Safeguarding citizens at the intersection of Generative AI and the Metaverse

There is limited research on how AI-generated content impacts people’s perception with regards to authenticity, trust, identity, privacy, and security. This is particularly true in the context of Metaversal experiences – experiences that use immersive displays, such as augmented reality (AR) and virtual reality (VR) devices. 

Given the plethora of data that AR/VR devices typically collect – about the users, their environments, and their interactions – AI models will soon be able to generate content that is hyper personalised. This means that the system will know exactly how to deliver a message to a specific person to maximise impact. In the wrong hands, this could have significant negative consequences, ultimately leading to an erosion of trust in public institutions.

For instance, imagine that “your most trusted friend” tells you that you need to do a specific task in a specific way, otherwise they (or someone you love, including yourself) would suffer negative consequences. Now imagine that in fact what you perceived to be “your most trusted friend” was in fact a realistic avatar generated to trick you. 

This project will investigate these types of scenarios to provide insights into how things could go wrong and how this should be prevented and mitigated.

Methodology

The methods used will be rooted in and further extend current HCI practices. Depending on the context, we will use co-creation and co-design workshops, interviews and surveys, prototyping and user studies, and scenario writing. Additionally, we may use theatre and performing arts as a way of provoking thought and evaluation of ideas.

Collaborators

There will be opportunities to collaborate with a wider community of HCI and cybersecurity researchers, including the SPRITE+ and REPHRAIN communities and CWI Amsterdam. Additionally, there will be input from a senior adviser from the BBC R&D.