🔬 physics

Intergroup violence in bursts

By adapting an Ising model to analyze street fight videos, the authors demonstrate that the proportion of group members unwilling or unable to contribute to collective victory determines whether intergroup violence erupts simultaneously in a burst or develops gradually.

Original authors: Jeroen Bruggeman, Don Weenink, Bram Mak

Published 2026-03-24

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Jeroen Bruggeman, Don Weenink, Bram Mak

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

The Big Question: Why Do Some Fights Explode, While Others Fizzle?

Imagine you are watching a street argument between two groups of young men. Sometimes, it starts with shouting and shoving, and suddenly—BOOM!—everyone in the group jumps in at the exact same time. It's a chaotic explosion of violence.

But other times, the same shouting and shoving happens, and only one or two guys get into a fight while the rest of their friends stand around watching, trying to calm things down, or just looking confused. The fight drags on slowly, or maybe it never really gets going.

The researchers (Bruggeman, Weenink, and Mak) wanted to know: What makes the difference between a sudden "burst" of violence and a slow, messy struggle?

The Secret Ingredient: The "Emotional Hill"

To understand this, the authors use a concept from physics called the Ising Model. Don't worry, you don't need a physics degree. Think of it like this:

Imagine every person in a group is standing at the bottom of a hill.

The Left Side of the Hill: Everyone is doing nothing (defecting). It's safe here. No one gets hurt.
The Right Side of the Hill: Everyone is fighting together (cooperating). This is the "victory" state.
The Top of the Hill: This is the scary part. To get from "doing nothing" to "fighting together," someone has to take a risk. If you jump in alone, you might get beaten up.

Usually, people are afraid to climb that hill. They wait for someone else to go first. This is the Dilemma of Collective Action: Everyone wants the group to win, but no one wants to be the first to get punched.

The Spark: "Turmoil" (The Heat)

So, how do they get over the hill? The researchers say it takes Turmoil.

Think of Turmoil as Heat or Noise.

Opponents yelling, pushing, pulling clothes, or tripping over creates "Heat."
As the Heat gets higher, people get more agitated, angry, and impulsive.
Eventually, the Heat gets so high that people start making "mistakes." They jump in accidentally because they are too worked up to think.

In physics, this is called a Phase Transition. It's like water. You can heat water slowly, and it just gets warmer. But once it hits 100°C, it suddenly turns into steam. The researchers found that violence works the same way. Once the "Heat" (agitation) hits a critical point, the whole group can suddenly flip from "doing nothing" to "all-out fighting."

The Deal-Breaker: The "Free Riders" (Unconditional Defectors)

Here is the most important part of the study. The researchers discovered that the "Heat" doesn't always work. It depends on who is in the group.

Imagine a group of friends. Some are "Conditional Cooperators" (they will fight if their friends fight). But some are "Unconditional Defectors." These are the people who will never fight, no matter what.

Maybe they are too scared.
Maybe they are trying to stop the fight (de-escalators).
Maybe they just don't care.

The researchers found a Magic Number: One-Third (33%).

Scenario A (The Burst): If less than 1/3 of the group are "Never-Fighters," and the Heat gets high enough, the group will explode into a burst of violence. The few brave (or crazy) ones start, and the rest follow instantly.
Scenario B (The Fizzle): If more than 1/3 of the group are "Never-Fighters," the Heat can get as high as you want, and nothing will happen. The group will never burst. The "Never-Fighters" act like anchors, dragging the group down and preventing the explosion.

The Analogy: The Campfire

Think of the group as a pile of wood trying to start a campfire.

The Wood: The people who might fight.
The Matches: The shouting and shoving (Turmoil).
The Wet Logs: The "Never-Fighters."

If you have a pile of dry wood and you throw a match on it, it bursts into flames instantly.
But if you have a pile of wood where 40% of the logs are soaked in water, you can throw a thousand matches on it. The fire might sputter, smoke, and glow, but it will never burst into a roaring flame. The wet logs absorb the heat and stop the fire from spreading.

