Project summary

Digital government initiatives in countries such as Australia and the UK increasingly depend on AI-enabled platforms to streamline public service delivery. However, while these systems have enhanced efficiency and access for many, their benefits are not equitably distributed. Marginalised populations—including elderly individuals, First Nations peoples, low-income households, and individuals with limited digital literacy—continue to face significant barriers in accessing and trusting digital services.

These challenges stem from multiple structural and technological factors. Limited access to digital devices, unstable internet connectivity in regional and remote areas, and low levels of digital confidence disproportionately affect these communities. Language barriers, physical or cognitive disabilities, and cultural diversity further complicate interactions with digital systems—particularly those relying on biometric authentication.

Face Anti-Spoofing (FAS) technologies, which are increasingly used to prevent fraud in identity verification, present a critical area of concern. These systems are often trained on non-representative datasets, resulting in significant performance disparities. Individuals with darker skin tones, older adults, and users wearing cultural attire or using outdated devices may experience high false positive rates—leading to unjust denial of access. Such experiences not only undermine service equity but can also erode trust in digital government systems and amplify feelings of exclusion and disempowerment.

This collaborative research project seeks to critically assess the accessibility and trustworthiness of AI-based FAS technologies in the context of marginalised communities in Australia and the UK. Leveraging an existing FAS prototype that integrates eye movement features to establish “liveness” in identity verification, we aim to evaluate the inclusivity and real-world usability of such systems.

We will address the following research questions:

• RQ1: What are the root causes of accessibility and usability challenges experienced by marginalised groups when interacting with FAS technologies?

• RQ2: How do socio-technical factors—such as digital literacy, cultural norms, and infrastructure quality—shape these challenges?

• RQ3: Does the integration of eye movement characteristics in FAS systems enhance both spoof resistance and user inclusivity, or does it introduce new usability or fairness issues?

To address the research questions, we propose a multi-method approach combining natural language processing, survey research, and simulation-based analysis, underpinned by AI-Bayesian inference techniques.