Analyzing the Engagement of Social Relationships During Life Event Shocks in Social Media

TL;DR

The paper addresses how online social relationships modulate responses to life shocks on Twitter. It combines a large, labeled dataset of shock disclosures with a quasi-causal framework (propensity score matching and diff-in-diff) and topic modeling to compare response rates, content, and topic shifts across five relationship types and four shock categories. Key contributions include a 13K-shock-event Twitter dataset with labeled dyads, evidence that response magnitude and content are relationship- and shock-dependent, and insights into how tie strength and embeddedness differentially influence online support. The work advances understanding of online social dynamics and suggests relationship-aware strategies to mobilize support in digital communities, with practical implications for design of interventions and social platforms.

Abstract

Individuals experiencing unexpected distressing events, shocks, often rely on their social network for support. While prior work has shown how social networks respond to shocks, these studies usually treat all ties equally, despite differences in the support provided by different social relationships. Here, we conduct a computational analysis on Twitter that examines how responses to online shocks differ by the relationship type of a user dyad. We introduce a new dataset of over 13K instances of individuals' self-reporting shock events on Twitter and construct networks of relationship-labeled dyadic interactions around these events. By examining behaviors across 110K replies to shocked users in a pseudo-causal analysis, we demonstrate relationship-specific patterns in response levels and topic shifts. We also show that while well-established social dimensions of closeness such as tie strength and structural embeddedness contribute to shock responsiveness, the degree of impact is highly dependent on relationship and shock types. Our findings indicate that social relationships contain highly distinctive characteristics in network interactions and that relationship-specific behaviors in online shock responses are unique from those of offline settings.

Figures (5)

• Figure 1: Comparison of metrics on the test set after performing additional rounds of active learning.
• Figure 2: Comparison of response metrics between shock tweets and control tweets. Shock tweets receive a substantially larger amount of retweets, replies, likes, and quotes compared to the control tweets.
• Figure 3: The average number of neighbors of each predicted relationship type for shock type.
• Figure 4: Changes in the volume of replies to shocked users across 3-hour blocks, measured using diff-in-diff relative to control users. The x-axis is the number of hours relative to the shock (e.g., the first hour of the shock corresponds to x=1). Statistically significant values (Bonferroni corrected) are colored in orange if before the shock and red if after the shock. The red dashed line indicates the time of posting the shock tweet.
• Figure 5: Coefficient sizes obtained from a logistic regression of response on mention frequency (tie strength) and Jaccard similarity (structural embeddedness). Values with solid colors indicate statistically significant (Bonferroni-corrected) coefficient values.