/ 28 September 2021

How to game Twitter’s algorithm – and hoodwink journalists

The Law Council of Australia warned the legislation could have "serious unintended consequences".
(Jonathan Nackstrand/AFP)

ANALYSIS

We all know that social media is not the real world. In theory, anyway. But sometimes, in the face of coordinated disinformation campaigns, it can be difficult to understand the difference. This story outlines some of the tactics used by unscrupulous politicians and lobbyists to manipulate the conversation on Twitter. As social media users, we all need to know how this works so that we don’t get fooled.

In March 2020 Bill Gates trended on Twitter in South Africa, after claims that he wanted to test his vaccine on Africans began circulating. People were outraged, and the claims made their way into conventional media.

A few days later, prominent South African news organisation News24 retracted a story headlined “Bill Gates confident a potential coronavirus vaccine will work in Africa, but Twitter does not think so”, and issued an apology to the Bill and Melinda Gates Foundation.

This story is a perfect illustration of how social media algorithms can be manipulated to trick users, readers and newsrooms.

The controversy was originally posted on Facebook, not Twitter, in French rather than English, and originated in the Democratic Republic of Congo, not South Africa.

The post started with a group of Facebook pages as part of a coordinated attempt to promote a local Congolese politician. They were later removed. The content claimed that controversial French physician Didier Raoult had warned Africans not to accept a vaccine from Bill Gates (Raoult’s employer denied he authored the claim, and Facebook later labelled the post as containing false information after AFP debunked it.)

The Congolese Facebook posts were shared tens of thousands of times, and subsequently picked up by clickbait websites and auto-translated into English. A link to the translated version on a clickbait website named EN24.news was shared by US politician Cynthia McKinney on Twitter on 29 March. Two days later the anonymous South African Twitter account known as @LandNoli retweeted McKinney’s link to the auto-translated disinformation piece by EN24.

Anti-Gates sentiment quickly started trending on South African Twitter. When President Cyril Ramaphosa posted a link to an interview Gates did with The Daily Show host Trevor Noah on 4 April this sentiment flared up even more: Responses to Ramaphosa’s tweet were flooded with anti-vaccine and anti-Gates narratives.

Prominent anonymous accounts used the tweet as evidence that the Bill and Melinda Gates Foundation was meddling in South Africa’s health system. This tweet and the anti-Gates sentiment that had already been circulating on Twitter were the source of the now-retracted News24 story, according to editor-in-chief Adriaan Basson.

The Congolese Facebook posts were shared tens of thousands of times, picked up by clickbait websites and auto-translated into English. A link to the translated version was shared by US politician Cynthia McKinney on Twitter on 29 March

Although the story was retracted and the narrative quickly investigated by News24, this case is a prime example of how disinformation can jump from social media platforms to traditional media, with potentially dangerous side-effects. 

A closer look at what happened during the two weeks of Gates-related confusion showed that there were no bots influencing the conversation.

Rather, anonymous influencers with hundreds of thousands of followers spread reactionary disinformation that was picked up by local media on the lookout for trends – and trending hashtags.

But hashtags can be artificially manipulated. Two of the most widespread ways of doing this are early morning hashtagging and click-to-tweet campaigns.

Early morning hashtagging

An easy way to manipulate Twitter’s algorithms is to start a hashtag campaign early in the morning. Twitter will list a hashtag as trending if it is tweeted multiple times, particularly if it is tweeted by a significant number of accounts, and if the hashtag makes up a significant proportion or percentage of recent tweets.

Because there is less activity when a country’s users are asleep, a hashtag needs a lot less engagement to reach the threshold required for the algorithm to mark it as trending. There are also fewer accounts available to counter the hashtag and the narrative being spread.

Malicious actors can purchase accounts to amplify a hashtag in the early morning. This makes the hashtag’s narrative appear to have grassroots support, when in reality bad actors have worked to game Twitter’s algorithm to push a specific agenda.

One of the primary issues with early morning hashtagging is the impact it can have on local media. Journalists may check their phones in the morning and notice that an inflammatory political hashtag is trending at 6am, even though it was only tweeted a few hundred times by a small number of accounts.

As a result, journalists might think the hashtag has more legitimate support than it actually does, and a narrative with very little actual support can subsequently find its way into the news cycle. After being amplified by local media it is likely to be picked up organically by legitimate social media accounts and maintain its trending status.

Click-to-tweet campaigns

Websites with pre-written tweets that allow social media users to share content with one click are referred to as click- to-tweet campaigns. Pre-written tweets are often utilised by activists and social justice groups to build support around pre-determined hashtags and connect with high-profile social media accounts.

Recently, members of the Ethiopian diaspora have utilised click-to-tweet campaigns to raise awareness about the conflict in Tigray.

Multiple websites, both supporting the government and opposing the conflict, were created within weeks after fighting broke out last year.

These websites organise dedicated click-to-tweet campaigns with the goal of getting hashtags to trend.

Trending topics in Ethiopia on 24 February 2021, were all from different click-to-tweet campaigns.

The websites have tutorials in Amharic and Tigrinya explaining how to set up Twitter accounts specifically to share the pre-written hashtags.

As a result, a significant number of accounts promoting the hashtags were created after conflict broke out in November 2020.

From the outside, may of the accounts appeared fake – they had little to no personal information, were following no other accounts and had no followers, and only posted pre-written hashtags

From the outside many of the accounts appeared fake – they had little to no personal information, were following no other accounts and had no followers, and only posted pre-written hashtags.

Accounts created to amplify pre-written hashtags from click-to-tweet campaigns also appeared to be fake.

Despite this, the campaigns were often successful – the hashtags frequently trended in Ethiopia and even other parts of the world, and by tagging journalists and news organisations the campaigns were able to reach a wider audience.

Although Twitter is not as popular as Facebook on the continent, these examples show how easily it can be used to manipulate local media. 

This article first appeared on The Continent, the pan-African weekly newspaper designed to be read and shared on WhatsApp. Download your free copy here.