Intergroup violence in bursts

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

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Intergroup violence in bursts".

Mira: Here's the gist of the paper. Intergroup violence in bursts examines how agitating stimuli can trigger collective violence even without usual ingroup cooperation mechanisms like norms with sanctions The proportion of group members who cannot or do not want to contribute to the public good predicts whether violence takes the form of a burst or not Core Model and Theoretical Framework The study adapts an Ising model to show that the proportion of group members who cannot or do not want to contribute to the public good—victory over opponents—predicts whether violence takes the form of a burst or not The core of the model is opponents’ agitation Collective violence involves a dilemma of collective action For an unarmed focal group, the ingroup ties that matter are face-to-face contacts in close proximity The model demonstrates that if turmoil increases, there is no gradual increase of accidental cooperators; instead, when turmoil reaches a critical level, fighting breaks out in a burst Empirical Methodology The researchers examined video recordings of street fights between groups of young men They coded violence as force being used against another’s body and or when another person’s body was forcefully moved Turmoil is calculated by multiplying the duration of each instance of turmoil by the number of individuals involved A burst is defined as an outbreak of violence by at least half of the group, or both individuals in a dyad, if they started fighting less than 2 seconds after the first with a 5% margin Key Findings on Burst Prediction The empirical distinction between bursts and non-bursts turns out to depend on the critical proportion of non-fighters versus conditional cooperators in the focal group The mean field analysis predicts that if C = 1 and D = 1/2, there is no burst of cooperation if p > 1/3 Simulations suggest that pc of simulated, fully connected networks is very close to the mean field calculation of pc, even in very small networks The critical threshold Tc for bursts is case-specific and depends on group size, both in absolute number and relative to the size of the opponent group, and on the proportion of non-fighters Limitations and Future Directions The study acknowledges limitations such as not selecting videos without violence, making it uncertain if violence is caused by turmoil Developing a proper scale of turmoil for field studies is exceedingly difficult due to variations across domains The Ising model does not predict the severity of violence, and future studies should expand the number and diversity of cases Extensions to norms (as external field-hNm) and noise in actors’ information about others’ behavior have been explored in simulations Data Compilation Details The dataset consists of 42 videos from websites like YouTube, LiveLeak, and WorldStarHipHop using search terms such as “brawl,” “street fight,” and “assault” The sample was random with respect to temporal unfolding and (sub)group size Most groups were small, 2 ≤ n < 10 (mean 3.6), but one had 14 members The coding process involved multiple rounds of discussion to ensure agreement on group memberships, fighters, and bursts The reliability of the coding was assessed using Krippendorff’s alphas which were.77 on average Descriptive Statistics Summary Table S1 provides an overview of the groups, including occurrence of collective violence (c.v.), maximum participation in violence (NC), time lag between the first participant and maximum participation in collective violence, burst, levels of opponents-produced turmoil (T) and of synchrony at the first outbreak of c.v., and number of non-fighters (ND) Table 2 provides descriptive statistics for the indicators used in the analysis The average time lag per participant in non-bursts was 5.12 seconds The level of turmoil prior to max participation in violence averaged 37, which is 97.4% Model Solving Results The Ising model was solved computationally using the Metropolis algorithm and analytically by a mean field analysis The mean field analysis for conditional cooperator i is S¯i = S−P(S−i) + S+P(S+i) = Mcc The self-consistency equation of Mcc is M = pS− + (1 − p)(S0 + ∆ tanh (∆M/T)) The critical proportion of non-fighters pc is reached at the right-hand end of the lines in Figure 13 Solutions of this equation become complex if p > S0∆, hence p ≤ S0∆ The prediction is that there is no (burst of) cooperation if pc > 1/3 Alternative Explanations A plausible alternative explanation for the onset of collective action is synchrony of motion which yields a feeling of oneness among group members The researchers measured the synchronization of focal group members pairwise if they engaged in behaviors like approaching members at the same pace or simultaneously engaging in aggressing Although some degree of synchronization preceded collective violence, it did not predict collective violence The Ising model can explain cooperation parsimoniously, based on agitating stimuli without recourse to strong rationality or norms The model also seems applicable to protests and revolts where agitation is due to the violation of social norms Data Availability All coded video data, the R script used to produce plots from the data (Supplementary Material), and a Fortran script for simulations of the Ising model are available at https://osf.

