Typology of transparency as best practice: evidence from facial recognition technologies in Australia

Abstract

Facial recognition technologies (FRTs) are commonly used in Australia in an under-regulated environment. Only two regulatory mechanisms currently direct FRT use–industry self-regulation and Australia’s privacy laws. Currently, best practices for FRT development and adoption are a key governance mechanism, including the recent guide to FRT privacy risks evaluation published by the Office of the Australian Information Commissioner in 2024. Transparency is lauded as one of the key principles for implementation, yet it can be an elusive best practice to operationalize. Transparency spans everything from technical design decisions and data governance to real-world applications, user communication, and regulatory compliance. Thus, our research critically considers the concept of transparency in FRT governance by asking: What does transparency as best practice mean in the development and implementation of facial recognition systems? This study sought to map multi-stakeholder viewpoints through problem-centered interviews (n = 30) with expert technical, policy, and academic stakeholders making and informing FRT policy in Australia. Based on their responses, we develop a typology of what transparency means in the context of operationalizing algorithmic surveillance technology governance, specifically FRTs. Our findings adopt a socio-technical perspective and identify five distinct modes of transparency in FRTs, including technical, functional, procedural, public and contextual transparency.