Academics Develop Plot to Scan Social Media, Flag Posts with “pre-crime” Machine Learning

Researchers from four US universities are floating the idea of employing machine learning (ML) in a pre-crime style in order to go through Instagram conversations and flag those the algorithms will be programmed to consider as "risky."

This is justified as a good way to protect young social media users from harassment and bullying, and advertised as a good balance between invading user privacy and "protecting" them by directly spying on their content.

The big picture of this effort seems to be looking for a way to bypass Instagram owner Meta’s intent to introduce end-to-end encryption as a way to protect privacy and safety of user data.

The proposal, that was recently published, comes from Drexel University, Boston University, Georgia Institute of Technology and Vanderbilt University researchers, and the model is based on utilizing metadata trails left by Instagram users to identify "content of interest" and still be able to automatically flag it, regardless of any encryption.

…