Loops, not groups: Long cycles are responsible for discontinuous phase transitions in higher-order network contagions

Laurent H\'ebert-Dufresne; Leah A. Keating

REVIEW 3 major objections 5 minor 1 cited by

Loops, not groups: Long cycles are responsible for discontinuous phase transitions in higher-order network contagions

T0 review · 3 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Discontinuous phase transitions in complex contagions are caused by long cycles where separate transmission chains meet inside a group, not by the group structure itself.

desk verdict A genuinely new mechanistic claim—loops, not groups, drive discontinuous cascades—supported by simulations, but the printed equations don't pass the α=1 sanity check and must be fixed before I'd trust the analytical branches. read the letter →

arxiv 2511.15688 v2 pith:W5HSRR45 submitted 2025-11-19 physics.soc-ph

Leah A. Keating , Laurent Hébert-Dufresne This is my paper

classification physics.soc-ph

keywords complexcontagionhigher-ordernetworksbranchingprocessesphasetransitionspercolationclusteredloopsglobalcascades

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that in higher-order network contagions the discontinuous jump in the fractional size of a global cascade is driven by loops—places where two different transmission chains meet in the same group—rather than by group interactions on their own. The authors build a self-consistent branching-process model on clustered networks that separates two kinds of second exposure: reinforcement along a single transmission chain inside a triangle, and reinforcement from a separate chain entering the same triangle. Comparing the three cases (both effects, group-only, loop-only) shows that only when the loop-induced enhancement α↷ is greater than 1 does the cascade size S jump discontinuously at a critical transmission probability; group-only enhancement raises the cascade size but leaves the transition continuous. If this is right, it redirects explanations of explosive contagion from the mere presence of groups to the statistical abundance of long cycles in the network.

What carries the argument

The load-bearing object is the inward percolation equation for triangles, Eq. (14), which computes v_in, the probability that a randomly chosen triangle fails to connect a node to the giant component. Its key input is ρ_in=1−G_t(u_in,v_in), the probability that a neighbor inside the triangle was activated by an external transmission chain—i.e., a loop—and the corresponding loop transmission probability T↷=T−(1−T)^{α↷}. The equation separates three cases: neither neighbor activated externally, one activated externally, and both activated externally, with different transmission probabilities TΔ and T↷ applied in each. This separation is what allows the model to attribute the discontinuity to t

What would settle it

Simulate the loop-only case (αΔ=1, α↷=10) on a network composed of many isolated triangles, so that the two neighbors of a node in a triangle are never independently activated by external chains, and measure the fractional cascade size as a function of transmission probability near the predicted critical point; if the curve is continuous, the independence closure is creating the discontinuity and the central claim is false.

Watch

Extended reading notes

Core claim

The central claim is that, in a self-consistent branching-process model of complex contagion on networks built from independent edges and triangles, a discontinuous phase transition in the mean size S of the giant component occurs precisely when transmission is enhanced along loops—that is, when a node in a triangle is activated by a transmission chain that came from outside the triangle (α↷>1). Group-level reinforcement (αΔ>1) amplifies the size and probability of cascades but does not by itself change the order of the transition. The model is solved by writing separate outward and inward percolation equations: the outward direction gives the probability R of a giant component, while the in

Load-bearing premise

The model's size equation assumes that the two neighbors of a node inside a triangle become activated independently, so their joint state factorizes; if correlations between these two neighbors are strong, the predicted discontinuity could be an artifact—and the printed equation also fails a consistency check at αΔ=α↷=1 against the known simple-contagion result, suggesting a possible algebraic error in the closure.

Editorial extensions

If this is right

If the central claim holds, models that explain explosive transitions by higher-order interactions alone need to be reconsidered: the order of the phase transition is controlled by α↷, the enhancement from separate chains meeting in a group, not by αΔ.
Hypergraph features that promote long cycles—such as low overlap between hyperedges—should promote bistability and discontinuous transitions, while features that hinder cycles should suppress them.
The inward/outward branching-process formalism yields quantitatively accurate predictions for both the size and probability of nonlinear cascades, validated by Monte Carlo simulations, and can serve as a reference for other self-consistent approximations.
The asymmetry between the size S and the probability R of a giant component means that experiments or simulations measuring only the probability of a global cascade may miss the discontinuous nature of the transition in the cascade size.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural testable extension: generate networks with identical triangle-degree distributions but different global clustering or cycle statistics (e.g., by rewiring triangles into larger loops) and check whether the discontinuous transition tracks the loop statistics rather than the group count.
The independence closure in Eq. (14)—treating the two triangle neighbors as activated independently—is the most sensitive approximation; a correlated closure that respects the triangle edge could be tested on small clustered networks where correlations are strongest and might shift the location or existence of the discontinuity.
The distinction between within-chain and between-chain reinforcement might transfer to other dynamical systems on networks, such as threshold or opinion models, where the convergence of two influence paths could be the generic cause of bistability.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