Title and authors: Kai: So, looking at the suggested improvements, they are basically trying to make this model more robust by testing it against different real-world conditions and adding more complexity <ref:2112.05088#pg6>.

Mira: They are exploring extending those norms as an external field hNm in their simulations, which means they’re asking what happens if the group starts having some implicit social rules instead of just relying on the model's core structure <ref:2112.05088#pg6>.

Lev: From an error-correction standpoint, adding noise in actors’ information about others’ behavior is something we deal with constantly when building systems for real hardware <ref:2112.05088#pg6>.

Kai: So that noise could represent uncertainty about what the other people in the fight are actually doing at any given moment, which is a realistic scenario for an experimental system <ref:2112.05088#pg6>.

Mira: And they're looking at how that noise interacts with the dynamics of turmoil to predict if you get a burst or not <ref:2112.05088#pg6>.

Lev: If we can model that uncertainty, it means we’re moving closer to simulating real human interaction where people aren't perfectly synchronized or perfectly rational <ref:2112.05088#pg6>.

Kai: It helps bridge the gap between the clean Ising model and the messy reality of a physical street fight <ref:2112.05088#pg6>.

Mira: It moves the model from being a purely theoretical tool to something that accounts for more of what’s happening in those confrontations <ref:2112.05088#pg6>.

Lev: But we still have to figure out how to scale these simulations up so they actually reflect the size and dynamics of real groups <ref:2112.05088#pg6>.

Kai: Yeah, that scaling issue is huge. How do you make a model that works for ten people translate to a thousand people in a crowd?

Mira: That’s exactly where the empirical data becomes important, because they need more cases to find those case-specific thresholds like Tc <ref:2112.05088#pg3>.

Lev: If you want to run this on real hardware, you need the dynamics to be fast enough that you can observe the transition from a slow buildup to a burst in real time <ref:2112.05088#pg3>.

Kai: So the paper is suggesting we need better data and more complex simulations to get these predictions into something actionable for understanding social dynamics <ref:2112.05088#pg6>.

The paper's summary: Mira: So, to wrap up the study on "Intergroup violence in bursts," the main implication is that when you’re looking at intergroup confrontation without strong external rules, the structure of cooperation within the group is what determines if violence explodes suddenly or builds up slowly <ref:2112.05088#pg3>.

Kai: It really boils down to that critical proportion of non-fighters versus conditional cooperators being the deciding factor for a burst versus a gradual escalation <ref:2112.05088#pg3>.

Lev: For us working on error correction, this suggests that if we can identify that critical threshold pc, we could potentially predict when a system’s state will suddenly transition into an unstable burst mode <ref:2112.05088#pg3>.

Mira: And the paper shows how this concept of collective action dilemma applies to things beyond just physical fights, like protests and revolts where agitation comes from violating social norms <ref:2112.05088#pg6>.

Kai: It’s a model that uses basic mechanics, like an Ising model, to explain complex human behavior without needing super-rational actors or strict rules <ref:2112.05088#pg3>.

Lev: We need to keep thinking about how external influences, like those norms or the noise in information flow, might shift that critical proportion pc over time <ref:2112.05088#pg6>.

Mira: Overall, this paper provides a formal way to think about sudden collective action by focusing on the internal dynamics of group contribution rather than just external threats <ref:2112.05088#pg3>.

Kai: That’s it for "Intergroup violence in bursts." We’ll keep looking at how these dynamics play out in more complex systems next time.

The paper's improvements: Kai: So, we just looked at how group members who don't want to help contribute predict whether violence happens in bursts or not <ref:2112.05088#pg3>.

