A unified framework for net effects in signed social and ecological networks

arXiv:2501.09190 · physics.soc-ph, q-bio.PE · Submitted 2025-01-15 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "A unified framework for net effects in signed social and ecological networks".

Paper summary: Kai: So, this paper is about creating a framework for measuring the net and indirect effects in networks where the links have signs, weights, and directions. Basically they want a way to look at these complex structures without having to simplify them too much <ref:2501.09190#pg1>.

Mira: Right, it’s about introducing this taxonomy for measuring net and indirect effects so we can capture both the direct and indirect interactions, as well as the effect of the whole network on a single node or vice versa <ref:2501.09190#pg2>.

Kai: They claim this approach is useful because it handles signed, weighted, and directed links while still being able to capture that direct interaction and those indirect ones <ref:2501.09190#pg1>.

Lev: From a simulation standpoint, they're defining the direct effect as the entry of the adjacency matrix A where entry aij represents the direct effect of node j on node i <ref:2501.09190#pg2>.

Mira: And then they define indirect effects by looking at them multiplicatively along all possible paths, and for a path of length one, they count it as a direct effect <ref:2501.09190#pg3>.

Kai: The net effect of node j on node i is defined as the sum of the indirect effects of node j on node i across all possible orders, provided that this series converges <ref:2501.09190#pg3>.

Lev: Convergence for this is guaranteed if rho of A is less than one, which they suggest you can check by looking at whether In minus A is strictly diagonally dominant by columns or rows <ref:2501.09190#pg3>.

Conclusion: Kai: So to wrap up this paper, the authors introduce a unified framework for net effects in signed social and ecological networks which unifies and extends existing approaches from network science and computational social science <ref:2501.09190#pg1>.

Mira: What they claim is that these tools can measure the effect of the network on a node, quantify the reverse measure, which is how much a single node affects the entire network, and then disentangle direct effects from indirect effects <ref:2501.09190#pg2>.

Kai: It matters because they show that this framework can be applied to ecological systems where net and indirect effects are critical yet difficult to measure <ref:2501.09190#pg1>.

Lev: The real implication here is that they generalize classical centrality measures like Katz’s and PageRank to handle signed, weighted, and directed networks <ref:2501.09190#pg2>.

Mira: This means we get a more complete perspective than just using unsigned measures in ecological simulations <ref:2501.09190#pg3>.

Kai: They also provide practical tools, like rescaling schemes, to address convergence issues for these measures and to recover those classical centrality measures when they are applied to these complex networks <ref:2501.09190#pg3>.

Carlos Gómez-Ambrosi, *Violeta Calleja-Solanas

University of Zaragoza · GOTHAM lab, Institute of Biocomputation and Physics of Complex Systems (BIFI), University of Zaragoza · Doñana Biological Station (EBD-CSIC)

physics.soc-ph, q-bio.PE

Submitted: 2025-01-15

Updated: 2026-11-16

Journal ref: Sci Rep (2026)

DOI: 10.1038/s41598-026-62819-2

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 83/100

The gist: Complex networks provide a powerful framework for encoding pairwise interactions between entities, represented as nodes and links The proposed framework introduces a taxonomy for measuring net and…

Key concepts

Direct Effect
This is the immediate influence of one node on another within the network structure. In a signed weighted directed network, it is represented by the entry in the adjacency matrix, showing how node j directly impacts node i.
Indirect Effect
These effects represent influences that propagate through multiple steps or 'walks' along the links of the network. They are calculated multiplicatively along these paths to determine how a node's influence spreads across the entire system.
Net Effect
The net effect of a node on another is the sum of all indirect effects originating from that node, across all possible path lengths. This captures the total, cumulative impact of a node on its target, provided this infinite series converges.
Rescaling Schemes
These are mathematical procedures (global, column, and row scaling) used to adjust the network matrix for convergence when calculating net effects. They are necessary because classical measures often fail on signed matrices and help recover meaningful centrality metrics.

Terminology

Summary

Complex networks provide a powerful framework for encoding pairwise interactions between entities, represented as nodes and links The proposed framework introduces a taxonomy for measuring net and indirect effects without simplifying these information-rich networks This approach is valuable because it captures both direct and indirect interactions, the effect of the whole network on a node, and conversely, the effect of a node on the entire network while accommodating signed, weighted, and directed links. p. 0

Where the paper says it

In this article, we introduce a framework for measuring net and indirect effects without simplifying these information-rich networks. It captures both direct and indirect interactions, the effect of the whole network on a node, and conversely, the effect of a node on the entire network, while accommodating the complexity of signed, weighted, and directed links.

