Joymanyasi/research-notes

A living archive of research memos, policy reflections, and exploratory notes on African genomics, AI governance, and scientific sovereignty." # Research Notes: Algorithmic Inequity and the Invisible Patient **Researcher:** Joy Manyasi Kabaka **Affiliation:** Sequence Africa **Location:** East Africa --- ## What This Is This repository grew out of a question that kept surfacing in my reading: > *What happens when artificial intelligence systems trained on one population are applied to another with fundamentally different biology, healthcare contexts, and data landscapes?* This is **not** a finished project. It's an open notebook where I: - Develop ideas and test arguments - Track what I'm learning - Connect threads across **AI, genomics, and health equity** in African contexts The focus is exploratory and reflective rather than polished or definitive. --- ## The Core Question AI systems in biomedical research are often treated as neutral technical instruments. In practice, they are built on data that reflect specific social, historical, and institutional conditions. When those conditions systematically exclude entire populations, two questions arise: 1. **What gets encoded into the models?** 2. **What and who gets left out?** I am especially interested in how this plays out for **triple‑negative breast cancer (TNBC) in East African women**, where: - TNBC has a **higher prevalence** in East Africa than in many European populations - The **genomic and biological landscape** differs in clinically significant ways - Most AI training data for oncology still comes from **European and North American cohorts** The concern is that AI systems built on these data may be: - Less accurate - Less generalisable - Potentially harmful when deployed in African clinical settings These notes treat that not simply as a performance issue but as a case of **algorithmic inequity** and a form of **invisibility**: East African women with one of the most aggressive breast cancer subtypes may barely register inside the systems that claim to make cancer care more precise. --- ## What's Here (Right Now) | Document …