According to a recent study, more than four times more tweets were made from automated accounts in favor of Donald Trump around the first presidential debate than from those supporting rival Hillary Clinton.
The study found that Donald Trump would have enjoyed more support on Twitter, even if the accounts – known as bots – were not active.
However, the study emphasizes that software has the ability to “manipulate public opinion” and “muddy political issues.”
The report has not yet been reviewed.
One critic noted that it is impossible to be completely sure which accounts are real and which are “web robots”.
The study was led by Professor Philip Howard of Oxford University and is part of a broader project exploring “computational propaganda.”
The investigation covers tweets posted on September 26, the day of the debate, plus the three days after, and relies on popular hashtags related to the event.
First, researchers identified accounts that published exclusively messages containing hashtags related to one candidate but not the other.
They represent about 1.8 million pro-Trump tweets and 613,000 publications in support of Clinton.
The researchers then analyzed which of them were published by bots. They identified an account as such if it tweeted at least 50 times a day during the period, which means a minimum of 200 tweets over the four days.
The results suggest that 32.7% of these tweets in support of Trump were posted by bots and 22.3% of those in support of Clinton.
In total, this represents a total of 576,178 tweets in favor of Donald Trump and 136,639 in support of Hillary Clinton.
Prof. Philip Howard said: “As for the probability that if you withdraw a highly automated account, the chances are four to one that you will find that this is a bot on Twitter in favor of Trump.”
However, there is no suggestion that the bots were generated by any of the official presidential campaign groups.
“We don’t look at the source, who is working on the bots or to what purpose, but only the data indicators,” said Prof. Philip Howard.
Looking more broadly – at accounts that have posted Twitter neutral hashtags or a combination of different types – the study suggests that 23% of all tweets were managed by bots.
One machine learning expert warns that the criteria used to identify bots may have been too inaccurate to weed out all human activity.
So, is it possible that Donald Trump’s supporters were just more enthusiastic than Hillary Clinton’s and did a better job of taking advantage of social media to their advantage?
Prof. Philip Howard said that was unlikely to be the only explanation.
“Most of the heavy automation and tweets happened overnight and shared similar hashtags and information.” he says.
“They show behavior that is not human and often have no comments [about other issues apart from] the specific topic in question. “
Prof. Philip Howard added that the rule of 50 tweets a day was confirmed by an analysis of posts made during the last elections in Venezuela and the Brexit vote.
In both cases, his team re-checked a sample of accounts that were marked as bots and confirmed that they showed other inhuman characteristics.