🔬 physics

Group dynamics shape contagion onsets and multistable active phases under collective reinforcement

This paper demonstrates that group dynamics and temporal turnover in interaction structures are central determinants of higher-order contagion transitions, revealing a richer phase diagram with multistable active phases and correcting previous static-model predictions regarding the conditions for discontinuous phase transitions.

Original authors: Santiago Lamata-Otín, Federico Malizia, Leah A. Keating, Guillaume St-Onge, Vito Latora, Jesús Gómez-Gardeñes, Laurent Hébert-Dufresne

Published 2026-03-31

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Santiago Lamata-Otín, Federico Malizia, Leah A. Keating, Guillaume St-Onge, Vito Latora, Jesús Gómez-Gardeñes, Laurent Hébert-Dufresne

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to convince your friends to adopt a new habit, like drinking a specific type of coffee or joining a new fitness challenge.

In the old way of thinking about how ideas spread, we imagined a static world: you sit in a room with the same group of people forever. If you need two friends to agree before you jump on board (a "complex contagion"), you wait. If no one else in that room changes their mind, you never will. The idea dies out.

But real life isn't static. People leave the room, new people enter, and groups form and dissolve constantly. This paper is about how that constant movement changes the rules of how ideas spread.

Here is the breakdown of their discovery, using some everyday analogies:

1. The "Static Room" vs. The "Dance Floor"

The Old View (Static/Freeze): Imagine a group of friends sitting at a table. If they are stubborn and need to see two people change their minds before they do, and no one changes, the idea is stuck. It's like a frozen pond; nothing moves, nothing spreads.
The New View (Dynamic/Dance Floor): Now, imagine that same group is at a dance floor where people constantly swap partners and join different circles. Even if a small circle doesn't have enough "converts" to change the mind of the person standing there, that person might move to a different circle where the energy is higher.
The Finding: The authors found that this constant swapping doesn't just help ideas spread a little bit; it can completely change the nature of the explosion. It can turn a slow, gradual adoption into a sudden, explosive "flash mob" effect where everyone jumps in at once.

2. The "Goldilocks" Speed of Swapping

One of the most surprising findings is that moving too fast can actually kill the idea.

Too Slow (The Frozen Pond): If people never switch groups, the idea gets stuck in small pockets and dies.
Too Fast (The Whirlwind): If people are swapping groups so fast that they can't stay in one place for more than a second, they never get the "reinforcement" they need. Imagine trying to learn a dance move by watching a video that changes every millisecond; you never get the rhythm. The idea needs a little bit of stability to "stick" and build momentum.
Just Right (The Sweet Spot): There is a "Goldilocks" speed of group switching. It's fast enough to bring new people into the mix, but slow enough that people can stay in a group long enough to feel the pressure to join in.

3. The "Three-Stage" Explosion

When you mix this movement with groups of different sizes (some small, some huge), the system gets weird and wonderful. They found four distinct ways the world can react:

The Slow Burn: A few people join, and it grows slowly. (Boring, but safe).
The Sudden Switch: The idea stays dead for a long time, then suddenly, boom, everyone joins at once. (Like a viral meme).
The "Two-Tier" Reality: This is the weird part. The system can get stuck in a state where two different levels of activity exist at the same time.
- Analogy: Imagine a party where half the room is dancing wildly, and the other half is sitting quietly. Both states are stable. Depending on how you start the party, you could end up with a wild dance party or a quiet room, and it's hard to switch between them once you're there.
The "Three-Way" Standoff: In the most complex scenario, the system can have three stable states at once:
- State A: Nobody cares (The Absorbing State).
- State B: A small, localized group cares (The "Local" Active State).
- State C: Everyone cares (The "Global" Active State).
- Analogy: Think of a town where the idea is dead, OR the idea is only popular in the wealthy neighborhood, OR the idea has taken over the whole town. All three scenarios can exist stably at the same time, depending on the history of the town.

