Can Technology Help Ease Ethnic Tension in Ethiopia?

Working with university students to flag hate speech on Facebook, Twitter, and other platforms.

With violence spurred by hate speech on the rise globally, many countries struggle to diffuse local tensions while still protecting freedom of expression. In Ethiopia, recent reforms have resulted in constructive dialogue. Yet ethnic- and religious-based dangerous speech has also increased, as have spikes in ethnic violence, displacing millions. And the country’s passage of a controversial law addressing hate speech and disinformation, intended to deescalate friction, has led to worries about restrictions on personal freedoms.

Dexis’ team is part of this effort to deescalate friction by implementing Promoting Religious and Ethnic Tolerance (PRETE), and activity in Ethiopia that determines how local actors can help address hate speech online and then designing approaches to counter hate speech. By working with Addis Ababa University’s Center for Human Rights, the Ethiopian Human Rights Commission, the Attorney General’s Office, and two regional universities, PRETE is helping local stakeholders to identify what constitutes hate speech, understand relevant legal definitions, support data collection of hate speech online through machine learning, develop and disseminate counter-messaging, and reduce inflammatory, dangerous speech. In partnership with a media intelligence firm, PRETE is developing an online dashboard utilizing machine learning technology to ingest data across social media platforms (i.e., Facebook, Twitter, YouTube) to map the hate speech landscape, monitor trends and patterns, identify key influencers, and other relevant analytics.

The link between online hate speech and violence is a vast and complex issue. PRETE is in its early stages and is tackling just one small aspect of it. What the project is rapidly learning has implications for multiple settings, such as:

• Technology is not everything—this work ultimately includes lots of interactions, engagement, and buy-in with communities to understand what resonates with audiences and tailor effective activities to counter hate speech. Navigating both bureaucratic institutions and mobilizing stakeholders requires dedicated time, face-to-face relationship building, and constant socialization with key interlocutors.
• How building sustainability into “high-tech” approaches is key to supporting long-term local solutions. Implementers must co-create solutions with stakeholders to ensure that technology platforms and approaches can and are funded and sustained locally. These solutions must be grounded in local equities and realities with practical ideas for how to expand, maintain, and evolve these tech approaches.
• Ways to adjust the model to low bandwidth settings. Programs must account for environments with weak ICT infrastructure, low internet penetration rates, and overall lack of access to resources like mobile smart phones and laptops. Training materials, adapting online tools to low bandwidth environments, and opening up IT access at the Universities are a few key areas PRETE has focused on to address these challenges.

PRETE works with university officials to identify a diverse team of select Jimma and Jigjiga University students to train on ways to identify hate speech online and develop effective counter-messaging. As Ethiopian universities are “de-ethnicized” through a lottery process for placement, universities have experienced tensions and even violence but present an opportunity for cross-ethnic dialogue and cooperation. The PRETE team trained over a dozen university students on polarizing issues in Ethiopia, hate speech, fake news, freedom of expression, and how and when to flag hate speech on Facebook, Twitter, and other platforms. The students were trained on using an online tool to flag hate speech content to support local data collection. This core group can also recruit other students and colleagues, training them in ways to get involved and broadening the overall knowledge base about hate speech.

The resulting data and analysis provide critical insights toward developing and delivering counter-messages and understanding who and what communities to engage with. This data is also intended to inform local and religious community leaders as they continue their outreach and messaging to stakeholders in local communities.

Due to COVID-19, discovering how to accommodate this work remotely when ICT infrastructure is underdeveloped in Ethiopia is now an even higher priority. Upcoming in-person workshops are on hold. The PRETE team is maintaining contact with university representatives and communicating directly with student leads of the working group via the Telegram messaging app. To ensure planning can continue, the PRETE team will work with students to maintain their flagging work through a mobile version of the tool and will keep open channels of communication with key stakeholders at Jimma and Jigjiga.

PRETE’s approach provides a concrete way for local students, activists, and leaders to contribute to countering hate speech and pragmatically support community stabilization. More broadly, it builds local capacity around digital literacy, civic education, machine learning, and analytics. Exploring how this model works in other contexts will be valuable to continue refining practical, locally-sustainable ways to invest in and sustain technological tools and platforms that address dangerous speech.

Sarah Kim is Senior Technical Advisor in the Monitoring, Evaluation, and Learning Division at Dexis where she provides business development and technical support for State and USAID programs focused on security assistance, governance, political transition, conflict/post-conflict, and stabilization sectors.

Photo by Eric Lafforgue / Hans Lucas / Hans Lucas via AFP