Twitter is a major platform for political communication and campaigning, used extensively by political parties, lobbying groups, and activists in many democratic debates around the world.
It has proven to be vulnerable to traffic manipulation by small but coordinated user groups, and those who control automated accounts, known as “bots”. These have demonstrated an ability to materially distort Twitter traffic, forcing chosen phrases and hashtags into the “trending” lists, and generating very high volumes of traffic from a very small base of human users.
This paper proposes a computational method to calculate the extent to which a given flow of Twitter traffic has been subject to manipulation by such groups. It examines three indicators:
- the average number of tweets (including retweets) per user;
- the percentage of retweets as a proportion of total traffic;
- the proportion of traffic generated by those fifty accounts which used the given term most often.
These three factors indicate whether Twitter traffic was generated organically by a large number of users or pushed by a small one; whether it was driven by a high proportion of original posts, or by large-scale retweeting; and whether it was driven by a small user group, or a broader movement.
We combine the three factors into a single metric, the Coefficient of Traffic Manipulation (CTM). This is a relative measure, rather than an absolute one. It allows us to compare different Twitter traffic flows against measurable criteria and assess which of those movements appear to have been most subject, or least subject, to manipulation.
As such, it can serve as an early-warning system for researchers, indicating Twitter traffic flows which do appear to have been manipulated. These can then be subject to further study, to determine what manipulation measures were used.
Ben Nimmo, “Measuring Traffic Manipulation on Twitter.” Working Paper 2019.1. Oxford, UK: Project on Computational Propaganda. comprop.oii.ox.ac.uk. 35 pp.