What should you know: this paper separates two second-exposure mechanisms in triangle-based clustered networks and argues that only the loop mechanism—where separate transmission chains intersect inside a group—creates discontinuous phase transitions. That is a real contribution, and I think the qualitative claim is likely right. The three-case comparison (groups+loops, groups only, loops only) is clean, and the Monte Carlo support is solid. The inward/outward percolation distinction is a useful technical step, and the result is new relative to Iacopini et al. and Malizia et al.

The problems are in the written equations. The definition T↷ = T − (1−T)^{α↷} is obviously wrong: for α↷=1 and T<0.5 it gives a negative probability. The intended form is almost certainly 1−(1−T)^{α↷}. But that is minor compared to Eq. (14), which does not reduce to the simple-contagion equation (6) when αΔ=α↷=1. Plugging in T=0.2 and G_t=0.5 gives v_in≈0.95 versus v≈0.79 from Eq. (6). That is not a typo in one coefficient; it means the printed closure is not self-consistent. The approximation that the two neighbors in a triangle are independently activated—ρ_in = 1 − G_t(u_in, v_in)—is another soft spot, since correlations induced by the triangle edge are ignored. As written, the unstable branches in Figs. 3 and 6 could be artifacts of the algebra. The fact that simulations reproduce the qualitative jump in both cases suggests the mechanism is real, but the derivation cannot be reproduced from the printed equations.

One more thing: the abstract says "only the latter mechanism can give rise to a discontinuous transition," but that is demonstrated for this model class, not as a general theorem. It should be hedged.

Bottom line: the idea is important, the simulations are evidence, and the paper deserves a serious referee. It needs major revision before publication: correct the typos, fix Eq. (14) so it reduces to simple contagion at α=1, and restate the scope of the claim. Send it to review, but ask for those changes.

Referee Report

3 major / 5 minor

Summary. The paper presents a self-consistent branching-process/perpcolation model for complex contagion on Newman-Miller networks composed of ordinary edges and triangles. It separates two higher-order mechanisms: an enhanced transmission probability T_Δ when a node receives a second exposure within the same triangle (group effect), and an enhanced probability T_↷ when the second exposure comes from a separate transmission chain meeting the triangle through a long cycle (loop effect). The authors claim that the size of the giant component S undergoes a discontinuous phase transition only when T_↷>T, whereas group effects alone yield continuous transitions. This is supported by a three-case comparison (α_Δ=α_↷=10; α_Δ=10,α_↷=1; α_Δ=1,α_↷=10) and by Monte Carlo simulations on doubly-Poisson clustered networks.

Significance. If the result holds, the paper offers a useful conceptual clarification of a much-debated question in higher-order contagion: whether explosive transitions are caused by groups themselves or by feedback loops. The separation of outward and inward percolation directions is a potentially valuable extension of clustered-network percolation to complex contagions. The explicit self-consistent equations and the three-case simulation comparison are strengths, and the base simple-contagion case is validated in Fig. 2. However, the printed analytical equations contain an algebraic inconsistency and the simulation methods are not described, so the manuscript in its current form does not yet provide a reproducible account of its central claim.

major comments (3)