What They Did to Prove It

The researchers didn't just guess; they watched 42 real videos of street fights (from YouTube, etc.). They coded every single person:

Who was fighting?
Who was trying to stop it?
How much "Heat" (shouting, pushing) was happening?
Did the fight start as a "Burst" (everyone at once) or a "Slow Burn"?

The Result: Their math was right.

Groups with few "Never-Fighters" (under 33%) almost always had Bursts.
Groups with many "Never-Fighters" (over 33%) almost always had Slow Starts or no collective violence at all.

Why Does This Matter?

This study helps us understand that violence isn't just about "bad people" or "angry mobs." It's about math and group dynamics.

Small Groups are Dangerous: Small groups (like 3 or 4 people) are more likely to have a "Burst" because it's easier for one person to influence the others. In a huge crowd, it's harder to get everyone to jump at once.
The "Guardian" Effect: If you want to stop a fight from exploding, you don't necessarily need to fight the aggressors. You just need to make sure there are enough people in the group who are refusing to fight (the "wet logs"). If you can get 35% of the group to stand back and say "No," the whole explosion is prevented.
It's Spontaneous: You don't need a leader to tell everyone to fight. If the "Heat" is high enough and the group is "dry" enough, the violence starts on its own, like a spark catching dry grass.

In a Nutshell

Violence in groups is like a pressure cooker.

Agitation is the heat.
Group members are the pressure.
The "Never-Fighters" are the safety valve.

If the safety valve is too small (too few people refusing to fight), the pressure cooker explodes (a burst of violence). If the safety valve is big enough (enough people refusing), the pressure just hisses and leaks, but never explodes.

1. Problem Statement

The paper addresses the mechanism behind the initiation of collective violence in small, non-professional groups (e.g., street fights). While existing literature (notably Randall Collins) emphasizes emotional barriers and "forward panics," it often lacks a formal model explaining why violence sometimes erupts in a sudden burst (simultaneous engagement by the majority) versus a gradual escalation (few participants, slow onset).

The core theoretical challenge is the dilemma of collective action: individuals face high risks in fighting alone. The authors seek to explain how groups overcome the "emotional barrier" and the "collective action hill" to engage in violence simultaneously without pre-existing norms, leaders, or long-term coordination. Specifically, they investigate the role of agitation (turmoil) and the proportion of unconditional defectors (those who refuse to fight regardless of circumstances) in determining the mode of violence onset.

2. Methodology

A. Theoretical Framework: The Adapted Ising Model

The authors adapt the Ising model from statistical physics to model social influence and collective action.

State Variables: Individuals ( $i$ $i$ ) are modeled as spins that can either Cooperate ( $C$ $C$ , fighting) or Defect ( $D$ $D$ , not fighting).
- Conditional Cooperators: Their state ( $S_i$ ) depends on the behavior of neighbors and external agitation.
- Unconditional Defectors: Their state is fixed at $S_j = -D$ regardless of agitation.
Hamiltonian ( $H$ ): The energy of the system is defined as $H = -\sum_{i \neq j} w_{ij} S_i S_j$ , where $w_{ij}$ represents the strength of the tie (proximity/visibility) between individuals. The system seeks to minimize $H$ .
Turmoil ( $T$ ): Adapted from "temperature" in physics, $T$ represents the level of agitation caused by opponents (threats, insults, physical provocation). High $T$ increases the probability of random behavioral changes ("trembling hands").
Mechanism: The model uses the Metropolis algorithm (Monte Carlo simulation). As $T$ $T$ increases, the system moves from a state of total defection. If $T$ $T$ crosses a critical threshold ( $T_c$ $T_{c}$ ), the system undergoes a phase transition.
- Burst: If the proportion of unconditional defectors ( $p$ ) is below a critical threshold ( $p_c$ ), the system jumps abruptly to a high-cooperation state (burst).
- Gradual Escalation: If $p > p_c$ , the system cannot overcome the energy barrier, and cooperation increases slowly or not at all.