Defining Net Effects in Signed Weighted Directed Networks

The authors develop a framework to compute a matrix of net effects for signed weighted directed networks This involves defining the direct effect as the entry of the adjacency matrix A, where entry aij represents the direct effect of node j on node i. Indirect effects are defined as multiplicative along walks, and for a walk of length 1, they are considered direct effects. p. 1

Where the paper says it

If i ← k ← j is a walk of length 2 from node j to node i, with node k as its middle node, then the indirect effect of node j on node i along this walk is defined as aikakj (Fig. direct effects are multiplicative along walks.

The net effect of node j on node i is defined as the sum of the indirect effects of node j on node i of all orders provided that the series converges. p. 1
Where the paper says it

Finally, we define the net effect of node j on node i as the sum of the indirect effects of node j on node i of all orders, i. X∞ 𝓁=1 [A𝓁 ]ij , (2) provided that this series converges.

Convergence is guaranteed under the hypothesis that rho(A) < 1, which can be checked by ensuring In - A is strictly diagonally dominant by columns or rows. p. 3
Where the paper says it

Consequently, we have to compute the spectral radius of A, and confirm that its value lies below one to assure convergence. Since computing the spectral radius can be very costly, it is often useful to apply the fact that if A has zero diagonal and In −A is strictly diagonally dominant (SDD) by columns, then ρ(A) < 1.

Rescaling Schemes for Convergence and Centrality Recovery

To address convergence issues for classical centrality measures like Katz's and PageRank, the authors propose three natural rescaling procedures to preserve network information.

** Global rescaling involves scaling A globally by a positive number alpha, resulting in Net(alphaA) which attenuates indirect effects by a factor of alpha l. p. 3

Where the paper says it

Global rescaling The simplest rescaling procedure is to rescale A globally by a positive number α, obtaining a rescaled interaction matrix αA = [αaij ]. If we choose α such that ρ(αA) = α ρ(A) < 1, then Net(αA) = −In + (In − αA) −1 = X∞ 𝓁=1 α 𝓁A𝓁 (18) is well defined, and we have at our disposal all the net effect notions introduced earlier, but with respect to αA.

Under specific conditions on alpha, this can recover classical measures like Katz centrality and Hubbell centrality. p. 8
Where the paper says it

Incoming net effects with global rescaling generalize Katz centrality (and, in its plusone form, Hubbell’s) to signed, weighted, and directed networks; column-rescaled plus-one net effects recover PageRank, and row-rescaled plus-one net effects yield its reverse PageRank counterpart. We show that other measures that have appeared in the social network literature, like PN centrality and truncated «net effect» scores, can be interpreted as particular or truncated instances of our Net/Net+1 series under appropriate rescalings.

The column rescaling procedure involves taking positive numbers d j and scaling column j of A by d j-1, which generalizes PageRank to signed weighted directed networks. p. 3
Where the paper says it

Therefore, incoming net effects (respectively, incoming plus one net effects) with respect to αA can be interpreted as a generalization of Katz centrality (respectively, Hubbell centrality) to signed weighted directed networks. Column rescaling Sometimes, a more sophisticated rescaling procedure is convenient.

The row rescaling procedure involves scaling row i of A by d i-1, which yields a reverse PageRank counterpart. p. 8
Where the paper says it

Our framework unifies and extends several classical measures. Incoming net effects with global rescaling generalize Katz centrality (and, in its plusone form, Hubbell’s) to signed, weighted, and directed networks; column-rescaled plus-one net effects recover PageRank, and row-rescaled plus-one net effects yield its reverse PageRank counterpart.

These rescalings are necessary because the path-sum interpretation is physically meaningful only for non-dimensional matrices. p. 9
Where the paper says it

When our analysis assumes a simple network (with zero diagonal), we make this explicit and use daggered matrices (†) to separate inter- from intraspecific terms. Because units matter, the path-sum interpretation is physically meaningful only for non-dimensional matrices; our global/row/column rescalings provide interpretable routes to non-dimensionality.

