A new study has shown that up to two thirds of links shared on micro-blogging site Twitter come from automated accounts.
It has been common knowledge for a while now that Twitter has a bot problem, with millions of automated accounts retweeting and sharing content. Research from the Pew Research Centre highlights the extent of the issue and the role in which these bot accounts play in spreading links to popular websites.
Pew released a report on Monday this week after researchers analysed over one million tweets from just over 140,000 Twitter accounts that were posted between 27th July and 11th September 2017. The study found that these tweets shared links to a total of 2,315 different popular websites, and that bots played a more significant role in sharing these links than humans did.
Out of all the tweeted links analysed, a total of 66% were shared by accounts that appeared to be automated. Furthermore, 22% of tweets that posted links to popular news sites and event sites were posted by some of the top 500 most active bot accounts. This is in comparison to the most active 500 human accounts, which were only responsible for a total of 6% of the tweeted links pointing towards this type of site.
To determine whether an analysed account was a bot or not, Pew used a tool named Botometer, which was developed by a team of professionals from the University of Southern California and Indiana University, which estimated the probability of an account being a bot. To calculate this probability, factors such as the age of the account, its followers and frequency of posting were all taken into consideration. Should an account score 43% or higher, it would be classified as a bot, although scores between the 40%-60% range can be difficult to classify. After testing, both Pew and Botometer found that the threshold of 43% and above provided the most accurate assessments.
As the spreading of fake news in terms of politics has been a taking point surrounding Twitter bots, Pew also looked at how often the suspected bot accounts tweeted links to political sites with clear political stances, and it found that there was no concrete evidence of political bias. The bots tweeted links to centrist and mixed audiences between 57% and 66% of the time, while tweeting links to sites to liberal and conservative sites 41%-44% of the time.
Although it seems that bots are doing most of the sharing, Pew states that it didn’t analyse the truthfulness of the content being tweeted, whether the bots were ‘good’ or ‘bad’, the location of the bots or the level of human engagement attracted by the tweets.
Pew released the below video detailing its methodologies and findings:
- Chrome to warn users regarding insecure web forms - August 20, 2020
- Google trials virtual business cards in India - August 14, 2020
- DuckDuckGo claims Google market share would drop if users given choice - August 11, 2020
