Agency in augmented reality: exploring the ethics of Facebook's AI-powered predictive recommendation system

The development of predictive algorithms for personalized recommendations that prioritize ads, filter content, and tailor our decision-making processes will increasingly impact our society in the upcoming years. One example of what this future might hold was recently presented by Facebook Reality Labs (FRL) who work on augmented reality (AR) glasses powered by contextually aware AI that allows the user to "communicate, navigate, learn, share, and take action in the world" (Facebook Reality Labs 2021). A major feature of those glasses is "the intelligent click" that presents action prompts to the user based on their personal history and previous choices. The user can accept or decline those suggested action prompts depending on individual preferences. Facebook/Meta presents this technology as a gateway to "increased agency". However, Facebook's claim presumes a simplistic view of agency according to which our agentive capacities increase parallel to the ease in which our actions are carried out. Technologies that structure people's lives need to be based on a deeper understanding of agency that serves as the conceptual basis in which predictive algorithms are developed. With the goal of mapping this emerging terrain, the aim of this paper is to offer a thorough analysis of the agency-limiting risks and the agency-enhancing potentials of Facebook's "intelligent click" feature. Based on a concept of agency by Dignum (Responsible Artificial Intelligence: How to Develop and Use AI in a Responsible Way. Springer International Publishing, Cham, 2019), the three agential dimensions of autonomy (acting independently), adaptability (reacting to changes in the environment), and interactivity (interacting with other agents) are analyzed towards our ability to make self-determining choices.