Developing ethical ML pipelines for Countering Child Exploitation
Ai Lin Soo1, Campbell Wilson1, Nina Lewis1, Greg Rolan1 1AiLECS Lab, Monash University
Abstract
AI tools are slowly becoming the norm but with their rise there is also the heightened awareness of their use and development within ethical, transparent and socially acceptable boundaries. It’s not just in our day to day lives where we see the integration of AI, machine learning (ML) is also being explored by law enforcement (LE) agencies to improve community safety. One such application being explored is in the classification of child sexual abuse material (CSAM) to support LE officers triage the rapidly increasing amount of CSAM online. While there is benefit to having such a tool in the arsenal to counter child exploitation, the same questions of ethical AI remain.
To address the increasing ethical and transparency concerns of AI particularly in law enforcement, the Ai for Law Enforcement and Community Safety (AiLECS) Lab has launched data management infrastructure that demonstrates its VALID (Veracity, Agency, Longevity and Integrity in Datasets) principles, together with the My Pictures Matter (MPM) crowd sourcing initiative. MPM aims to curate an ethical dataset of benign child images without compromise on the dignity of participants or infringing on social expectations. VALID/MPM ensures consentful use of data is embedded at all points of the AI pipeline and is one tool that supports the growing need for community education on ML given its increasing prevalence in communities.
AiLECS Lab is a partnership between Monash University and the Australian Federal Police.
Biography
Ai-Lin Soo is the Operations Manager for the Ai for Law Enforcement and Community Safety (AiLECS) Lab at Monash University and Project Officer at UNSW. She applies her project management skills to support a diverse range of activities in areas such as research data management and operational governance. Ai-Lin is also heavily involved in the Research Data Culture Conversation (www.researchdataculture.org) as the RDCC coordinator and product manager of the Macro View.