}^n h(x_j)-h(\mathbf{x}^n), \qquad \mathrm{dtc}(\mathbf{x}^n)=h(\mathbf{x}^n)-\sum_{j=1}^n h(x_j\mid \mathbf{x}^n_{-j}), \] where \[ h(\mathbf{x}^n)=-\log p(\mathbf{x}^n). \] The local O-information is then \[ \omega(\mathbf{x}^n)=\mathrm{tc}(\mathbf{x}^n)-\mathrm{dtc}(\mathbf{x}^n) =(n-2)h(\mathbf{x}^n)+\sum_{j=1}^n\big(h(x_j)-h(\mathbf{x}^n_{-j})\big). \] Positive $\omega(\mathbf{x}^n)$ indicates that the specific pattern is redundancy-dominated, negative values indicate synergy-dominated structure, and values near zero correspond to more balanced or weakly structured patterns [2108.11625]. A central property is that the pointwise quantity averages back to the global one: \[ \Omega(\mathbf{X}^n)=\mathbb{E}\{\omega(\mathbf{X}^n)\}. \] Accordingly, \[ \inf_{\mathbf{x}}\omega(\mathbf{x})\le \Omega(\mathbf{X}^n)\le \sup_{\mathbf{x}}\omega(\mathbf{x}). \] The finite-alphabet bounds \[ -(n-2)\log|\mathcal{X}| \le \Omega(\mathbf{X}^n) \le (n-2)\log|\mathcal{X}| \] apply to the average $\Omega$, not necessarily to each individual $\omega(\mathbf{x})$. This separates the pattern-level manifestation of synergy and redundancy from the system-level average and makes explicit that a single global statistic can conceal heterogeneous local structure [2108.11625]. The three-spin Ising example clarifies the distinction. For ferromagnetic coupling $J>0$, the global O-information is positive; for frustrated negative $J$, it becomes negative. Yet the local analysis shows that configurations with all spins aligned are redundancy-dominated, whereas the six mixed configurations are synergy-dominated regardless of the sign of $J$. What changes with $J$ is the frequency with which these states are visited. The pattern-level sign is therefore not reducible to the global average [2108.11625]. The four-voice Bach chorale application gives a concrete high-dimensional use case. The analysis preprocesses 172 chorales in major mode, transposes them to C major, encodes each voice as a 13-symbol alphabet, estimates the empirical joint distribution over four-note chords, and computes $\omega(\mathbf{x})$ for each observed chord. Frequency correlates weakly but positively with local O-information; dissonance significantly lowers $\omega$; root-position triads tend to be more redundant than first- or second-inversion forms; harmonically closer chords have higher $\omega$ than harmonically distant ones; and the outer voices display relatively weak redundancy but substantial synergy. The same framework is also applied to words associated with chords, where many common theological words are redundancy-associated while some, such as “Jesu,” are synergy-associated [2108.11625]. ## 4. Gradients and dynamical structured O-information Localization can also be expressed by finite differences of the global O-information. The first-order gradient with respect to variable $X_i$ is \[ {}_i\Omega(\mathbf{X}^n)=\Omega(\mathbf{X}^n)-\Omega(\mathbf{X}^n_{-i}), \] and the second-order gradient for a pair $(X_i,X_j)$ is \[ {}_{ij}^2\Omega(\mathbf{X}^n)={}_i\Omega(\mathbf{X}^n)-{}_i\Omega(\mathbf{X}^n_{-j}). \] The latter is symmetric and can be written as a discrete second difference: \[ {}_{ij}^2\Omega(\mathbf{X}^n)=\big[\Omega(\mathbf{X}^n)-\Omega(\mathbf{X}^n_{-ij})\big] -\big[\Omega(\mathbf{X}^n_{-i})-\Omega(\mathbf{X}^n_{-ij})\big] -\big[\Omega(\mathbf{X}^n_{-j})-\Omega(\mathbf{X}^n_{-ij})\big]. \] The first-order quantity can also be expressed as \[ {}_i\Omega(\mathbf{X}^n)=(2-n)I(X_i;\mathbf{X}^n_{-i})+\sum_{k\ne i} I(X_k;\mathbf{X}^n_{-ik}), \] or as the difference between nonnegative gradients of total correlation and dual total correlation: \[ {}_i\Omega={}_iTC-{}_iDTC,\qquad {}_iTC=I(X_i;\mathbf{X}_{-i})\ge 0,\qquad {}_iDTC=\sum_{k\ne i} I(X_k;X_i\mid \mathbf{X}_{-ik})\ge 0. \] Positive gradients indicate localized redundancy-dominated contributions; negative gradients indicate localized synergy-dominated contributions [2207.03581]. The gradient framework is designed to map high-order effects onto low-order objects. In the seven-spin frustrated Ising model, first-order gradients identify the central spin as the main source of synergy, while peripheral spins are mostly redundant; second-order gradients show redundant pairwise descriptors at low temperature and synergistic pairs at higher temperature, especially the central spin with a peripheral spin and neighboring peripheral-spin pairs. In US macroeconomic data, seven indicators have significant positive first-order gradients, GPDI is significantly synergistic, and GDP has significantly negative second-order gradients with four other variables. The paper argues that these pairwise gradients yield a sparser and more parsimonious network than local O-information alone [2207.03581]. For time series, the dynamical O-information extends the same logic to directed prediction. Given lagged state vectors \[ X_k(t)=\big(x_k(t),x_k(t-1),\ldots,x_k(t-m+1)\big), \qquad Y(t)=\big(z(t),z(t-1),\ldots,z(t-m+1)\big), \] and future target $y(t)=z(t+1)$, the conditioned dynamical quantity is \[ d\Omega_n=(1-n)I(y;\mathbf{X}\mid Y)+\sum_{j=1}^n I(y;\mathbf{X}\setminus X_j\mid Y). \] Conditioning on the target’s past removes effects due to autoregressive memory, common history, and shared inputs. For two drivers, \[ d\Omega_2=I(y;X_1\mid Y)+I(y;X_2\mid Y)-I(y;X_1X_2\mid Y), \] which is the second-order term in the transfer-entropy expansion. The framework uses maximization of $d\Omega_k$ to identify the most redundant multiplet and minimization to identify the most synergistic multiplet; it is insensitive to adding an unrelated variable, and $d\Omega_1=0$, so it is not a measure of pure pairwise effects [2007.16018]. The neural-spiking application illustrates the scope of the dynamical formulation. Using 169 channels and 1778 trials from a macaque performing a random dot motion discrimination task, with neurons categorized as H, M, and L, the method shows that both redundant and synergistic circuits influencing H targets peak around 300 ms after the go cue, synergy decays more slowly, redundant circuits are mostly composed of H and M neurons, and L neurons become important in synergistic circuits for larger multiplets. A representative H target supports a redundant circuit of 7 drivers and a synergistic circuit of 5 drivers under the surrogate-based stopping rule [2007.16018]. ## 5. Information-geometric decomposition on structured spaces A more abstract precursor of structured O-information appears in the information geometry of partially ordered sets. Instead of using the full Boolean lattice of all subsets, the framework begins with an arbitrary finite poset $S$ with bottom element $\bot$ and defines the probability simplex \[ \mathcal{S}=\{\,p \mid p(x)>0 \text{ for all } x\in S,\ \sum_{x\in S}p(x)=1\,\}. \] For any $I\subseteq S$, the lower and upper sets are \[ \downarrow I=\{x\in S\mid x\le s \text{ for some } s\in I\}, \qquad \uparrow I=\{x\in S\mid x\ge s \text{ for some } s\in I\}. \] Principal ideals $\downarrow x$ and principal filters $\uparrow x$ become the structural primitives of the decomposition [1601.05533]. Two dual coordinate systems are assigned to distributions on $S$. The exponential-family form uses feature functions \[ F_s(x)= \begin{cases} 1 & \text{if } s\le x,\\ 0 & \text{otherwise}, \end{cases} \qquad \psi(\theta)=-\log p(\bot), \] which gives \[ p(x;\theta)=\exp\Big(\sum_{s\in S^+}\theta(s)F_s(x)-\psi(\theta)\Big), \qquad \log p(x)=\sum_{s\le x}\theta(s), \] with recursion \[ \theta(x)=\log p(x)-\sum_{s