The truncated versions of these net effects can be used as proxies for potentially undefined net effects with respect to A. p. 4
Where the paper says it

For example, the L-truncated vector of outgoing net effects with respect to αD−1A is defined by NetL (αD−1A, all ← ∗) = 1 T [ X L 𝓁=1 α 𝓁 (D−1A) 𝓁 ]. (30) This definition will be used in Eq.

The L-truncated vector of outgoing net effects is defined by NetL(alphaD-1A, all ← **

Ecological Applications and Extinction Prediction

The framework is applied to ecological networks governed by generalized Lotka–Volterra dynamics. The alpha matrix, which measures the direct effect of species j on species i relative to its self-inhibition, is defined as alpha = D-1A = -In + α†. p. 5

Where the paper says it

The alpha matrix is then defined by α = D−1A = −In + D−1A† = −In + α† , so that α = [αij ] = [−aij/aii]. Since αij = −aij/aii measures the direct effect of species j on species i relative to the direct effect, in absolute value, that species i has on itself, α and α† are non-dimensional, and therefore amenable to meaningful definitions of net effects —recall the small aside after Eq.

The net effects matrix of alpha in the sense of [30] is given by-alpha-1 = (In - α†)-1 = X∞ l=0 (α†)l. p. 5
Where the paper says it

When M is defined as: (i) M = J(x ∗ ) (the community matrix), (ii) M = A (the interaction matrix), or (iii) M = α (the alpha matrix); then entry (i, j) of −M−1 measures the response of the equilibrium abundance of species i to a small press perturbation in (respectively) (i) the growth rate [30, 31], (ii) the intrinsic growth rate [30, 32], or (iii) the carrying capacity [19, 30] of species j. Critically, the connection of net effects as defined from −M−1 with net effects in the sense of «overall effects through both direct and indirect pathways» [33] (Section I), while explicitly suggested in many ecological articles [19, 23, 30– 32, 34], has just been made mathematically explicit in [14] for the alpha matrix.

A sufficient condition for Net+1(α†) to be well defined is that In - α† or, equivalently, A itself, be SDD by rows. p. 6
Where the paper says it

This has led to the introduction in [14] of the «collectivity parameter» ϕ = ρ(α† ), as a measure of the degree to which net effects in an ecological community (in the sense of the net effects matrix −α−1 ) emerge as infinite aggregates of indirect effects, and cease to do so when ϕ ≥ 1. In the light of Section I, a sufficient condition for Net+1(α† ) to be well defined is that In −α† or, equivalently, A itself, be SDD by rows, i.

The vector Net(1nβ†, ∗ ← all) of incoming net effects with respect to 1nβ† is a structural predictor of species extinction because species with negative incoming net effect have a higher probability of becoming extinct under gLV dynamics. p. 7
Where the paper says it

Here, we no longer assume that our equations have a feasible equilibrium. We claim that the vector Net( 1 n β † , ∗ ← all) of incoming net effects with respect to 1 n β † is a structural predictor of species extinction, in the sense that species that suffer a negative incoming net effect have a higher probability of becoming extinct under gLV dynamics or, equivalently, under replicator dynamics.

Social Network Insights and Model Validation

The framework is tested on the social network of the 18 monks of Sampson’s monastery using replicator dynamics. The vector Net(1 18A, ∗ ← all) captures Basil, Elias and Simplicius as the three monks with the lowest (negative) incoming net effect. p. 8

Where the paper says it

Next, we have computed the vector Net( 1 18A, ∗ ← all) of incoming net effects with respect to 1 18A, whose values can be seen in Fig. Of the four monks that were expelled from the monastery, this measure captures Basil, Elias and Simplicius as the three monks with the lowest (negative) incoming net effect, but fails to capture Gregory, who has a high (positive) incoming net effect.

The simulation in agreement with the incoming net effect scores shows that Basil, Elias, and Simplicius are the first three monks whose popularity dwindles. p. 8
Where the paper says it

Finally, identifying fitness with popularity and starting in the center of the simplex, we have simulated the replicator dynamics with interaction matrix A. In agreement with the incoming net effect scores, Basil, Elias, and Simplicius are the first three monks whose popularity dwindles, while Gregory is the only one among them who ends up with a positive popularity score, as represented in Fig.