4. The "Real World" Warning

The authors looked at real data from schools, hospitals, and conferences. They found a crucial warning for scientists and policymakers:

If you ignore time, you get it wrong.

Many computer simulations treat social networks like a static map (a frozen snapshot). The authors show that if you use these static maps to predict how a virus or a trend will spread, you might think it will explode easily. But in reality, because people move around, the "explosion" might require much stronger social pressure than we thought.

The Metaphor: It's like trying to predict how a fire spreads.
- Static Model: You look at a map of dry wood and say, "It will burn everywhere instantly!"
- Dynamic Model: You realize the wind is blowing the embers around. Sometimes the wind helps the fire jump to new trees; other times, it blows the fire out before it can catch. You can't predict the fire just by looking at the wood; you have to watch the wind.

The Big Takeaway

Time is not just a background detail; it is a main character.

How fast people move between groups determines whether an idea spreads slowly, explodes suddenly, or gets stuck in a weird limbo where multiple versions of reality coexist. If you want to understand how behaviors, viruses, or trends spread in the real world, you can't just look at who knows whom; you have to watch when and how often they change their company.

1. Problem Statement

The paper addresses a critical gap in the understanding of complex contagion (the spread of behaviors or ideas requiring multiple exposures) within higher-order networks (interactions involving groups of $n \geq 2$ individuals).

The Conflict: Previous studies on higher-order contagion have yielded conflicting results regarding the impact of temporal variability (time-varying group structures).
- Pairwise interactions: Temporal variability typically enhances mixing, lowering the invasion threshold compared to static networks.
- Higher-order interactions: Some studies suggest temporal variability raises the invasion threshold, while others (using mean-field/annealed approximations) suggest it lowers it.
The Limitation: Most existing models rely on either static (quenched) structures (groups never change) or mean-field (annealed) approximations (groups reshuffle infinitely fast). These extremes neglect the realistic scenario where group composition changes at finite rates comparable to the contagion dynamics.
The Question: How does the specific rate of group switching ( $\omega$ ) interact with nonlinear collective reinforcement to determine the macroscopic phase transition (continuous vs. discontinuous) and the stability of active states in real-world social systems?

2. Methodology

The authors develop a novel, analytically tractable framework to bridge the gap between static and mean-field limits.

The $\omega$ -AME Model: They extend the Approximate Master Equation (AME) framework to include a group switching rate ( $\omega$ ).
- Contagion Dynamics: Modeled as a higher-order process where a susceptible individual in a group of size $n$ with $i$ adopters becomes infected at rate $\beta(n, i) = \lambda i^\nu$ . Here, $\lambda$ is the intrinsic adoption rate and $\nu$ is the synergy exponent (nonlinearity).
- Group Dynamics: Individuals swap between random groups at rate $\omega$ . This is modeled as a node-level switching process that regulates the turnover of group composition.
Analytical Derivation:
- The model tracks the density of susceptible individuals with membership $k$ ( $s_k$ ) and the fraction of groups of size $n$ with $i$ adopters ( $f_{n,i}$ ).
- They derive a set of coupled differential equations (Eqs. 5–8) and solve for stationary states to determine the invasion threshold ( $\lambda_c$ ) and the persistence threshold ( $\lambda_p$ ).
- They analyze the tricritical point (where continuous transitions become discontinuous) and the conditions for multistability (coexistence of multiple stable active states).
Empirical Validation: The theoretical framework is applied to seven real-world face-to-face datasets (e.g., high schools, hospitals, workplaces) to extract empirical group-size distributions $\{p_n\}$ , membership distributions $\{g_k\}$ , and effective switching rates $\langle \omega \rangle$ .

