Order‑r Interaction in Complex Systems

• Order‑r interaction is a framework that describes precise interdependencies among exactly r entities, distinguishing it from pairwise or higher‐multiplicity interactions.
• It is applied in various fields including hypergraph filtering in network science, high‐order spin models in physics, and synergy measurement in game theory.
• Analytical and computational methods for order‑r interactions enable scale‑specific inference, revealing emergent dynamics and complex system structures.

An order‑r interaction refers to any interaction, coupling, or effect that intrinsically involves exactly r constituents, entities, or attributes, as opposed to pairwise (r=2) or higher multiplicity (r>2) interactions. This concept appears across statistical physics (spin models, quantum criticality), network science (hypergraph data, synchronization models), feature attribution and game theory (interaction indices), and experimental design. Analytical and computational frameworks for capturing order‑r interactions are crucial for modeling, inference, and unveiling system structure at the correct scale.

1. Formal Definitions Across Domains

Hypergraphs and Complex Systems

In higher-order data structures, an order‑r interaction is a hyperedge connecting r distinct nodes. For a hypergraph $H=(V,E)$, an edge $e\in E$ is of order r if $|e|=r$. The collection of all order‑r interactions forms an r-uniform hypergraph, and filtering to isolate such interactions enables scale-specific analysis of group dynamics (Landry et al., 2023).

Statistical Physics and Spin Models

In statistical mechanics, the prototypical order‑r interaction is a term in the system Hamiltonian or Gibbs measure coupling r degrees of freedom. The p-spin Curie–Weiss model provides a canonical example, with Hamiltonian

$$H_p(\bm x) = -\frac{\beta}{N^{p-1}} \sum_{1 \leq i_1 < \cdots < i_p \leq N} x_{i_1} \cdots x_{i_p},$$

corresponding to a pure interaction of order $p$ among spin variables (Mukherjee, 2024).

Game Theory and Feature Attribution

Order‑m interactions quantify the additional value generated when specific entities (e.g., players or features) act jointly in the context of size-m coalitions. The order‑m Shapley interaction component $I^{(m)}(i, j)$ is defined as the average marginal contribution of the pair $(i, j)$ across all coalitions of size $m$ drawn from $N \setminus \{i, j\}$ (Zhang et al., 2020).

2. Mathematical and Algorithmic Characterizations

Hypergraph Order‑r Filtering

Order‑r interactions in a dataset are extracted by defining

$$E_r = \{e \in E : |e| = r\}, \quad V_r = \bigcup_{e \in E_r} e,$$

yielding the filtered hypergraph $H_{(=, r)} = (V_r, E_r)$. Generalizations allow greater-than, less-than, or not-equal-to filters (Landry et al., 2023).

High-Order Regression and Experimental Design

In paired comparison designs, models may include order‑(r+1) interactions encoded as tensor products of regressor vectors. Design matrices are constructed so that the information matrix is block-diagonal, each block corresponding to parameters for interactions up to order r+1, enabling optimal estimation of high-order effects (Nyarko, 2019).

Game-Theoretic Interaction Decomposition

Order‑m interaction indices are derived as

$$I^{(m)}(i, j) = \frac{1}{\binom{n-2}{m}} \sum_{S \subseteq N \setminus \{i, j\}, |S| = m} \left[v(S \cup \{i, j\}) - v(S \cup \{i\}) - v(S \cup \{j\}) + v(S)\right],$$

admitting linearity, symmetry, and accumulation properties (Zhang et al., 2020).

3. Physical Models: High-Order Terms in Interactions

Nematic Quantum Criticality and Marginal Order‑N Interactions

At a two-dimensional nematic quantum critical point, the effective N-point coupling among order-parameter fluctuations becomes singular for $N \geq 4$. Explicitly, the N-point fermion loop

$\Pi_N \sim \lambda^{2(3-N)} \quad (\lambda \to 0)$

diverges in the collinear low-energy regime, making all such interactions marginal under the anisotropic scaling $$q_0 \sim \lambda^3, q_x \sim \lambda^2, q_y \sim \lambda$$. Consequently, the effective action cannot be truncated at any finite order—infinitely many order‑N interactions survive (Thier et al., 2011).

