A new study in Frontiers in Physics has exposed the presence of short-term behavioral tendencies in humans, which are absent in social media bots, offering an example of a human signature on social media that could be leveraged to develop more refined bot detection strategies. The analysis is the first study to apply person behavior over a social media session to the problem of bot detection.
In this work, the researchers studied how the behavior of people and bots changed over the course of an activity session utilizing a large Twitter dataset associated with current political events. Over the course of those sessions, the researchers measured various factors to seize consumer behavior, along with the propensity to engage in social interactions and the amount of produced content, after which compared these results between bots and people.
To study the behavior of bot and human users over an exercise session, the researchers focused on indicators of the quantity and high quality of social interactions a user engaged in, along with the number of retweets, replies, and mentions, in addition to the length of the tweet itself.
They then used these behavioral outcomes to inform a classification system for bot detection to observe whether the inclusion of options describing the session dynamics could enhance the performance of the detector.
A range of machine learning techniques had been used to coach two completely different units of classifiers: one including the features explaining the session dynamics and one with-out those options, as a baseline.