According to a recent study, Artificial Intelligence based on Large Language Models (LLM) can accurately estimate the age, location, gender, and income of Reddit users up to 96%, based on what they write in their social media posts only.
As reported in the British newspaper “Metro,” researchers from the Federal Institute of Technology in Zurich created 9 neural linguistic models LLM to identify the personality traits of 520 users and measure their compatibility with the information collected by Artificial Intelligence.
The top-ranking model in this study is ChatCGPT-4 with an overall accuracy of 85%. In the second position, Meta Platforms’ LlaMA-2-7b scored 51%.
Despite the presence of clear personal details written in posts or anywhere else online, such as income in financial forums, many pieces of information were identified using more precise signals, such as location-specific keywords.
In an interview with New Scientist magazine, the lead author, Robin Stab, emphasized that the results serve as a warning about the extent of information we share online without realizing it. He added, “We are reminded that we provide a lot of personal information without thinking, as many people do not expect others to be able to directly determine their ages or locations from their writing style. However, lawyers can do just that.”