The results demonstrate that full net-effects measures better align the ranking with observed outcomes in the Sampson monastery example. p. 9
Where the paper says it

In a signed social network, full net-effects measures (rather than their truncated versions) better align the ranking with observed outcomes in the Sampson monastery example, and simple social replicator dynamics mirrors those predictions. These results provide a powerful lens for ecosystem conservation and underscore the relevance of net effects for understanding phenomena such as influence propagation and power dynamics.

This approach is pertinent to any differentiable model in which the interaction matrix encodes direct effects. p. 9
Where the paper says it

These results provide a powerful lens for ecosystem conservation and underscore the relevance of net effects for understanding phenomena such as influence propagation and power dynamics. Importantly, our results do not rest on the assumption of underlying Lotka-Volterra or replicator dynamics; the approach is pertinent to any differentiable model in which the interaction matrix encodes direct effects.

Discussion and Future Directions

The unified framework successfully unifies and extends several classical measures, showing that incoming net effects with global rescaling generalize Katz centrality and Hubbell’s to signed, weighted, and directed networks. The results provide a more complete perspective than traditional unsigned measures in ecological simulations. p. 9

Where the paper says it

Our applications illustrate a more complete perspective than traditional unsigned measures. In ecological simulations, the incoming net effect with respect to β † acts as a structural predictor of extinction risk: species with more negative incoming net effect go extinct more often, and the species with the minimum incoming value is frequently the first to vanish.

Challenges remain regarding empirical errors and the choice among global vs. p. 9
Where the paper says it

First, empirical links may contain sign/weight errors, and then sensitivity analyses (perturbing entries and averaging, comparing to truncated versions) are warranted. Second, the choice among global vs.

column vs. row rescaling which can change rankings. p. 9
Where the paper says it

row rescaling is problem-dependent (whether attenuating by walk length vs. normalizing by in- or out-strength) and can change values and rankings; reporting both the choice and its rationale is advised.

Further research should focus on integrating this framework with empirical datasets containing signs, weights, and directions of links. p. 9
Where the paper says it

For instance, we expect that the role of indirect interactions in shaping coexistence and community assembly [46, 47] will change once interaction weights are incorporated into the networks. Integrating this framework with empirical datasets that contain information on signs, weights, and directions of links may challenge the views obtained from the unsigned, unweighted, or undirected versions of these networks.

The authors provide two practical tools: rescaling choice and incoming vs. p. 9
Where the paper says it

Second, the choice among global vs. row rescaling is problem-dependent (whether attenuating by walk length vs.

outgoing aggregation to match the metric to the research question while preserving interpretability across domains. p. 8
Where the paper says it

outgoing aggregation). We have (i) defined Net matrices that incorporate direct and indirect effects across paths of all lengths; (ii) derived the conditions under which net effects can be defined —non-dimensional interaction matrices with spectral radius less than one—; and (iii) introduced three rescaling schemes —global, column-wise, and row-wise— that guarantee convergence while preserving the information of signs, directionality, and relative weights.

The inclusion of sign, weight, and direction in links makes our framework broadly applicable to real-world networks where relationships are inherently heterogeneous. p. 9
Where the paper says it

In an ecological context, note also that the indirect effects captured here are path-mediated via pairwise links; they should not be confused with traitmediated interactions or interaction modifications that alter the direct links themselves. The inclusion of sign, weight, and direction in links makes our framework broadly applicable to real-world networks where relationships are inherently heterogeneous.

p. 9
Where the paper says it

In an ecological context, note also that the indirect effects captured here are path-mediated via pairwise links; they should not be confused with traitmediated interactions or interaction modifications that alter the direct links themselves. The inclusion of sign, weight, and direction in links makes our framework broadly applicable to real-world networks where relationships are inherently heterogeneous.

The path-sum interpretation is physically meaningful only for non-dimensional matrices through the proposed rescaling methods. p. 9
Where the paper says it

When our analysis assumes a simple network (with zero diagonal), we make this explicit and use daggered matrices (†) to separate inter- from intraspecific terms. Because units matter, the path-sum interpretation is physically meaningful only for non-dimensional matrices; our global/row/column rescalings provide interpretable routes to non-dimensionality.

The inclusion of sign, weight, and direction in links makes our framework broadly applicable to real-world networks where relationships are inherently heterogeneous.

