Repression, Emotion, Contagion : Analyzing the (De)Escalation of Protest Violence

A surge in violent protests has spread across the globe in recent years. Governments have responded by deploying various tactics to de-escalate conflicts, but not all measures are effective. Under some circumstances, protest policing has backfired, provoking widespread anger and protester defiance in response to the repression. The literature continues to debate why people engage in violent protests, how they react to the risk of repression, and under which conditions policing tactics de-escalate violence. Taking into account individual psychology and group dynamics, we propose that: 1) police tactics can influence individual stances on protest violence, but individuals’ emotions mediate this influence; and 2) collective identity, appraisals of public support, and violence contagion can also sway people’s stance on protest violence, offsetting the impact of police tactics.

To examine the above propositions, we trained recurrent neural networks to detect the emotional states, violent inclination, and group identity strength of the online forum LIHKG, which was the main mobilizing platform during the 2019 Hong Kong Protests. Our big data and regression analyses reveal emotions are critical mediators between police tactics and endorsement of violent protest. Anger, disgust, and awe can contribute to the endorsement of violence, while fear and sadness can attenuate such inclination (study 1). Against this backdrop, crowd control agents could indirectly escalate violence in protests by provoking anger and reducing fear. In contrast, mass arrest could deter the escalation of violence by inducing fear and decreasing disgust (studies 2 and 3). Mass protest can effectively relieve all negative emotions, and therefore protest ban would risk escalating violent conflicts further. Besides, violence contagion also plays an important role, but it only exerts influence when an individual’s collective identity is salient (study 2). For activists/sympathizers who do not have a strong collective identity, the violence contagion will not sway their minds about violent protest. To assure the robustness of model results, we conduct a sensitivity analysis on the previous three models (study 4).

Taken together, our findings imply that violence de-escalation efforts should not focus solely on short-term issues, such as crowd dispersion. Instead, they should also account for social contagion and the erosion of trust in the long run. This research offers three contributions to the knowledge of conflict processes. First, it disentangles the previously inconclusive interplay between repression and violence inclination by shedding light on the mediating roles played by emotions and contagions. Second, we investigated diverse emotions and a wide range of contextual factors. Little attention has been paid to emotions other than anger,7,8 although many scholars have suggested that threat-related emotions, such as fear, could be as important as anger during social unrest in non-democracies. At the same time, previous studies on protest violence, typically focused on individual factors while ignoring contextual factors and the interchange between individual and contextual factors. Finally, and relatedly, we drew on a large amount of observational data, which allowed us to deploy richer models with multiple levels and interesting cross-level interactions to unveil nuances that might have otherwise been ignored or confounded.