Mira: Right, and now they’re suggesting ways to make that model more realistic by adding things like external social rules or noise into the system <ref:2112.05088#pg6>.

Kai: They're saying they can extend the Ising model by treating those social norms as an external field hNm, which is essentially putting some implicit group behavior into the equation <ref:2112.05088#pg6>.

Lev: So, if we treat those norms as a field, it means we’re looking at how those rules change the critical proportion pc that triggers a burst <ref:2112.05088#pg6>.

Kai: And they also mentioned adding noise into how actors know what others are doing, which is like uncertainty about what’s actually happening right now <ref:2112.05088#pg6>.

Mira: That noise could represent the imperfect information people have about their neighbors' actions in a real fight <ref:2112.05088#pg6>.

Lev: If we model that uncertainty, then the system has to account for how that noise interacts with the dynamics of turmoil to predict if you get a burst or not <ref:2112.05088#pg6>.

Kai: That sounds like it moves us closer to simulating real human interaction where people aren't perfectly synchronized or perfectly rational <ref:2112.05088#pg6>.

Mira: Exactly, it takes the clean Ising model and tries to bridge the gap with more of what’s happening in those actual confrontations <ref:2112.05088#pg6>.

Lev: We still have to figure out how to scale these simulations up so they actually reflect the size and dynamics of real groups <ref:2112.05088#pg3>.

Kai: That scaling issue is huge, because we need more cases to find those case-specific thresholds like Tc <ref:2112.05088#pg3>.

Mira: And you also have to worry about the time lags in real life, where local circumstances can cause delays beyond that two-second limit the model has right now <ref:2112.05088#pg3>.

Kai: So, it sounds like the next step is building an agent based model that incorporates those turmoils into individual decision making to distinguish between a gradual buildup and a sudden outbreak <ref:2112.05088#pg4>.

Conclusion: Kai: So, to wrap up, this paper on "Intergroup violence in bursts" shows that we can use an Ising model to predict if violence will explode suddenly based on how many people in a group aren't contributing <ref:2112.05088#pg3>.

Mira: Right, and they showed that the key is finding that critical proportion of non-fighters versus conditional cooperators which determines whether you get a burst or just a slow start <ref:2112.05088#pg3>.

Lev: From my side, it’s interesting how they try to add noise and external social rules to see if that prediction holds up in a more complex environment <ref:2112.05088#pg6>.

Kai: They're looking at how those social norms change the critical proportion pc that triggers a burst, which is a big step because it moves beyond just the raw dynamics <ref:2112.05088#pg6>.

Mira: And they also brought up time lags in real life where local circumstances can cause delays over two seconds which means their current model has limitations there <ref:2112.05088#pg3>.

Lev: If you want to run this on real hardware, that time lag is something we’d have to account for when thinking about the latency of action <ref:2112.05088#pg3>.

Kai: So, it sounds like the next big push is building an agent based model that incorporates those turmoils into individual decision making to see if they can distinguish a gradual buildup from a sudden outbreak <ref:2112.05088#pg4>.

Mira: That’s the way forward, moving from just identifying the pattern to actually simulating how the system reacts when we add those realistic delays and uncertainties <ref:2112.05088#pg6>.

Lev: I think that modeling the noise correctly is where we spend most of our time trying to keep it from being too computationally expensive for real hardware <ref:2112.05088#pg3>.

Kai: So, the whole point of this paper on "Intergroup violence in bursts" is setting up a framework where we can test these complex social dynamics with more realistic parameters <ref:2112.05088#pg3>.

Mira: It gives us a tool to look at collective action without needing strong rationality or norms and it points toward how external influences actually shape those dynamics <ref:2112.05088#pg6>.

Lev: We’ll keep pushing on those simulations, but we’ve got to remember that the fidelity of the measurement is everything when you try to map this onto a physical system <ref:2112.05088#pg3>.