For example FBI trawling through social media and flagging accounts for review doesn't seem like a strict constitutional violation unless you can prove some coercion; but it does seem like a questionable use of the FBI to be doing content moderation for Twitter absent a criminal investigation.
Maybe this sort of stuff (i.e. government/law enforcement communications with social media companies) should be automatically in the public domain if it isn't part of a criminal investigation.
Law enforcement, people in corporate risk, and business intelligence groups very often use social media to perform open source research. In fact, when they don't, people exclaim that the police are incompetent for not knowing the shooter said he was going to shoot a place up after a history of deranged posting and a call to violence on facebook or something.
There are platforms built on top of Facebook/Twitter/Reddit/Etc that will do automated sentiment analysis, key into keyword trends, etc. You can build reports, see how much impact or reach certain calls to violence have, etc. and then triage the threats that you think are credible vs. those that are full of shit.
There are people out there trying to recruit people to firebomb tax buildings. Telling Twitter to shadowban them before they develop networks and get off the platform is a lot more innocuous than raiding the guy's house to see if he has bombmaking equipment.