Ethiopia polls: Facebook tames social media political hate speech

​Social media giant Facebook has moved with speed to tame political hate speech and misinformation in Ethiopia ahead of the upcoming General Elections on June 21.

The company said the efforts will be in force before, during and after the vote, and will target to control the spread of content that incites voters to violence on Facebook and Instagram.Mr Nathaniel Gleicher, Head of Security Policy at Facebook told Nation.Africa that the company has removed a network of accounts, pages and groups in the Horn of Africa nation for coordinated inauthentic behaviour (CIB) that targeted domestic audiences.

 “We have removed 65 Facebook accounts, 52 pages, 27 groups, and 32 accounts on Instagram for violating our policy against CIB. The operation used duplicate and fake accounts — some of which were already detected and disabled by our automated systems,” he said.

In a targeted political advertising spending budget of Sh668,000, about 1.1 million accounts followed one or more of the pages, while 766,000 accounts joined one or more of the groups.  1,700 people followed one or more of the Instagram accounts.

Accounts pulled down

Accounts that were pulled down had gone through significant name changes over time, using spam-like inauthentic distribution tactics to post the same content across multiple Pages and Groups simultaneously.

The misinformation network posted primarily in Amharic about news and current events in Ethiopia, including the Prosperity party, Prime Minister Abiy Ahmed, and criticism of Egypt and Sudan related to Ethiopia’s mega dam project. 

“They also posted critical commentary about various opposition politicians and groups in Ethiopia, including Oromo Liberation Front, Ethiopian Democratic Party, and the Tigray People’s Liberation Front among others,” Facebook said in a statement.

Although the people behind it attempted to conceal their identities and coordination, Mr Gleicher said Facebook’s investigation found links to individuals associated with the Information Network Security Agency (INSA), in Ethiopia.

Ms Mercy Ndegwa, Facebook’s Head of Public Policy for East and Horn of Africa said the company would remain committed to election integrity.

 “These efforts are being implemented by a team purpose-built to focus on the Ethiopian election. Local understanding is critical to doing this work effectively, therefore the team includes a number of people in and from Ethiopia including experts in topics like misinformation, hate speech, elections and disinformation,” she told Nation.Africa.

Political debates

Facebook said it believes political discussions and debates should be transparent to every voter, having introduced a number of tools that provide more information about political ads on Facebook and Instagram.

 “In March this year, Facebook made these political ads transparency tools mandatory in Ethiopia,” the statement reads.As a result, any politician who wants to run political ads in Ethiopia must now go through a verification process to prove who they are and that they live in Ethiopia.

Facebook runs additional checks to ensure compliance with its policies, with political ads in Ethiopia will be labelled with a “paid by” disclaimer, users can see who paid for them.The social network also revealed that it puts political ads that run in Ethiopia in its ads library so that everyone can see what ads are running, what types of people saw them and how much was spent.

“This fully searchable archive will store these ads for seven years. These changes mean that political advertising on Facebook and Instagram is now more transparent than other forms of election campaigning, whether that’s billboards, newspaper ads, direct mail, leaflets or targeted emails.

”Facebook’s Community Standards - which set out what is and isn’t allowed on Facebook - cover a number of areas relevant to elections, including policies against harassment and incitement to violence, as well as detailed hate speech policies that ban attacks on people based on characteristics like ethnicity or religion.

Facebook removes any content that it's made aware of that violates these rules.

Policies in Ethiopia

To further broaden awareness of its policies in Ethiopia, Facebook has run online ad and radio campaigns in Ethiopia and held training sessions with activists, civil society organizations, small and medium sized businesses owners, government agencies and members of the local media. 

The company has also established dedicated reporting channels for specialized international and local human rights and civil society organizations to make sure they can quickly review problematic content they identify for possible violations, and continue to work with local partners who provide Facebook with feedback that the company incorporates into its policies and programs.

Facebook also invested in Natural Language Processing, a sub-set of Artificial Intelligence technology to proactively identify hate speech in Amharic and Oromo languages in Ethiopia, alongside over 40 other languages globally.

Over the last few years, Facebook said it had tripled the size of the global team working on safety and security to over 35,000 and hired more content reviewers who are native speakers of Amharic, Oromo and Somali, while having the capacity to review content in Tigrinya.

Fake news

Through such efforts, Facebook removed 87,000 pieces of hate speech in Ethiopia, about 89 per cent of which were detected proactively, between March 2020 and March 2021.For content which doesn't violate these particular rules, Facebook has partnered with independent third-party fact-checking partners in Ethiopia - Pesa Check and AFP -- to ascertain whether articles propagate misinformation or false news. 

When they review and rate a piece of content as false, Facebook uses AI to reduce its distribution so fewer people see it and add a warning label with more information for anyone who does see it.

“In general, when a warning screen is placed on a post, 95 per cent of the time people don’t see it,” the company said.

While only 5.8 per cent of Ethiopia’s 112 million population use social media, according to the Digital 2021 Ethiopia report from DataReportal, the country has been criticised for shutting down the internet for the 5.5 million citizens who need access to information. 

According to British research firm Comparitech, a combined total of 3,657 hours were lost in early 2020 during an internet shutdown, with the first one happening from January to the beginning of April in Western Oromia.

The government cited security reasons for that. But in June 2020, the entire country went into a 23-day internet outage following the shooting of Haacaaluu Hundeessaa, a prominent singer. Another one followed in November in the region of Tigray after war broke out and lasted till December 15 when some services were restored.

[email protected]