The Personal Writes the Political: Securely Rendering Black Lives Legible Through the Application of Advanced Machine Learning (ML) to Anti-Apartheid Solidarity Letters

The Personal Writes the Political (PWP) is a digital humanities project that applies advanced machine learning (ML) models to anti-apartheid solidarity letters predominantly authored by Black South African women. We began work on the project in July 2022 when we were awarded an ACLS Digital Justice Seed Grant. Our initial objective was to develop software that automates the transcription of the Programme II Collection (P2C), a unique collection of letters that emerged from an elaborate welfare program that secretly provided financial support to South African families of political detainees during the last three decades of the apartheid era. Most of these letters were authored by Black women heading households while their loved ones were caught in apartheid’s labyrinthine carceral state. By our recent estimates the P2C contains nearly 233,000 letters written by the members of 4672 families who received Programme II support. Their stories are largely unknown in the historical record. We spent the prior 18 months cataloging the entire collection, prototyping Careful Recall (Caracal) a software workflow that automates the responsible extraction of data from these handwritten materials and thinking through data ethics and security challenges with local stakeholders. During the forthcoming development phase, we will finalize and install Caracal on a secure server located at the Mayibuye Centre Archives, University of the Western Cape (Cape Town, South Africa). We will also disseminate our work on both sides of the Atlantic in a series of research presentations and project meetings. Concurrently, we will identify ‘focal clusters’ of letters for screening, and once approved, apply advanced handwritten text recognition and named entity recognition models to automate the transcription, redaction, and output of structured data from their contents. We believe that with additional software development and model training, Caracal will securely render the experiences of letter-writers legible in the form of structured data, for the first time.