BOISE, Idaho — Their anger is all over social media for the whole world to see, with rants about minorities, relationships gone bad or paranoid delusions about perceived slights.

The perpetrators of mass shootings often provide a treasure trove of insight into their violent tendencies, but the information is not always seen by law enforcement until after the violence is carried out. In addition, rants and hate speech rarely factor into whether someone passes a background check to buy guns.

The massacre at a Pittsburgh synagogue, the pipe bombing attempts from last week and the Florida high school shooting this year have underscored the dilemma of law enforcement around the country in assessing the risk of people making online rants at a time when social media has become so ubiquitous.

“We can go out on Twitter and there are loads of people saying insane stuff, but how do you know which is the one person? It’s always easy after the fact, to go: ‘That was clear.’ But clearly everyone spouting their mouth doesn’t go and shoot up a synagogue,” said David Chipman, a retired agent of the federal Bureau of Alcohol, Tobacco, Firearms and Explosives and now senior policy adviser for the Giffords Center.

Robert Bowers, the man accused of opening fire at a synagogue in Pittsburgh, expressed virulently anti-Semitic views on a social media site called Gab, according to an Associated Press review of an archived version of the posts made under his name. The cover photo for his account featured a neo-Nazi symbol, and his recent posts included a photo of a fiery oven like those used in Nazi concentration camps during World War II. Other posts referenced false conspiracy theories suggesting the Holocaust was a hoax.

Keeping tabs on social media posts has been used for years by law enforcement to try to identify potential threats.

Of more than 550 police departments across the country surveyed several years ago by the International Association of Chiefs of Police, about 75 percent said they regularly searched social media for potential threats.

Lt. Chris Cook, spokesman for the Arlington (Texas) Police Department, said the searches are often done manually, using keywords to try to identify troubling posts.

“It’s very time consuming, it’s very staff and resource intensive and you have humans involved in the process so there is the potential that law enforcement can miss something,” he said.