Interpersonal conflicts occur frequently in both offline and online groups, with conditions for conflict especially ripe online. This research attempts to understand the consequences of online group conflict and reporting it to group administrators, both for the protagonists in the conflict and observers. If group conflict is aversive, then group members should reduce their group participation after observing conflict. Theories of imitation and behavioral mimicry suggest that even onlookers will exhibit more conflict and negative language after observing conflict conversations in their group. In contrast, theories of deterrence suggest that both the instigator of the conflict and onlookers will reduce their conflict and onlookers might even increase their engagement if conflicts are reported to group administrators. The current study uses de-identified and aggregated data from Facebook group conversations and Mahalanobis distance matching to test these ideas. Results are consistent with the hypothesis that conflict in group conversations reduces engagement within the group and increases the amount of conflict and the negativity of language users express in the group. However, inconsistent with deterrence theories, conflict and language negativity increase and group engagement decreases when conflict is reported to group administrators.

With the rapid development of digital platforms, users can now interact in endless ways from writing business reviews and comments to sharing information with their friends and followers. As a result, organizations have numerous digital social networks available for graph learning problems with little guidance on how to select the right graph or how to combine multiple edge types. For example, while user-to-user interactions are directed in nature, many graph learning approaches use the undirected version of the network. In this paper, we introduce edge direction, edge weight, and multi-relational data for node prediction tasks. We first adapt an existing node attribute prediction method for binary prediction, LINK-Naive Bayes, to account for both edge direction and weights on single-layer networks. We compare predictive performance metrics across various node attribute prediction tasks for an ads click prediction task on Facebook and for a publicly available dataset from the Open Graph Benchmark (OGB). We observe meaningful predictive performance improvements when incorporating edge direction and weight, and performance that's competitive with the OGB Leaderboard. We then introduce an approach called MultiLayerLINK-NaiveBayes that can combine multiple network layers during training and observe superior performance over the single-layer results. Ultimately, whether edge direction, edge weights, and multi-layers are practically useful will depend on the particular setting. Our approach enables practitioners to quickly combine multiple layers and edge types.

With the rise of social media, users from across the world are able to connect and converse with each other online. While these connections have facilitated a growth in knowledge, online discussions can also end in acrimonious conflict. Previous computational studies have focused on creating online conflict detection models from inferred labels, primarily examine disagreement but not acrimony, and do not examine the conflict’s emergence. Social science studies have investigated offline conflict, which can differ from its online form, and rarely examines its emergence. The current research aims to understand how online conflicts arise in online personal conversations. Our ground truth is a Facebook tool that allows group members to report conflict to administrators. We contrast discussions ending with a conflict report with paired non-conflict discussions from the same post. We study both user characteristics (e.g., historical user-to-user interactions) and conversation dynamics (e.g., changes in emotional intensity over the course of the conversation). We use logistic regression to identify the features that predict conflict. User characteristics such as the commenter’s gender and previous involvement in negative online activity are strong indicators of conflict. Conversational dynamics, such as an increase in person-oriented discussion, are also important signals of conflict. These results help us understand how conflicts emerge and suggest better detection models and ways to alert group administrators and members early on to mediate the conversation.

Social network platforms have enabled large-scale measurement of user-to-user networks such as friendships. Less studied is user sentiment about their networks, such as a user’s satisfaction with their number of friends. We surveyed over 85,000 Facebook users about how satisfied they were with their number of friends on Facebook, connecting these responses to their on-platform activity. As suggested in prior work, we’d expect users who are not satisfied with their friend count to have a higher probability of experiencing the friendship paradox: “your friends have more friends than you”. However in our sample, among users with more than 3,500 friends, no user experiences the friendship paradox. Instead, we still observe that those users with more friends would prefer to have even more friends. The friendship paradox also contributes to local perception bias, defined as the difference between the average number of friends among a user’s friends and the average friend count in the population. Users with a positive perception bias – their friends have more friends than others – are less satisfied with their friend count. We then introduce a weighted perception bias metric that considers the fact that different friends have different effects on an individual’s perception. We find this new weighted perception bias better distinguishes friend count satisfaction outcomes for users with high friend count when compared to the original perception bias metric. We conclude with modeling the behavior interactions via a machine learning model, demonstrating the heterogeneity in the interactions across users with different perception biases. Altogether, these findings offer more insights on users’ friend count satisfaction, which may provide guidelines to improve the user experience and promote healthy interactions.

Randomized experiments, or “A/B” tests, remain the gold standard for evaluating the causal effect of a policy intervention or product change. However, experimental settings, such as social networks, where users are interacting and influencing one another, may violate conventional assumptions of no interference for credible causal inference. Existing solutions to the network setting include accounting for the fraction or count of treated neighbors in a user’s network, yet most current methods do not account for the local network structure beyond simply counting the number of neighbors. Our study provides an approach that accounts for both the local structure in a user’s social network via motifs as well as the treatment assignment conditions of neighbors. We propose a two-part approach. We first introduce and employ “causal network motifs”, which are network motifs that characterize the assignment conditions in local ego networks; and then we propose a tree-based algorithm for identifying different network interference conditions and estimating their average potential outcomes. Our approach can account for social network theories, such as structural diversity and echo chambers, and also can help specify network interference conditions that are suitable to each experiment. We test our method on a synthetic network setting and on a real-world experiment on a large-scale network, which highlight how accounting for local structures can better account for different interference patterns in networks.

Social media provides the government with novel methods to improve regulation. One leading case has been the use of Yelp reviews to target food safety inspections. While previous research on data from Seattle finds that Yelp reviews can predict unhygienic establishments, we provide a more cautionary perspective. First, we show that prior results are sensitive to what we call “Extreme Imbalanced Sampling”: extreme because the dataset was restricted from roughly 13k inspections to a sample of only 612 inspections with only extremely high or low inspection scores, and imbalanced by not accounting for class imbalance in the population. We show that extreme imbalanced sampling is responsible for claims about the power of Yelp information in the original classification setup. Second, a re-analysis that utilizes the full dataset of 13k inspections and models the full inspection score (regression instead of classification) shows that (a) Yelp information has lower predictive power than prior inspection history and (b) Yelp reviews do not significantly improve predictions, given existing information about restaurants and inspection history. Contrary to prior claims, Yelp reviews do not appear to aid regulatory targeting. Third, this case study highlights critical issues when using social media for predictive models in governance and corroborates recent calls for greater transparency and reproducibility in machine learning.

Graph smoothing methods are an extremely popular family of approaches for semi-supervised learning. The choice of graph used to represent relationships in these learning problems is often a more important decision than the particular algorithm or loss function used, yet this choice is less well-studied in the literature. In this work, we demonstrate that for social networks, the basic friendship graph itself may often not be the appropriate graph for predicting node attributes using graph smoothing. More specifically, standard graph smoothing is designed to harness the social phenomenon of homophily whereby individuals are similar to “the company they keep.” We present a decoupled approach to graph smoothing that decouples notions of “identity” and “preference,” resulting in an alternative social phenomenon of monophily whereby individuals are similar to “the company they're kept in,” as observed in recent empirical work. Our model results in a rigorous extension of the Gaussian Markov Random Field (GMRF) models that underlie graph smoothing, interpretable as smoothing on an appropriate auxiliary graph of weighted or unweighted two-hop relationships.