[Complex contagion, Eqs. (13)-(14)] Equation (14) does not reduce to the simple-contagion equation (6) when α_Δ=α_↷=1, which is a required self-consistency check. With T_Δ=T and the presumably intended T_↷=T, substituting T=0.2, G_t=0.5 (so ρ_in=0.5) gives v_in≈0.954 from Eq. (14), whereas Eq. (6) gives v≈0.794. The discrepancy is the factor 2 in the first term, 2(1−T)(1−T_↷)ρ_in²; removing that factor makes Eq. (14) algebraically identical to Eq. (6) in the α=1 limit. Additionally, the printed definition T_↷ = T − (1−T)^{α_↷} is not a probability (for α_↷=1 and T<0.5 it is negative); it should be 1−(1−T)^{α_↷}, as in Eq. (8). Since Eqs. (14)-(15) generate the S-curves in Figs. 3 and 6 from which the central conclusion is drawn, the manuscript as printed cannot be reproduced. The authors must correct both errors and rerun/verify the theoretical curves; the qualitative claim stands only if the corrected equations retain the
[Monte Carlo validation, Figs. 2, 3, 5, 6] The Monte Carlo simulations are central to validating Eqs. (11) and (15), but no simulation methodology is given. The manuscript does not specify the network generation algorithm for Newman-Miller networks with doubly-Poisson p_{s,t}, the update rule implementing T_Δ and T_↷ in finite populations, the definition of a 'global cascade' or of the 'largest connected component' in a direction-dependent complex contagion, or the number of network realizations and seeds per point. Without these details, the claimed agreement between theory and simulations cannot be assessed, and the numerical evidence is not reproducible. Please provide a full simulation description or a supplementary implementation.
[Abstract and Conclusions] The abstract and conclusions make a categorical claim ('only the latter mechanism can give rise'; 'Group effects alone, without long cycles, produce standard continuous phase transitions'), but the evidence is three parameter regimes for one doubly-Poisson degree distribution (ν=1,2,3, μ=6.5−2ν). The three cases demonstrate that α_↷>1 can produce a discontinuity and that α_↷=1 does not in these examples, but they do not establish that group effects alone can never produce a discontinuous transition. A broader parameter scan in (α_Δ, α_↷, T, ν) and/or a local expansion of Eq. (15) near the critical point would be needed to support the universal wording.

minor comments (5)

[Throughout] The labels 'groups, no loops' and 'loops, no groups' may mislead: the network always contains triangles and long cycles; what changes is whether the higher-order transmission probability is enhanced along those paths. Consider rephrasing, e.g., 'no loop-enhanced transmission', to avoid implying that cycles are structurally absent.
[Fig. 3, caption] The statement 'The unstable equilibria were calculated as in [18]' is too vague. The authors should give the method or equation used to locate unstable branches, since [18] concerns a different generalized contagion model.
[Eq. (14) typesetting] The expressions 'ρ2in' and '2ρin(1−ρin)' should be typeset with explicit superscripts and multiplication signs; the current form makes the algebra harder to follow, especially around the factor 2.
[Figs. 2, 3, 5, 6] The captions say 'networks with at least 100,000 nodes' but do not state how many networks were generated per parameter point, how many initial seeds were used, or whether error bars are omitted because they are smaller than the markers. Please specify.
[Sec. 'Complex contagion'] The phrase 'changing every second exposure to transmit with T_Δ' is unclear. Define precisely when a transmission uses T vs. T_Δ vs. T_↷ in a triangle; this is especially important for the readers to understand the difference between the outward and inward percolation constructions.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the loops-vs-groups conclusion is a computed property of a parameterized model, not a fitted or definitional input.

full rationale

The central claim is derived from a model whose transmission parameters alpha_Delta and alpha_curvearrowright are stated postulates, not fitted quantities. The discontinuity is obtained by solving the fixed-point equations (9)-(15) and is then compared with Monte Carlo simulations; it is not assumed as an input. The three-way comparison (alpha_Delta=alpha_curvearrowright=10; alpha_Delta=10, alpha_curvearrowright=1; alpha_Delta=1, alpha_curvearrowright=10) is a parameter scan that exhibits the phase-transition distinction, rather than a restatement of the conclusion. Citations to the authors' earlier branching-process work [14,15] are methodological background and are not used to justify the central 'only loops' result. The apparent algebraic inconsistency of Eq. (14) with the simple-contagion limit Eq. (6) at alpha=1 is a serious correctness/artifact concern, but it is not a circularity: the paper does not define its prediction in terms of that equation, nor fit the discontinuity to data. Under the circularity rubric, therefore, the derivation is not self-referential in a load-bearing way.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

Central claim rests on the Newman-Miller network ansatz, the doubly-Poisson example distribution, the independent-neighbor closure in Eq. (14), and the postulated transmission-boost functional forms. None of these are fitted, so the circularity burden is low; but the closure approximation is a significant modeling assumption that could influence the phase-transition character.

