A study by Northeastern University in 2018 analyzed 17,260 popular Android apps, which revealed that while no secret audio recording was found, approximately 9,000 apps had the ability to take screenshots and send them to third parties. In one instance, the food delivery service GoPuff sent screen activity recordings to the analytics company Appsee, potentially exposing search queries, cart data, and payment information.
The Federal Trade Commission (FTC) in August 2026 concluded a case against Cox Media Group, MindSift LLC, and 1010 Digital Works LLC. These companies sold a product called 'Active Listening' which claimed to analyze conversations near smart devices for targeted advertising. However, the same investigation revealed the same product actually used broker data, such as email lists, and did not use voice data at all. The companies paid a total of $930,000 in fines, with Cox Media Group paying $880,000 and the others paying $25,000 each.
Modern recommendation systems use the concept of 'latent intent' to predict user behavior. A 2013 Cambridge University and Microsoft study of over 58,000 Facebook users showed that likes alone could predict personality traits, political views, and openness to new experiences. In 2025, researchers proposed the DUIP (Dynamic User Intent Prediction) framework, which combines an LSTM sequence model with a Large Language Model (LLM) to predict the next object a user will likely choose, outperforming previous methods on ML-1M, Games, and Bundle datasets.
Another 2025 study published in ACM Transactions on Information Systems described a model that uses LLMs to extract multi-faceted latent intent from a current session to provide recommendations. Additionally, a January 2026 study in the International Journal of Human-Computer Studies involving 231 participants showed that persuasion techniques like social proof, authority, and scarcity create a personalized approach to influence user choice, often leading users to prefer less useful options.
The process of recommendation systems creates a loop where the system reflects and simultaneously shapes user interests. By controlling the information environment and personalizing narratives, these systems can influence long-term identity, knowledge, and critical thinking. This is further exemplified by China's social credit system, which uses data on purchases, movements, and online activity to affect access to credits, travel, and government services.
