Have You Heard?: How Gossip Flows Through Workplace Email

We spend a significant part of our lives chatting about other people. In other words, we all gossip. Although sometimes a contentious topic, various researchers have shown gossip to be fundamental to social life—from small groups to large, formal organizations. In this paper, we present the first study of gossip in a large CMC corpus. Adopting the Enron email dataset and natural language techniques, we arrive at four main findings. First, workplace gossip is common at all levels of the organizational hierarchy, with people most likely to gossip with their peers. Moreover, employees at the lowest level play a major role in circulating it. Second, gossip appears as often in personal exchanges as it does in formal business communication. Third, by deriving a power-law relation, we show that it is more likely for an email to contain gossip if targeted to a smaller audience. Finally, we explore the sentiment associated with gossip email, finding that gossip is in fact quite often negative: 2.7 times more frequent than positive gossip. Introduction a Email 1: hey seems like we aren’t the only ones that think kyle is an arrogant asshole anymore – susan told me that on saturday night she had it out with him and doesn’t want him around – also, apparently he’s been treating dana like shit and it’s starting to get noticed by other people – just thought this was an interesting development. Email 2: Here’s the third party assessment of current western supply issues that we’re most in agreement with. Sam Van Vactor’s group publishes the daily energy market report that’s widely read, and Pickel is with Tabors Caramanis, consultants that we have employed on several issues. In both of these messages, the sender discusses someone who is not on the email. Anthropologists call conversations like these gossip: the absence of a third party from the conversation (Besnier 1989; Hannerz 1967). Despite some negative social connotations, gossip is fundamental to healthy societies—from small groups to large, formal organizations (Feinberg et al. 2012). Simply put, we use it to trade social information, information we may find very useful in the future. Copyright © 2012, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. In fact, Dunbar (1994) goes so far as to suggest that language itself developed so we could gossip about one another. Entire technologies have even arisen to support the practice, such as the dumbwaiter, a rope-and-pulley system designed in the eighteenth century to guard the masters’ gossip from nosy servants (Goffman 1959). Following in this line of scholarship, this paper presents the first study of gossip in a large corpus of computer-mediated communication (CMC). Using the Enron email dataset of 517,431 messages, we look to answer the following research questions. In systems characterized by power and hierarchy— like workplaces—what role does hierarchy play in shaping how people gossip? Going further, can we infer someone’s corporate rank from their gossip behavior? Is gossip limited to personal email exchanges, or does it leak into more formal business communication? How does the size of an email’s recipient list affect the likelihood of gossip in the body of the message? And finally, do certain phrases and emotions characterize gossip via email? Though sometimes overlooked in an always-changing internet, email was the internet’s first widespread social medium (Henderson Jr. and Myer 1977). Email affords conversations among both small and large groups. Networks of contacts form over time, like Twitter. Unlike Twitter, however, 92% of online adults use email (Purcell 2011). Madden and Jones (2008) recently reported a sharp increase in the number of adults who “constantly” check their work email, a figure that has almost certainly risen as smartphones find their way into more and more pockets. In other words, it may be fair to call email the world’s most successful and pervasive type of social media. With this as a backdrop, we turn to natural language methods—specifically, Named Entity Recognition—to identify gossip in the Enron corpus. We find it present at all levels of the corporate hierarchy. We demonstrate hierarchical signatures of gossip, showing specific pathways for the transmission of gossip via email. People belonging to certain ranks are the major sources of these messages, while other ranks silently receive it. Yet others do both. We find that people gossip most with their peers, indicating their tendency to gossip within their own group, the ones belonging to the same rank. Interestingly, people have a greater likelihood to send gossip messages to smaller audiences: a fact demonstrated by deriving a power law relation between the frequency of gossip email and the number of recipients on an email. After exploring gossip as framed by hierarchical structure, we take a closer look at the content of gossip messages. Using sentiment analysis, we search for emotional signals in gossip. We explore the sentiment associated with gossip email, finding that gossip is in fact quite often negative: 2.7 times more frequent than positive gossip. We see the primary contribution of this paper as an exploratory study of an important social process, albeit one that is sometimes hidden.