UMATI- Monitoring Dangerous Speech Online

Juliana Rotich
Dec 19, 2012

This is a blogpost by Angela Crandall, one of the Research Leads at iHub Research, Nairobi.

While most projects related to hate speech have been looking at mainstream media, we are aware of the influence—positive and negative—that New Media such as the blogosphere and online forums had on the 2007 Post Election Violence in Kenya. Therefore, our flagship Umati project seeks to monitor and report, for the first time, the role of new media on an election. Our Kenya-based project has citizens at its core and uses relevant technologies to collect, organize, analyze, and disseminate the information collected.

For this project, we are looking specifically at “dangerous speech”, or hate speech with a potential to cause violence. We have adopted the framework created by Professor Susan Benesch of American University (Washington, DC, USA), which defines dangerous speech as a subset of hate speech, and comprising of five criteria:

- A powerful speaker with influence over an audience;

- An audience vulnerable to incitement;

- Content of the speech that may be taken as inflammatory or inciteful;

- A conducive social and historical context of the speech; and

- An influential means of disseminating the speech.

The goals of the Umati project are three-fold:

To cite incidences of dangerous speech from the Kenyan online ecosystem and thus galvanize a definition that the relevant parties can act upon;

To forward incidences of dangerous speech to the Uchaguzi platform (www.uchaguzi.co.ke) as soon as they are identified, so that further harm may be circumvented, limited or thwarted;

To develop and test a new methodology/process of monitoring inflammatory speech online that could be used in other applicable instances.

Umati, Uchaguzi, Ushahidi

Monitoring Process

The online monitoring process is being carried out by five monitors, representing the five largest ethnic groups in Kenya to enable the translation of cited incidences from vernacular to the country’s official language, English. The Umati project provides a perfect opportunity to test new Ushahidi software, SwiftRiver. The ideal application of SwiftRiver for this project is to collect data streams from various identified sources (such as blog posts, comments on online newspapers, Twitter, and Facebook) to aggregate the data onto one platform. Once the data is aggregated, a categorisation process that facilitates the categorisation of the cited speech into respective buckets is necessary before bucketing the incidences or “drops” of dangerous speech into three categories:

Bucket One: Offensive speech (lowest level);

Bucket Two: Moderately Dangerous speech;

Bucket Three: Extremely Dangerous speech (highest level).

Despite the appropriateness of the SwiftRiver technology, unfortunately, there have been some challenges in the practical implementation. For example, we have found it difficult to monitor blogs that don’t have RSS feeds (especially the comments section). Blogs without RSS feeds are quite common in this project, especially with vernacular language sites. Additionally, we have encountered challenges in the Facebook integration between SwiftRiver and Facebook. This has forced us to use other applications in order to monitor Facebook’s group comments and page comments. Finally, a big issue has been the reliability of Twitter feeds. This is especially true when large amounts of data are aggregated on SwiftRiver.

In addition to the challenges faced with the integration across various data sources, we have also identified certain aspects of SwiftRiver that we need to be customized for online media monitoring on Swift. These include a form pop-up prior to categorizing drops into buckets, color coding of drops, and easy integration into the Uchaguzi platform.

We are currently using work-around methods while we collaborating with our technology partner, Ushahidi, to address these challenges.

Outputs

We will be producing and releasing monthly reports that consolidate incidences of dangerous speech that have been identified in Kenya’s webosphere from three main sources: social media (Facebook and Twitter), online blogs and comments sections of online newspaper. The next report  has been published and is available on this link[pdf].

We hope that the work of this project will lead to the inclusion of a more elaborate definition of illegal speech in the current constitution of Kenya, and that findings will be used to educate the Kenyan public on what type of speech has the potential to disrupt peace and security in the country. Through this project, we aim to create a process and accompanying technology (A variation of SwiftRiver which complements the Ushahidi platform), and recommend a methodology that can be replicated in other countries to monitor dangerous speech leading up to pivotal national events, such as elections and referendums. We look forward to continuing to share our experience, challenges, and insights through this blog and welcome your comments and feedback. *Generous support for Umati work comes from several partners, including those that support SwiftRiver app development.