[Paper Review] Signed Networks in Social Media

Jure Leskovec, Daniel P. Huttenlocher+1 more|arXiv (Cornell University)|Mar 11, 2010

Opinion Dynamics and Social InfluencePhysics and Astronomy18 references157 citations

TL;DR

This paper investigates signed social networks in online platforms like Epinions, Slashdot, and Wikipedia, analyzing how positive and negative relationships shape network structure. Using structural balance and status theories, it finds that while weak structural balance explains undirected triad patterns, status theory better predicts directed link signs, revealing that users' perceptions of social status drive link formation more than balance alone.

ABSTRACT

Relations between users on social media sites often reflect a mixture of positive (friendly) and negative (antagonistic) interactions. In contrast to the bulk of research on social networks that has focused almost exclusively on positive interpretations of links between people, we study how the interplay between positive and negative relationships affects the structure of on-line social networks. We connect our analyses to theories of signed networks from social psychology. We find that the classical theory of structural balance tends to capture certain common patterns of interaction, but that it is also at odds with some of the fundamental phenomena we observe --- particularly related to the evolving, directed nature of these on-line networks. We then develop an alternate theory of status that better explains the observed edge signs and provides insights into the underlying social mechanisms. Our work provides one of the first large-scale evaluations of theories of signed networks using on-line datasets, as well as providing a perspective for reasoning about social media sites.

Motivation & Objective

• To understand how positive and negative relationships coexist and shape large-scale online social networks.
• To evaluate classical structural balance theory and an alternative status theory in explaining signed network structures in real-world online platforms.
• To investigate whether the interplay between positive and negative links reflects social psychological mechanisms like trust, distrust, and perceived status.
• To compare how different platforms (Epinions, Slashdot, Wikipedia) use signed edges and how this affects network structure and dynamics.
• To provide one of the first large-scale empirical validations of signed network theories using real online datasets.

Proposed method

• Analyzing three large-scale online datasets: Epinions (trust/distrust), Slashdot (friend/foe), and Wikipedia (support/oppose votes) to extract signed, directed network structures.
• Applying structural balance theory to predict the frequency of different signed triads, particularly focusing on triangles with one, two, or three positive edges.
• Testing status theory, which posits that a positive link from A to B indicates A perceives B as having higher status, and negative links indicate lower perceived status.
• Comparing observed network structures with randomized versions to assess whether deviations from chance are statistically significant.
• Measuring the impact of edge sign visibility on outlier behavior (e.g., users with unusually high in- or out-degree) across platforms.
• Using mutual relationships with third parties to infer status levels and predict link signs, validating the status model’s predictive power.

Experimental results

Research questions

• RQ1 To what extent does structural balance theory explain the distribution of positive and negative links in real online signed networks?
• RQ2 How do the signs of directed links in online social networks align with predictions from status theory versus structural balance theory?
• RQ3 How does the visibility of link signs (public vs. private) affect user behavior, such as the prevalence of highly active users in positive or negative linking?
• RQ4 What role do common neighbors play in influencing the likelihood of a positive link between two users?
• RQ5 How do differences in platform design (e.g., explicit friend/foe tagging vs. implicit voting) affect the structure and dynamics of signed networks?

Key findings

• Triangles with exactly two positive edges are massively underrepresented in all three datasets compared to random expectations, supporting a weak form of structural balance.
• Triangles with three positive edges are significantly overrepresented, indicating a tendency for clusters of mutual trust or friendship to form.
• Users with multiple common neighbors (of any sign) are significantly more likely to form positive links, linking balance to social capital theory.
• In directed networks, structural balance theory fails to explain link signs, while status theory accurately predicts deviations in link signs based on users’ mutual relationships with third parties.
• On Wikipedia, where link signs are highly public, the fraction of users with high positive in-degree exceeds random expectations, suggesting conformity to existing positive outcomes.
• On Epinions and Slashdot, where signs are less public, the fraction of users with high negative out-degree is lower than expected, indicating suppression or avoidance of prolific negative evaluators.