White Paper: Human-AI Collaboration in Conflict Analysis: Text Classifier Development with Peacebuilders

This paper documents a collaborative research process involving peacebuilders and data scientists in Kenya and Sudan to develop AI-based text classifiers for monitoring online polarization and hatespeech. The method describes a participatory annotation process in which practitioners and domain experts contributed to problem definition, annotation design, iterative validation, and model evaluation. Fine-tuned BERT-based classifiers were … Read more

Case Study on Social Media Analysis of Elections in Kenya (2026)

This paper was written by Caleb Gichuhi in the book Disinformation: A Multidisciplinary Analysis. This case study analyzes how disinformation was used on social media to influence perceptions during Kenya’s 2022 general election, specifically examining disinformation tactics across three electoral phases; pre-election, during-election, and post-election. Data gathered from Facebook, TikTok, and Twitter reveal a persistent … Read more

The Polarisation Footprint – Kenya

The polarization footprint is to the attention economy what the carbon footprint is to the industrial economy. Join us for a webinar to explore this idea, including results of the Kenya polarization footprint. Register here. 🤓 If you want to read ahead, we’ve just published our work to measure polarization footprints, both a replicable, public … Read more

The Kenya Polarization Footprint Results Report (2025)

The Kenya Polarization Footprint measures how social media is dividing Kenyans. Specifically, we measure “affective polarization”, where people dislike and distance themselves from others based on their identity. This is not the same as a difference of opinion: differences of opinion are good for democratic debate, and for our diverse society. Identity-based polarization is not: … Read more

Smart Social Media Analysis (Part 2)

Webinar on annotation Strategies for Text Classification. Join is for this webinar to explore practical strategies for creating high-quality labeled datasets that reflect the nuanced realities of conflict-sensitive topics, like developing annotation guidelines, managing inter-annotator agreement, and how to handle edge cases. Whether you’re preparing to fine-tune models for your own organization’s needs, are curious … Read more

Smart Social Media Analysis (Part 1)

15 October 2025 Join us for an interactive session exploring Phoenix’s newest feature: open-source language model integration for text classification. Learn how to move beyond simple keyword searches to understand meaning, context, and nuanced patterns in social media data. We’ll cover the fundamentals of text classifiers, demonstrate how to apply pre-trained models from our curated … Read more

Exploring Online Polarizing Content: LGBTQ+ Rights in Kenya

This report is part of a broader initiative to analyze online discourse and specifically monitor harmful content e.g., hate speech in Kenya on various topics in order to help mitigate the threat of polarizing content in fueling offline and online violence. Insights gathered from the monitoring efforts are shared with civil society actors to inform … Read more

Mapping of Legal Framework and Responses by Actors to Address Harmful Content Online in Kenya

Over the last 15 years, the Kenyan government has responded to hate speech and misinformation on social media through new legislation and direct work by governmental institutions established through that legislation. Most legislation, especially legislation pre-dating 2013, does not explicitly address or name social media; that said, any legislation that addresses harmful speech, whether hate … Read more