Understanding the dynamics of social media influence : empirical analysis of the determinants of retweeting

This graduate thesis is part of a wider research project whose aim is to analyze influence dynamics on social media. Literature focused its efforts into studying the role of influencers, i.e. users who have a broad audience (e.g. they possess many followers on Twitter). The term influence, instead, refers to the social impact of the content shared, regardless of who published it: the key point is the ability of the message subject to raise audience attention on its own. Though we are aware of the importance of being an influencer, our assertion is that message content possesses a decisive role in generating influence, irrespective of its author. Hypotheses testing is here performed with the aim of evaluating content significance in influence generation. An in-house software was built in order to support all the stages this thesis work consists of. The first step assesses the weight of content specificity (i.e. level of detail) at user level, considering also tweeting volumes. Empirical results highlight a positive correlation between specificity and influence, while high volumes show a strongly negative connection with the possibility of being retweeted. Even when the popularity of the user is taken into account, specificity is shown to keep holding a positive effect over messages distribution. The following section analyzes influence dynamics at single post level, without the bias of author variables: this is a crucial stage, where a clear distinction between influence and influencers is performed. Sentiment (i.e. feelings being conveyed) and specificity are the employed variables. Data show a perfect fit to the model, validating the positive relationship between specificity and influence. As regards sentiment, the need of a few negative messages is displayed while seeking for a larger amount of retweets. The final step of the work exploits data clustering, with the intention of verifying at which level specificity stops playing a significant role in influence generation. Empirical findings are converted into guidelines, useful for both private users and corporations as a starting point for building a self-promoting strategy.