You talkin’ to me? Exploring Human/Bot Communication Patterns during Riot Events

Highlights：

• Human-human and human-bot conversation on Twitter can be characterized by specific patterns that emerge as bot and human accounts exchange emotion-conveying messages.

• These patters come in form of statistically-significant subgraphs that we call emotion-exchange motifs, which are an extension of the traditional network motifs.

• There are distinct emotion-exchange motifs that are characteristic for a human-like communication. These motifs include self-loops, reciprocal edges, and transitive triads.

• Human users tend to use self-loops when communicating anger, a mechanism used to bypass the 140-character restriction on Twitter.

• As they communicate with humans, bot accounts form only a smaller subset of emotion-exchange motifs that, unlike the ones found in a human-human communication, involve only one-way edges (message chain motifs, broadcasting motifs, and a message-receiver motifs without reciprocal edges).

• Specific to the events considered in this study (riot events), bots were responsible for a dissemination of messages conveying fear. These messages receives a considerable high number of retweets in our data-sets. The use of fear is also evident in the fear-exchange motifs where bots consistently take over a message-sender role (whereas a human is consistently a message receiver).

摘要

•Human-human and human-bot conversation on Twitter can be characterized by specific patterns that emerge as bot and human accounts exchange emotion-conveying messages.•These patters come in form of statistically-significant subgraphs that we call emotion-exchange motifs, which are an extension of the traditional network motifs.•There are distinct emotion-exchange motifs that are characteristic for a human-like communication. These motifs include self-loops, reciprocal edges, and transitive triads.•Human users tend to use self-loops when communicating anger, a mechanism used to bypass the 140-character restriction on Twitter.•As they communicate with humans, bot accounts form only a smaller subset of emotion-exchange motifs that, unlike the ones found in a human-human communication, involve only one-way edges (message chain motifs, broadcasting motifs, and a message-receiver motifs without reciprocal edges).•Specific to the events considered in this study (riot events), bots were responsible for a dissemination of messages conveying fear. These messages receives a considerable high number of retweets in our data-sets. The use of fear is also evident in the fear-exchange motifs where bots consistently take over a message-sender role (whereas a human is consistently a message receiver).