A team of researchers from the University of Southern Carolina and Indiana University set out to determine how many Twitter accounts were bots. It turns out, 15% of all accounts are not humans.
To do so, they created a framework to identify bot accounts. The framework itself utilised more than a thousand tools which looked at timing between tweets, sentiment, and tweet content. Presenting their findings, the researchers estimate that “between 9% and 15% of active Twitter accounts are bots.”
The figures may actually be even higher – the researchers advised that sophisticated bots may have been categorised as humans in their model, so 48 million bot accounts may well be a conservative number.
This new evaluation, is far higher than the figures previously released by Twitter. In a filing last month, the company said only 8.5% of all accounts do not showcase, in their own words, “discernible additional user-initiated action”.
A Twitter spokesperson commented that while bots have largely negative connotations, many are useful to “automatically alert people in case of natural disasters” or for customer service purposes. While the USC researchers supported that statement, they insisted that a large chunk of bot accounts perform toxic functions. Fake political support or promotion of terrorist propaganda are just two examples amongst many.
In recent weeks, Twitter has manoeuvred itself to make the reporting of abusive activity easier, providing more ways for users to rid Twitter of trolls and bots. However, considering that nearly 50 million accounts are not humans, Twitter has a real mountain to climb. And these new findings have made that mountain a lot higher.