Digital Watchdogs: Developing early warning detection systems to counter hate speech and toxic online

WanaData
Hacks/Hackers Africa
7 min readSep 22, 2023

By Community Coordinators

The Nairobi HackHackers virtual session in August focused on “Digital Watchdogs: An overview of early warning mechanisms in the detection of hate speech and toxic content.” This enlightening session featured John Ndung’u, an Investigative Data Analyst within the iLAB team at CfA (Code for Africa). John’s work primarily involves tracking hate speech, incitement, mapping online disinformation, coordinated inauthentic behaviour, and transnational organised crime. The session shed light on the complexities of information disorder and the crucial role diverse toolsets play in combating it. The HacksHackers East Africa session happened on the 24th of August 2023 and there were 18 participants in total.

John Ndung’u introduced the session by highlighting the pervasive issue of information disorder. He emphasised that “fake news” is not solely the creation of malicious actors deliberately spreading false information. It encompasses a wide range of problems, from innocent mistakes to toxic hate speech aimed at inciting violence. To effectively combat this multifaceted problem, diverse toolsets and strategies are necessary.

During the session, John delved into the concept of hate speech and its various forms. In common language, “hate speech” refers to offensive discourse targeting a group or an individual based on inherent characteristics, such as race, religion, or gender, that may threaten social peace.

Recognising the gravity of the issue, Code for Africa (CfA) has developed a comprehensive 4-pronged approach to combating information manipulation:

  1. Debunk Claims: CfA works to verify or debunk claims circulating online, ensuring that misinformation is corrected promptly.
  2. Expose Networks: Identifying and exposing the networks behind disinformation campaigns is crucial to understanding their reach and impact.
  3. Follow the Money: Tracking the financial aspects of information manipulation helps uncover hidden motives and funding sources.
  4. Monitor Narratives: CfA continuously monitors and analyses the narratives surrounding information manipulation, shedding light on evolving strategies.

The speaker noted that effective early detection of online hate speech requires a methodological approach.

Forensic teams at Code for Africa identify potential “bad actors” or systemic weaknesses by conducting tests on both media data and social media data. These tests provide valuable insights into the information environment and reveal the presence of hate speech or toxic content.

One of the key techniques discussed was Trend Analysis. Identifying spikes in certain types of posts is a vital aspect of this approach. A spike in similar posts often indicates the onset of a coordinated hate speech campaign or the spread of toxic content. Analyzing these spikes closely helps pinpoint the key drivers behind sudden increases in such content.

Network Analysis also plays a pivotal role in identifying coordinated harmful behaviour online. CfA employs this technique to analyse the relationships between different users on social media platforms. It helps identify coordinated harmful behaviour, such as online harassment campaigns or the spread of disinformation.

The session also touched upon other critical aspects of hate speech detection, such as the development of lexicons, sentiment analysis, actor watchlists, puppets, and networks, queries creation, and mapping networks.

In conclusion, the Nairobi HackHackers August virtual session provided valuable insights into the world of digital watchdogs. John Ndung’u’s presentation highlighted the multifaceted nature of information disorder and the importance of diverse toolsets in combating it. Attendees learned about the methodological approach, trend analysis, trigger language, network analysis, and other techniques crucial for early detection of hate speech and toxic content online.

The session’s participants came from diverse backgrounds, including journalists, data analysts, and technology enthusiasts, who shared their experiences and perspectives on the evolving challenges of

information manipulation in the digital age. This diversity of voices added depth to the discussion and emphasised the need for collaborative efforts in addressing this critical issue.

The upcoming session promises to delve even deeper into the topic, covering tools for identifying online hate speech, sample cases in Kenya, techniques used to spread hate speech and incitement during elections, counter-messaging efforts, and solutions implemented by CfA partners to tackle disinformation and toxic networks. As digital watchdogs, it is our collective responsibility to stay vigilant and work together to safeguard the integrity of online information.

Intro to Relational Database Management Systems

In his opening talk, Joseph Dokhare described relational database management systems as an organised collection of logically related data. Varying in size, complexity, and use. Dokhare spoke at the virtual Hacks/Hackers monthly meetup on 25 August 2023, surrounded by 18 participants.

A Data Analyst at Code for Africa, Dokhare took the audience through the five major components of a database. These components included Hardware, Software, Data, Procedure, and Database Access language.

Displaying impressive knowledge of the subject, he took on the task of explaining the difference in databases and spreadsheets. Although both are convenient ways of storing data, deciding which one to use depends on how much data can be stored and who has access to the data. He further provided a detailed breakdown of the history of databases, and types of databases.

During his presentation Dokhare shared 8 types of databases.

Preparing to focus on Relational Database Management Systems (RDBMS), Dokhare opened the floor to the community to discuss real-life application of databases. The audience provided examples of Wikipedia, survey to go, Crawtangle, and Jumia.

The speaker used this as a stepping stone to discussing the relational model introduced by E.F Codd, called entity relation model (ER model).

To understand RDBMS you need to understand the ER model. The ER model defines the data elements and relationship for a specified system. It further assists in describing entities, attributes, relationships between data, and building databases quickly.

Dokhare then explained entities as objects, classes, people, or places. Entities are presented through rectangles. Within the entity is an attribute, represented by a circle. A rhombus or diamond describes the relationship between the two.

He then took the audience through how information is displayed, on a RDBMS spreadsheet.

A rhombus or diamond describes the relationship between the two.
Dokhare shared a diagram of how information is displayed on an RDBMS spreadsheet

The talk progressed to understanding primary keys and foreign keys. A primary key provides a unique identification to a record. Meanwhile, a foreign key is a primary key in another table. On this note Dokhare had to pause the talk, as a cloud of confusion filled the room.

He moved the audience to Airtable in an attempt to explain this concept. Accessible language and visuals came in handy. He focused on the example of a salary. He pointed out that a salary is an entity. A staff ID is a primary key. The staff ID connects other parts that link to the salary. The foreign key using the salary example, is anything that is not directly related to information about the salary.

It was touch and go as the audience struggled to process the technical information. As a way of properly unpacking this topic, the community requested two additional meetups.

This means there will be a second and third instalment of the talk around RDBMS. The intention of the follow-up talks is to allow ample time to engage with RDBMS, LinkTree, and Airtable. Ensuring that the community is following and participating in the talk.

In this first session, Dokhare shared an introduction to RDBMS. The community is now familiar with this concept and how it relates to databases. Hacks/Hackers West Africa looks forward to hosting Dokhare again, to receive practical guidance on applying RDBMS in the media field.

The worlds of hackers and journalists are coming together, as reporting goes digital and Internet companies become media empires.Journalists call themselves “hacks,” someone who can churn out words in any situation. Hackers use the digital equivalent of duct tape to whip out code.

Hacker-journalists try to bridge the two worlds. Hacks/Hackers Africa aims to bring all these people together — those who are working to help people make sense of our world. It’s for hackers exploring technologies to filter and visualise information, and for journalists who use technology to find and tell stories. In the age of information overload and collapse of traditional business models for legacy media, their work has become even more crucial.

Code for Africa, the continent’s largest #OpenData and civic technology initiative, recognises this and is spearheading the establishment of a network of HacksHackers chapters across Africa to help bring together pioneers for collaborative projects and new ventures.

Follow Hacks/Hackers Africa on Twitter and Facebook and join the Hacks/Hackers community group today.

--

--