3. Key Contributions

Analytical Interpolation: The paper provides the first analytical model that continuously interpolates between the quenched (static) and annealed (mean-field) limits via the group switching rate $\omega$ .
Reconciliation of Contradictions: It resolves the conflicting literature on temporal effects by showing that the impact of time-varying structures depends fundamentally on the nonlinearity of reinforcement ( $\nu$ ) and the group size.
Discovery of Multistable Active Phases: The authors identify a regime of multistability (multiple coexisting active states) that is absent in both static and mean-field limits, emerging only at finite switching rates in heterogeneous systems.
Methodological Warning: They demonstrate that aggregating temporal data into static hypergraphs can lead to qualitatively misleading predictions regarding critical thresholds and transition types.

4. Key Results

A. Reshaping the Phase Transition (Homogeneous Systems)

Non-Monotonic Invasion Threshold: In systems with strong nonlinear reinforcement ( $\nu > 2$ $ν > 2$ ), the invasion threshold $\lambda_c(\omega)$ $λ_{c} (ω)$ is non-monotonic.
- Initially, increasing $\omega$ lowers $\lambda_c$ (enhancing contagion by exposing individuals to more groups).
- Beyond an optimal switching rate ( $\omega^*$ ), further increasing $\omega$ raises $\lambda_c$ . Rapid reshuffling prevents individuals from staying in a group long enough to accumulate the necessary reinforcement, thereby suppressing adoption.
Transition Order: Group dynamics can transform a continuous phase transition into a discontinuous one. The region of bistability (coexistence of absorbing and active states) widens as $\omega$ and $\nu$ increase.

B. Heterogeneity and Multistability (Real-World Systems)

When applied to real-world data with heterogeneous group sizes and memberships:

Four Distinct Dynamical Regimes: The phase diagram reveals four regimes not seen in simpler models:
1. Continuous Transition: Standard SIS-like behavior (low $\nu$ ).
2. Discontinuous Transition with Absorbing-Active Bistability: Classic explosive adoption (intermediate $\nu$ ).
3. Hybrid Continuous Transition with Active Bistability: Activity emerges continuously from zero, but two stable active branches coexist (high $\nu$ ).
4. Discontinuous Hybrid Transition with Three Stable States: The system exhibits three coexisting stable states: the absorbing state and two distinct active states.
Mesoscopic Localization: The two active states in the multistable regime differ in their spatial distribution. One state is delocalized (spread across many groups), while the other is localized (sustained primarily by the largest groups). This localization is driven by the interplay between group-size heterogeneity and temporal switching.

C. Implications for Real-World Systems

Stronger Reinforcement Required: In real-world systems with measured temporal scales, the collective reinforcement ( $\nu$ ) required to trigger a discontinuous transition is significantly stronger than predicted by static or mean-field models.
Time Scales Matter: The nature of the transition is governed by the competition between the contagion timescale ( $\mu^{-1}$ $μ^{- 1}$ ) and the group switching timescale ( $\omega^{-1}$ $ω^{- 1}$ ).
- If switching is very fast (annealed), the system behaves like a mean-field model.
- If switching is very slow (quenched), fragmentation prevents invasion.
- Intermediate rates create the complex multistable behaviors.

5. Significance and Conclusion

This work fundamentally shifts the understanding of higher-order contagion by establishing that temporality is not a secondary correction but a primary dynamical mechanism.

Theoretical Impact: It extends the classical absorbing-active bistability to multistable active phases, revealing that time-varying structures can sustain multiple distinct levels of collective activity simultaneously.
Practical Implications:
- Policy/Intervention: Strategies to stop or promote contagion (e.g., viral marketing, disease control) must account for the rate of group turnover, not just the network structure. There is an "optimal" mixing rate for maximizing adoption; too much or too little movement can be detrimental.
- Data Analysis: The study warns against using aggregated static networks to model social dynamics. Such aggregation suppresses temporal features and can artificially lower the predicted threshold for explosive adoption, leading to flawed predictions.

In summary, the paper demonstrates that the interplay between collective reinforcement and group dynamics creates a rich phase space of behaviors, including multistability and localization, which are invisible to static or mean-field approximations.