The Neighborhood: An Interactive Social Media Experience for Increasing Awareness of Automated User Profiling

Abstract

Internet users leave a range of digital traces that allow for automated profiling, often without their knowledge. Advancements in artificial intelligence (AI) make it easier than ever to automate the inference of user attributes that are not even directly shared, exposing users to privacy risks. One means of protection against this is the widespread promotion of AI literacy to social media users. As an example of how this might be achieved, we present an interactive online simulation with integrated AI tools, designed to raise awareness of AI capabilities for user profiling. Within this experience, users can take actions similar to those that they would everyday: make a post, update their profile, and like existing posts. After the users finish their experience, they may choose to receive a report about an AI model's inferences about their personal attributes along with information about how similar inferences are made about them when using major social media platforms. We find that the level of interaction within the simulation correlated with large language model (LLM) accuracy in profiling. We also observe high accuracy in profiling despite many participants self-identifying as protective of their data. Finally, a majority of participants reported that as a result of taking part in our study, they expect to think more carefully about their online interactions and what they might reveal about themselves in the future. This aligns with our aim to increase AI literacy and empower people to make more informed decisions on what they disclose on the internet.