assumptions (5)

domain assumption The network is a Newman-Miller network composed only of single edges and triangles (3-cliques).
The generating functions and percolation equations are built exclusively for this joint distribution of single edges and triangles (Section 'Simple contagion', Eqs. 1-3).
domain assumption The examples use a doubly-Poisson degree distribution ps,t = e^{-μ} μ^s/s! e^{-ν} ν^t/t!.
All figures and the phase-transition conclusions are drawn for this distribution with ν=1,2,3 and μ=6.5−2ν (Eq. 4).
domain assumption In the inward percolation equation, the two neighbors of node i in a triangle are treated as independently activated with probability ρ_in.
Eq. (14) uses products ρ^2 and ρ(1−ρ), ignoring correlations induced by the triangle edge; this is the central closure approximation.
domain assumption Second-exposure boost is described by T_Δ = 1−(1−T)^{α_Δ} and (presumably) T_↷ = 1−(1−T)^{α_↷}.
These functional forms are postulated as the complex-contagion mechanism (Eq. 8 and the definition near Eq. 14); no empirical derivation is given.
standard math The system is in the thermodynamic limit (infinite population, branching process).
Branching processes operate over an infinite population; the authors state this in the abstract and introduction.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Loops, not groups: Long cycles are responsible for discontinuous phase transitions in higher-order network contagions." pith.science (2026). https://pith.science/paper/W5HSRR45

@misc{pith2026251115688,
  author       = {Pith},
  title        = {Pith review of: Loops, not groups: Long cycles are responsible for discontinuous phase transitions in higher-order network contagions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W5HSRR45}},
  note         = {Machine review of arXiv:2511.15688}
}

read the original abstract

We study a self-consistent approach to introduce higher-order effects in a branching process model of complex contagion on clustered networks. Branching processes operate over an infinite population such that they never circle back and interact with previously exposed parts of the system. This infinite, treelike, structure makes it tricky to account for complex contagion mechanisms such as group effects, peer pressure, or social reinforcement where multiple exposures interact in synergistic ways. Here we present a self-consistent solution that accounts for local group structure and global cycles where the process can feedback on itself. This allows us to distinguish multiple exposures that stem from a single transmission chain, from those occurring at the intersection of different transmission chains. We find that only the latter mechanism can give rise to a discontinuous phase transition in the size of global cascades, which is a defining feature of complex contagions. Group effects alone, without long cycles, produce standard continuous phase transitions.

Figures

Figures reproduced from arXiv: 2511.15688 by the authors.

**Figure 2.** FIG. 2. The average size [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 4.** FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗

Figures from the paper (2 more)

**Figure 3.** Figure 3: FIG. 3. The case where [PITH_FULL_IMAGE:figures/full_fig_p003_3.png]

**Figure 5.** Figure 5: FIG. 5. The case where [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Nested hyperedges promote the onset of collective transitions but suppress explosive behavior
physics.soc-ph 2026-01 conditional novelty 6.0 of 10

Nestedness between pairwise and three-body hyperedges promotes earlier epidemic onset while raising the three-body infectivity required for bistability, thereby smoothing explosive transitions.

Reference graph

Works this paper leans on

18 extracted references · 1 linked inside Pith · cited by 1 Pith paper

[1]

Majhi, M

S. Majhi, M. Perc, and D. Ghosh, Dynamics on higher- order networks: A review, Journal of the Royal Society Interface19, 20220043 (2022)

2022
[2]

C. Bick, E. Gross, H. A. Harrington, and M. T. Schaub, What are higher-order networks?, SIAM Review65, 686 (2023)

2023
[3]

Ferraz de Arruda, A

G. Ferraz de Arruda, A. Aleta, and Y. Moreno, Conta- gion dynamics on higher-order networks, Nature Reviews Physics6, 468 (2024)

2024
[4]

St-Onge, R

J. St-Onge, R. Harp, G. Burgio, T. M. Waring, J. Lovato, and L. H´ ebert-Dufresne, Defining and classifying models of groups: The social ontology of higher-order networks, arXiv preprint arXiv:2507.02758 (2025)

arXiv 2025
[5]

Iacopini, G

I. Iacopini, G. Petri, A. Barrat, and V. Latora, Simplicial models of social contagion, Nature Communications10, 2485 (2019)

2019
[6]

Battiston, C

F. Battiston, C. Bick, M. Lucas, A. P. Mill´ an, P. S. Skardal, and Y. Zhang, Collective dynamics on higher- order networks, arXiv preprint arXiv:2510.05253 (2025)

arXiv 2025
[7]

Battiston, E

F. Battiston, E. Amico, A. Barrat, G. Bianconi, G. Fer- raz de Arruda, B. Franceschiello, I. Iacopini, S. K´ efi, V. Latora, Y. Moreno,et al., The physics of higher-order interactions in complex systems, Nature Physics17, 1093 (2021)

2021
[8]