B. Empirical Data Collection

Source: 42 video clips of street fights sourced from public domains (YouTube, LiveLeak, WorldStarHipHop).
Sample: 59 distinct groups (focal groups) involving 406 individuals. Groups were small ( $2 \le n < 10$ , mean $n=3.6$ ), mostly young men, unarmed, and non-professional.
Coding: Using Noldus Observer XT 14, researchers coded:
- Violence: Forceful contact or movement.
- Turmoil: Opponent behaviors (aggression, gestures, encroachment, stumbling).
- Synchrony: Simultaneous movement patterns.
- Defection: Non-participation, de-escalation attempts, or physical inability to fight.
Definition of a Burst: A collective violent event where $\ge 50\%$ of the focal group engages within 2 seconds of the first participant.

3. Key Contributions

Formalizing the "Burst" Phenomenon: The paper provides the first formal, quantitative model (Ising-based) distinguishing between bursty and gradual collective violence, moving beyond qualitative descriptions.
Critical Threshold of Defectors ( $p_c$ ): The study derives a specific mathematical prediction: a burst of violence is only possible if the proportion of unconditional defectors in the group is below $p_c \approx 1/3$ (specifically calculated as $1/3$ analytically and $0.34$ in simulations).
Role of Turmoil: The model demonstrates that agitation acts as a catalyst. Violence does not increase linearly with agitation; rather, once agitation crosses a critical level ( $T_c$ ), a phase transition occurs, triggering a burst—provided the group composition allows it.
Rejection of Alternative Explanations: The study empirically tests and rejects synchrony (moving in unison) as a primary predictor of violence onset. While synchrony often precedes violence, it also precedes non-violent outcomes and does not distinguish between bursts and gradual starts.

4. Results

Empirical Validation of $p_c$ :
- Groups exhibiting bursts had a mean proportion of unconditional defectors of 0.19.
- Groups with non-bursts (gradual or solitary violence) had a mean proportion of 0.49.
- A Welch's t-test confirmed a statistically significant difference ( $P < 0.001$ ). The critical threshold of $1/3$ effectively separated the two groups (AUC = 0.774).
Turmoil and Group Size:
- Smaller groups (dyads/triads) required lower levels of turmoil to initiate violence compared to larger groups, consistent with the model's prediction that smaller networks have lower critical thresholds.
- Turmoil preceded all collective violence events (with one exception).
Model Accuracy:
- The model correctly predicted the burst/non-burst distinction in the majority of cases.
- Exceptions: 14 out of 59 groups were misclassified. Most errors were attributed to local physical constraints (e.g., obstacles, slippery floors, intoxication, or being in a vehicle) that delayed participation beyond the 2-second "burst" window, rather than a failure of the theoretical model.
Synchrony Findings: Synchrony was observed in 21/23 bursts but also in 18 non-violent or gradual cases, confirming it is a correlate of group cohesion but not a causal predictor of the onset mode of violence.

5. Significance and Implications

Theoretical Advancement: The paper bridges physics (phase transitions) and sociology (collective action), showing that complex social behaviors like spontaneous violence can be modeled using parsimonious physical principles without assuming strong rationality or complex reputation systems.
Policy and Criminology: The findings suggest that preventing collective violence in crowds or small groups may be most effective by:
1. Reducing the proportion of "unconditional defectors" (e.g., through de-escalators or guardians who physically prevent fighting).
2. Managing the level of "turmoil" (agitation) to keep it below the critical threshold ( $T_c$ ).
Broader Applications: The authors argue the model is applicable to other forms of collective action, such as protests, revolts, and lynchings, where the presence of "unconditional defectors" and the level of agitation determine whether an event explodes into a mass action or fizzles out.
Limitations: The study relies on observational data (cannot prove causality definitively), uses a cumulative measure of turmoil (which may miss the intensity of specific triggers), and excludes norms/reputations due to data limitations, though the model can be extended to include them.

In conclusion, the paper successfully demonstrates that the composition of the group (specifically the ratio of willing participants to unconditional defectors) combined with external agitation determines whether intergroup violence manifests as a sudden, explosive burst or a slow, limited escalation.