Elasticity of Colloids in Nematic Liquid Crystals

Higher-order multipole expansions describe interactions between axially symmetric colloidal particles. For a particle of radius $a$, the director perturbation is expanded as

$$n_\mu(\mathbf{r}) = \sum_{l=1}^N a_l (-1)^l \partial_\mu \partial_z^{l-1} \frac{1}{r}.$$

For instance, boojum-decorated spheres effectively realize order‑6 multipoles ($2^6 = 64$ multipolarity), contributing nontrivial angular dependence up to $1/r^7$ (Chernyshuk, 2012).

4. Order‑r Interactions in Dynamical Systems

In coupled oscillator models, generic r‑body Kuramoto dynamics incorporate terms

$$\frac{K_r}{N^r} \sum_{j_1, \dots, j_r} \sin(\theta_{j_1} + \cdots + \theta_{j_r} - r\theta_i),$$

generating intricate collective phenomena. In D=2, higher-order interactions do not shift the synchronization threshold due to a cancellation; in D>2, three-body terms shift the critical coupling as $\Delta K_c = \frac{D-2}{D}K_2$, and strong higher-order interactions induce bistability and hysteresis (Fariello et al., 2024).

5. Practical Methods and Empirical Significance

Filtering and Structure in Higher-Order Data

Analysis restricted to order‑r interactions reveals scale-dependent organization. For example, effective information, degree assortativity, and betweenness centrality exhibit nontrivial dependence on r in email communication networks, indicating that roles and connectivity structure vary dramatically when aggregating only same-sized group interactions, as opposed to full hypergraph aggregation (Landry et al., 2023).

Experimental Design

Optimal paired comparison designs accounting for up to order‑r interactions require allocation across pairs with varying comparison depths; explicit algorithms maximize information for main effects through to order‑r by solving convex maximization over weights on orbits (partitioned by difference count) (Nyarko, 2019).

6. Learning and Identification of Interaction Order

In high-dimensional statistical models, consistent estimation of interaction order is often impossible unless sufficient system parameters are known a priori. In the p-spin Curie–Weiss model, simultaneous estimation of the inverse temperature $\beta$ and interaction order p is unfeasible. If $\beta$ is known and above a critical threshold $\beta^*(p)$, exponentially consistent estimators for p based on sample magnetization become possible except at a countable set of coincidence points. For $\beta < \beta^*(p)$, the Fisher information collapses and p cannot be identified (Mukherjee, 2024).

7. Structural and Theoretical Implications

Order‑r interactions generate unique phenomena inaccessible to pairwise models: nontrivial emergent dynamics (e.g., higher-order synchronization), singularities in effective field theory (necessitating infinite action terms), and stratified hypergraph and network structure invisible at aggregate levels. Marginality and nonlocal singular behavior, particularly in quantum critical or strongly interacting regimes, invalidate traditional truncation schemes (e.g., Hertz–Millis), demanding nonperturbative or infinite-order descriptions (Thier et al., 2011).

Summary Table: Core Implementations of Order-r Interaction

Domain	Order‑r interaction definition	Key implication
Hypergraph/network science	Hyperedge of size r	Reveals scale-specific connectivity patterns
Spin systems/statistical mech.	Term in Hamiltonian coupling r variables	Determines phase, criticality, inference
Game theory/feature attribution	Marginal synergy in coalitions of size r+2	Dissects feature/player synergy by scale
Elastic multipole expansions	Tensor of rank r in field expansion	Describes high-order colloidal interactions
Coupled oscillator models	r-body phase-coupling term in dynamics	Induces shifted thresholds, new phases

Order‑r interaction analysis is thus indispensable for multiscale modeling, inference, and understanding of both emergent and fundamental system properties across scientific disciplines.