Bridging or Breaking: Impact of Intergroup Interactions on Religious Polarization

TL;DR

The paper investigates how intergroup interactions influence religious polarization on social media during crises, using Indian COVID-19 discourse and inferring religion from names. It introduces the Group Conformity Score ($GCS$) based on contextualized tweet embeddings and compares it to a Bag-of-Words polarization measure, applying a meta-learning framework to estimate heterogeneous treatment effects across seven events. Through a T-Learner and Oaxaca-Blinder decomposition, it shows that intergroup interactions generally reduce polarization, but can backfire for communal events (e.g., Tablighi) and political contexts, with Muslims experiencing increased conformity in certain settings. The work highlights context-aware metrics and a scalable framework for monitoring polarization, with implications for platform design and broader applicability beyond the Indian context.

Abstract

While exposure to diverse viewpoints may reduce polarization, it can also have a backfire effect and exacerbate polarization when the discussion is adversarial. Here, we examine the question whether intergroup interactions around important events affect polarization between majority and minority groups in social networks. We compile data on the religious identity of nearly 700,000 Indian Twitter users engaging in COVID-19-related discourse during 2020. We introduce a new measure for an individual's group conformity based on contextualized embeddings of tweet text, which helps us assess polarization between religious groups. We then use a meta-learning framework to examine heterogeneous treatment effects of intergroup interactions on an individual's group conformity in the light of communal, political, and socio-economic events. We find that for political and social events, intergroup interactions reduce polarization. This decline is weaker for individuals at the extreme who already exhibit high conformity to their group. In contrast, during communal events, intergroup interactions can increase group conformity. Finally, we decompose the differential effects across religious groups in terms of emotions and topics of discussion. The results show that the dynamics of religious polarization are sensitive to the context and have important implications for understanding the role of intergroup interactions.

Figures (18)

• Figure 1: 7-day Exponential Moving Average of daily polarization estimated using contextualized approach $\hat{\pi}^{LO}$ vs. bag-of-words approach $\hat{\pi}^{LO, BOW}$ along with the number of COVID-related tweets by both religious groups. The COVID-related events are marked with green vertical lines and major festivals are marked with yellow vertical lines.
• Figure 2: Decomposition of difference in the effect of interaction on $GCS$ between Muslims and non-Muslims using Oaxaca-Blinder decomposition. The red bars show the extent to which the effect is explained by topics and emotions. The error bars represent 95% confidence intervals.
• Figure 3: Sensitivity, specificity, Youden index, and geometric mean by prediction threshold.
• Figure 5: Coefficient Plot of Covariates when Treatment Effect is regressed on them.
• Figure 6: Contribution of emotions and topics towards the explained component of difference in the effect of interaction on $\Delta GCS$ across Muslims and non-Muslims using Oaxaca-Blinder decomposition. Errors bars show 95% confidence intervals.