Meta Patents AI That Listens and Tracks User Emotions

Meta Platforms filed a patent application this week for an artificial intelligence system capable of continuously listening to user conversations and analyzing their emotional states. The proposed AI would process vocal cues to infer how a person is feeling and maintain a timestamped log of these emotional readings. This log would be associated with specific moments, including the time, the user's location, their activity, and how they were interacting with their device.

The patent, titled "Systems and Methods for Inferring Emotional State from Vocal Characteristics," outlines systems that could operate throughout the day, potentially capturing a comprehensive emotional profile of the user. Some described embodiments would continuously monitor audio, while others might activate based on specific triggers. The AI would analyze various vocal characteristics such as pitch, tone, speaking rate, and pauses to determine emotional states like happiness, sadness, anger, or stress.

Beyond emotional inference, the patent details how the system could correlate these emotional readings with contextual information. This includes data points like the user's geographical location, their current activity (e.g., working, commuting, exercising), and their device usage patterns. The goal is to create a rich, time-series dataset of a user's emotional journey throughout their day, potentially for personalized user experiences or targeted advertising. The patent application was filed on May 23, 2024, with the United States Patent and Trademark Office (USPTO).