St-Onge, I

G. St-Onge, I. Iacopini, V. Latora, A. Barrat, G. Petri, A. Allard, and L. H´ ebert-Dufresne, Influential groups for seeding and sustaining nonlinear contagion in het- erogeneous hypergraphs, Communications Physics5, 25 (2022)

2022

Show all 18 references

[9]

Bret´ on-Fuertes, C

E. Bret´ on-Fuertes, C. Clemente-Marcuello, V. Sanz- Arqu´ e, G. Tom´ as-Delgado, S. Lamata-Ot ´ ın, H. P´ erez- Mart ´ ınez, and J. G´ omez-Garde˜ nes, Explosive adoption of corrupt behaviors in social systems with higher-order interactions, Chaos: An Interdisciplinary Journa...

2025
[10]

Malizia, S

F. Malizia, S. Lamata-Ot ´ ın, M. Frasca, V. Latora, and J. G´ omez-Garde˜ nes, Hyperedge overlap drives explosive transitions in systems with higher-order interactions, Na- ture communications16, 555 (2025)

2025
[11]

Lamata-Ot ´ ın, F

S. Lamata-Ot ´ ın, F. Malizia, V. Latora, M. Frasca, and J. G´ omez-Garde˜ nes, Hyperedge overlap drives syn- chronizability of systems with higher-order interactions, Physical Review E111, 034302 (2025)

2025
[12]

M. E. Newman, Random graphs with clustering, Physical Review Letters103, 058701 (2009)

2009
[13]

J. C. Miller, Percolation and epidemics in random clus- tered networks, Physical Review E80, 020901 (2009)

2009
[14]

L. A. Keating, J. P. Gleeson, and D. J. O’Sullivan, Mul- titype branching process method for modeling complex contagion on clustered networks, Physical Review E105, 034306 (2022)

2022
[15]

L. A. Keating, J. P. Gleeson, and D. J. O’Sullivan, A generating-function approach to modelling complex con- tagion on clustered networks with multi-type branch- ing processes, Journal of Complex Networks11, cnad042 (2023)

2023
[16]

M. E. Newman, Spread of epidemic disease on networks, Physical Review E66, 016128 (2002)

2002
[17]

Kenah and J

E. Kenah and J. M. Robins, Second look at the spread of epidemics on networks, Physical Review E76, 036113 (2007)

2007
[18]

P. S. Dodds and D. J. Watts, Universal behavior in a generalized model of contagion, Physical Review Letters 92, 218701 (2004)

2004

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [1]

Majhi, M

S. Majhi, M. Perc, and D. Ghosh, Dynamics on higher- order networks: A review, Journal of the Royal Society Interface19, 20220043 (2022)

2022

[2] [2]

C. Bick, E. Gross, H. A. Harrington, and M. T. Schaub, What are higher-order networks?, SIAM Review65, 686 (2023)

2023

[3] [3]

Ferraz de Arruda, A

G. Ferraz de Arruda, A. Aleta, and Y. Moreno, Conta- gion dynamics on higher-order networks, Nature Reviews Physics6, 468 (2024)

2024

[4] [4]

St-Onge, R

J. St-Onge, R. Harp, G. Burgio, T. M. Waring, J. Lovato, and L. H´ ebert-Dufresne, Defining and classifying models of groups: The social ontology of higher-order networks, arXiv preprint arXiv:2507.02758 (2025)

arXiv 2025

[5] [5]

Iacopini, G

I. Iacopini, G. Petri, A. Barrat, and V. Latora, Simplicial models of social contagion, Nature Communications10, 2485 (2019)

2019

[6] [6]

Battiston, C

F. Battiston, C. Bick, M. Lucas, A. P. Mill´ an, P. S. Skardal, and Y. Zhang, Collective dynamics on higher- order networks, arXiv preprint arXiv:2510.05253 (2025)

arXiv 2025

[7] [7]

Battiston, E

F. Battiston, E. Amico, A. Barrat, G. Bianconi, G. Fer- raz de Arruda, B. Franceschiello, I. Iacopini, S. K´ efi, V. Latora, Y. Moreno,et al., The physics of higher-order interactions in complex systems, Nature Physics17, 1093 (2021)

2021

[8] [8]

St-Onge, I

G. St-Onge, I. Iacopini, V. Latora, A. Barrat, G. Petri, A. Allard, and L. H´ ebert-Dufresne, Influential groups for seeding and sustaining nonlinear contagion in het- erogeneous hypergraphs, Communications Physics5, 25 (2022)

2022