Methods for Simulation

Simulations of species extinctions with replicator dynamics involve transforming the gLV equations into replicator equations on the n-dimensional simplex Sn. The replicator equations used in simulations are z˙i = zi((βz)i − z Tβz), i = 1,. p. 7

Where the paper says it

We claim that the vector Net( 1 n β † , ∗ ← all) of incoming net effects with respect to 1 n β † is a structural predictor of species extinction, in the sense that species that suffer a negative incoming net effect have a higher probability of becoming extinct under gLV dynamics or, equivalently, under replicator dynamics. To provide supporting evidence for our claim, we have conducted the computer simulations described in the Methods Section.

.., n, where β is the interaction matrix. p. 1
Where the paper says it

Our intended interpretation is that the nodes in the network interact, and that aij measures the strength of the interaction i ← j (including whether this interaction is positive or negative). Since A represents direct effects or interaction strengths rather than mere adjacency, we call it the matrix of direct effects or the interaction matrix of the network.

The simulation procedure involves selecting a connected matrix β with off-diagonal entries randomly chosen from a uniform distribution in the interval (−1, 1) and then computing Net(1nβ†, ∗ ← all). p. 3
Where the paper says it

Since computing the spectral radius can be very costly, it is often useful to apply the fact that if A has zero diagonal and In −A is strictly diagonally dominant (SDD) by columns, then ρ(A) < 1. To see this, let B = [bij ] be any n × n matrix, and let Cj (B) denote the sum of the absolute values of the non-diagonal entries in the j-th column of B.

Species are identified as extinct if their densities fall below 10-5/n and their densities are reset to zero. p. 7
Where the paper says it

Here, we no longer assume that our equations have a feasible equilibrium. We claim that the vector Net( 1 n β † , ∗ ← all) of incoming net effects with respect to 1 n β † is a structural predictor of species extinction, in the sense that species that suffer a negative incoming net effect have a higher probability of becoming extinct under gLV dynamics or, equivalently, under replicator dynamics.

Data and Code Availability

Notebooks to reproduce all the results are available online at https://zenodo.org/records/17288130 You can also find there ready-to-go functions of all the measures of net effects, and the eight matrices of Sampson’s monastery. p. 7

Where the paper says it

We then embed these measures within our framework and test them against our measures of net effects in a standard example. Works like [8], [9] or [39] illustrate their analyses with data from Sampson’s well-known study of monk relationships in a monastery [40], a key example for signed social networks, much like Zachary’s karate club for community detection [41].

The author contributions include Conceptualization (supporting), Formal analysis (lead), Methodology (lead), Software (equal), Writing - original draft (equal), and Writing - review and editing (equal). The authors wish to thank Oscar Godoy, Fernando DiazDiaz, as well as Victor Maull and the Complex Systems Lab members (ICREA) for insightful conversations. p. 0
Where the paper says it

Measures of net effects in signed social and ecological networks Carlos Gómez-Ambrosi 1, 2 and Violeta Calleja-Solanas 3, ∗ 1Department of Mathematics, University of Zaragoza, 50009 Zaragoza, Spain 2GOTHAM lab, Institute of Biocomputation and Physics of Complex Systems (BIFI), University of Zaragoza, 50018 Zaragoza, Spain 3Department of Ecology and Evolution, Doñana Biological Station (EBD-CSIC), 41092 Seville, Spain (Dated: October 10, 2025) With improvements in data resolution and quality, researchers can now represent complex systems as signed, weighted, and directed networks. In this article, we introduce a framework for measuring net and indirect effects without simplifying these information-rich networks.

CGA acknowledges financial support from the Ministerio de Ciencia, Innovación y Universidades (grant PID2023-147734NB-I00) and VCS acknowledges financial support from the Ministerio de Ciencia e Innovación (grant PID2021-127607OB-I00).

