Social networks such as Twitter and Facebook come in various forms depending on the cohesiveness and size - from the most intimate to tenuous relationships. In the context of Twitter, the flexibility of establishing connections, such as a directed tie like following, enables the proliferation of tenuous relationships. This study observes that the implication of such flexibility poses challenges to data mining tasks, such as detection of socially cohesive groups, or content veracity. A small group of interconnected users or Simmelian ties are more intimate with a high degree of familiarity due to strong social cohesion. Such groups are considered homogeneous for many socio-demographic, behavioural, and intrapersonal characteristics. In the context of content veracity, anecdotal and cognitive evidence suggests that users are more likely to believe information shared by closely related individuals. Thus, the study is based on the premise that by recognising users who reciprocate friendships, some of the challenges will be mitigated. However, in social platforms such as Twitter, where flexible and transitory connections are prevalent, it is challenging to identify Simmelian ties. In this study, we present an empirical analysis of datasets consisting of 9300 Simmelian ties retrieved from over 30m Twitter accounts. Noting the challenges in identifying reciprocal relationships on a large scale, we propose a useful prediction model. As a result, the detection of socially cohesive communities is enhanced, thus providing a valuable analysis tool and strengthening the validity of online content. To evaluate the efficacy of the approach, we apply two state-of-the-art community detection algorithms on different datasets and achieve promising results. We further describe how to enhance content veracity and information diffusion by leveraging Simmelian connections. To the best of our knowledge, this study provides the first large scale dataset of Simmelian ties on Twitter.