References

[1] P. Chesson, Mechanisms of maintenance of species diversity, Annual Review of Ecology, Evolution, and Systematics 31, 343 (2000)

[2] O. Godoy, I. Bartomeus, R. P. Rohr, and S. Saavedra, Towards the integration of niche and network theories, Trends in Ecology & Evolution 33, 287 (2018)

[3] D. Soriano-Paños, L. Lotero, A. Arenas, and J. GómezGardeñes, Spreading processes in multiplex metapopulations containing different mobility networks, Physical Review X 8, 031039 (2018)

[4] A. Vendeville and F. Diaz-Diaz, Modeling echo chamber effects in signed networks, Physical Review E 111, 024302 (25)

[5] M. Newman, Networks, 2nd ed. p. 0

Where the paper says it

Measures of net effects in signed social and ecological networks Carlos Gómez-Ambrosi 1, 2 and Violeta Calleja-Solanas 3, ∗ 1Department of Mathematics, University of Zaragoza, 50009 Zaragoza, Spain 2GOTHAM lab, Institute of Biocomputation and Physics of Complex Systems (BIFI), University of Zaragoza, 50018 Zaragoza, Spain 3Department of Ecology and Evolution, Doñana Biological Station (EBD-CSIC), 41092 Seville, Spain (Dated: October 10, 2025) With improvements in data resolution and quality, researchers can now represent complex systems as signed, weighted, and directed networks. In this article, we introduce a framework for measuring net and indirect effects without simplifying these information-rich networks.

(Oxford University Press, 2018)

[6] L. Katz, A new status index derived from sociometric analysis, Psychometrika 18, 39 (1953)

[7] D. F. Gleich, Pagerank beyond the web, SIAM Review 57, 321 (2015)

[8] M. G. Everett and S. P. Borgatti, Networks containing negative ties, Social Networks 38, 111 (25)

[9] W.-C. Liu, L.-C. Huang, C. W.-J. Liu, and F. Jordán, A simple approach for quantifying node centrality in signed and directed social networks, Applied Network Science 5, 1 (25)

[10] J. Wootton, The nature and consequences of indirect effects in ecological communities, Annual Review of Ecology and Systematics 25, 443 (1994)

[11] P. R. Guimaraes Jr, M. M. Pires, P. Jordano, J. Bascompte, and J. N Thompson, Indirect effects drive coevolution in mutualistic networks, Nature 550, 511 (2017)

[12] E. Omodei, M. De Domenico, and A. Arenas, Characterizing interactions in online social networks during exceptional events, Frontiers in Physics 3, 59 (25)

[13] T. J. Case, An Illustrated Guide to Theoretical Ecology (Oxford University Press, 2000)

[14] Y. R. Zelnik, N. Galiana, M. Barbier, M. Loreau, E Galbraith, and J.-F Arnoldi, How collectively integrated are ecological communities? p. 0

Where the paper says it

Measures of net effects in signed social and ecological networks Carlos Gómez-Ambrosi 1, 2 and Violeta Calleja-Solanas 3, ∗ 1Department of Mathematics, University of Zaragoza, 50009 Zaragoza, Spain 2GOTHAM lab, Institute of Biocomputation and Physics of Complex Systems (BIFI), University of Zaragoza, 50018 Zaragoza, Spain 3Department of Ecology and Evolution, Doñana Biological Station (EBD-CSIC), 41092 Seville, Spain (Dated: October 10, 2025) With improvements in data resolution and quality, researchers can now represent complex systems as signed, weighted, and directed networks. In this article, we introduce a framework for measuring net and indirect effects without simplifying these information-rich networks.

, Ecology Letters 27, e14358 (2024)

[15] C. D. Meyer, Matrix analysis and applied linear algebra (SIAM, 2000)

[16] C. H. Hubbell, An input-output approach to clique identification, Sociometry 28, 377 (1965)

[17] S. Kéfi, Ecological networks: from structure to dynamics, in Theoretical ecology: concepts and applications, edited by K. p. 5

Where the paper says it

The matrix A = [aij ] is the interaction matrix of our ecological network [30]. As is common in theoretical ecology, we will assume that all species have negative self-regulation, i.

S. McCann and G. Gellner (Oxford University Press, 2020)

[18] C. J. Puccia and R. Levins, Qualitative modeling in ecology: loop analysis, signed digraphs, and time averaging, in Qualitative simulation modeling and analysis, edited by P. A Fishwick and P. A Luker (Springer, 1991)

[19] E. A. Bender, T. J. Case, and M. p. 1

Where the paper says it

X∞ 𝓁=1 [A𝓁 ]ij , (2) provided that this series converges. The convergence of this number series for all i and j is equivalent to the convergence of the matrix series X∞ 𝓁=1 A𝓁 , (3) in which case we have the equality X∞ 𝓁=1 [A𝓁 ]ij = [X∞ 𝓁=1 A𝓁 ]ij .

E Gilpin, Perturbation experiments in community ecology: theory and practice, Ecology 65, 1 (1984)

[20] J. Roughgarden, A comparison of food-limited and spacelimited animal competition communities, in Community ecology, edited by J. Diamond and T. Case (Harper & Row, 1986)

[21] K. E. Miller, G. Aguilera, R. Bommarco, and T Roslin, Landuse intensity affects the potential for apparent competition within and between habitats, Journal of Animal Ecology 90, 1891 (2021)

[22] S. S. Browett, R. Synnott, D. B O’Meara, R E Antwis, S S Browett K J Bown O S Wangensteen D A Dawson J B Searle J M Yearsley and A D McDevitt Resource competition drives an invasion-replacement event among shrew species on an island, Journal of Animal Ecology 92, 698 (2023)

[23] J. M Montoya, G Woodward, M C Emmerson and R V Solé Press perturbations and indirect effects in real food webs, Ecology 90, 2426 (2009)

[24] E. J Lundgren, D Ramp, O S Middleton E I F Wooster E Kusch M Balisi W J Ripple C D Hasselerharm J N Sanchez M Mills and A D Wallach A novel trophic cascade between cougars and feral donkeys shapes desert wetlands, Journal of Animal Ecology 91, 2348 (2022)

[25] T. M Knight, M W McCoy J M Chase K A McCoy and R D Holt Trophic cascades across ecosystems Nature 437, 880 (2005)

[26] R. H Loyn, R G Runnalls G Y Forward and J Tyers Territorial bell miners and other birds affecting populations of insect prey Science 221, 1411 (1983)

[27] P. W Price, M Westoby B Rice P R Atsatt R S Fritz J N Thompson and K Mobley Parasite mediation in ecological interactions Annual review of ecology and systematics 17, 487 (1986)

[28] M Loreau, The ecosystem, in Unsolved problems in ecology, edited by A Dobson R D Holt and D Tilman (Princeton University Press, 2020)

[29] S. Kéfi, V Domínguez-García I Donohue C Fontaine E Thébault and V Dakos Advancing our understanding of ecological stability, Ecology letters 22, 1349 (2019)

[30] M Novak, J D Yeakel A E Noble D F Doak M Emmerson J A Estes U Jacob M T Tinker and J T Wootton Characterizing species interactions to understand press perturbations: what is the community matrix?

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Enhanced network analysis for complex systems allows for measuring net and indirect effects without simplifying information-rich networks, capturing both direct and indirect interactions, the effect of the whole network on a node, and conversely, the effect of a node on the entire network.

  2. Improved AI systems can quantify influence propagation and power dynamics in social networks by identifying informative rankings derived from these net effects measures.

  3. The framework enables ecologists to predict community response to perturbations by using species with negative incoming net effect as a structural predictor of species extinction, suggesting that species that suffer a negative incoming net effect have a higher probability of becoming extinct under gLV dynamics or, equivalently, under replicator dynamics.

  4. AI models can be adapted to analyze ecological communities governed by generalized Lotka-Volterra equations by using the alpha and beta matrices to determine the response of equilibrium abundance to changes in carrying capacities or intrinsic growth rates.

  5. The system allows for the interpretation of ambiguous net effects from ecological literature by connecting them explicitly with mathematically well-defined net effects from matrices like alpha dagger and beta dagger, clarifying heuristic links that previously lacked rigorous mathematical foundations.

Abstract

With improvements in data resolution and quality, researchers can now represent complex systems as signed, weighted, and directed networks. In this article, we introduce a framework for measuring net and indirect effects without simplifying these information-rich networks. It captures both direct and indirect interactions, the effect of the whole network on a node, and conversely, the effect of a node on the entire network, while accommodating the complexity of signed, weighted, and directed links. Our taxonomy unifies and extends existing approaches and measures from network science, computational social science, and ecological networks. We demonstrate its value in ecological systems, where net and indirect effects are critical yet difficult to quantify. Using generalized Lotka-Volterra dynamics, we find a strong correlation between negative net effects and species extinction. We further apply the framework to a real-world social network, where it identifies informative rankings that illuminate influence propagation and power dynamics.

Related papers