CPD Twitter fail bIn all the grateful examination of police response since the Boston Marathon bombing, when two Cambridge men were identified as suspects and led police in a bloody chase through Watertown, there is one fact that has never been mentioned publicly: An automated alert system put in place by police broke.

The detail has been lost in the year since the April 15 bombings; April 18 armed robbery of a 7-Eleven, killing of MIT police officer Sean Collier and carjacking; and April 19 chase of the suspects into Watertown, where there were firefights, a 20-block search and finally the arrest of suspect Dzhokhar Tsarnaev, now 20. This is despite discussion of police response at a May 6 Public Safety Special Presentation; a May 13 debriefing before the City Council; and a Sept. 16-17 Police Innovation Conference co-hosted by Cambridge Police that featured two 45-minute discussions of the technology. One session, called “Twitter Automation of Dispatch Calls,” was devoted fully to a description of Cambridge’s program, including the offer of the city’s custom software code for use by police nationwide.

The conference was held four months after the department received questions about the Twitter outage.

Growing interest

The Cambridge Police Department’s Twitter account was updated by hand until Feb. 19, 2013, when officials followed the example of Seattle police and added custom software to also send certain tweets to the public automatically. The system takes calls in 17 categories from the department’s emergency dispatch center computers, gives officers a five- to 10-minute time delay to get to a crime scene and sends out an alert about what incident has been reported and roughly where. When the feature was rolled out, subscribers to the police Twitter account rose to more than 7,000 from 3,600, police officials said.

About two months later came the Boston Marathon bombing and its aftermath, and Twitter followers surged to nearly 14,500, said Dan Riviello, director of communications and media relations at the police department until a November move to a public relations firm.

The automated-tweet categories include armed robbery and attempted armed robbery; carjacking in progress; assault and battery with a dangerous weapon in progress; and “person with a gun,” and in the course of about three hours over April 18-19 all five might have been used – had an error not shut the automated system down.

Riviello and Police Commissioner Robert Haas denied requests in the months after the bombing to take questions in person or via email, but a Freedom of Information request resulted in responses to a fraction of the topics raised, including a Twitter error message dated April 19 but bearing no timestamp: “Twitter servers are up, but overloaded with requests.”

Not a critical system

Why did the automated system fail while other Twitter accounts, including non-automated Cambridge police accounts, kept working? Twitter didn’t respond to emailed questions, but Ken Pitts, public safety systems manager for Cambridge’s Emergency Communications Department, said blaming a surge in Twitter use was plausible.

“We didn’t build a whole lot of resiliency into our system. We don’t rely on Twitter for public safety purposes … it’s not critical to our operations,” Pitts said.

As a result, his department hasn’t changed anything to guard against a future shutdown and wasn’t sure if Twitter had, he said, because it wasn’t a priority.

“We spend a lot of time making sure critical infrastructure works,” he said.

The city also sends text message alerts to phones and email addresses through its Code Red, Citizen Observer and E-Line systems, referred to together as the Cambridge Alert Network. But in neither public marketing for the network nor in the two relevant sessions at the department’s Police Innovation Conference did Riviello or Haas make clear or even discuss a hierarchy among the options.

Described as important

“This is now one of the most important tools that we use to get our information out to the public,” Riviello said of the automated tweets. “If something serious occurs outside of business hours those who follow us on Twitter will know about it within minutes … in an emergency situation like that where people don’t have power they can turn to their smartphones and still be getting information from the police if they can’t turn on their TV or they can’t go online.”

That was consistent with what Haas said in introducing the automated Twitter system Feb. 19, 2013: “We believe strongly in the ability of social media to communicate with the public in a timely manner, and tweeting information about serious incidents will better inform our residents about what types of incidents police are responding to in their community.”

In offering the Cambridge software to other departments, Riviello acknowledged that “with any automated system there are bugs that we try to work out,” but he described a problem unrelated to the April 18-19 shutdown and assured listeners, “we’ve worked through that now.”

The bug that went unaddressed:

041714i-Twitter-error-message

Riviello said the automated system was started in part because he was an office of one, and police wanted the public to be informed of vital safety information even when the department’s director of communications and media relations was off duty. It was Pitts who took Twitter code and customized it to work with the department’s computer-aided dispatch, replacing a module that sent automated texts to cellphones. “The idea that I had for the system was that the command staff and other members of the department are able to get automated texts whenever we have certain serious calls that are entered into the dispatch system,” Riviello said. “That’s what we used as the basis for this system.”

Endorsements for the approach

Social media experts Dan Riviello, then of Cambridge Police Department, Cheryl Fiandaca, then of the Boston Police Department, and Trooper Dustin Fitch of Massachusetts State Police, during a September police technology conference. (Photo: Police Innovation Conference)

Social media experts Dan Riviello, then with Cambridge Police, Cheryl Fiandaca, then with Boston Police, and Trooper Dustin Fitch of the State Police, at a September police technology conference. (Photo: Police Innovation Conference)

Other area police public relations experts agreed with the approach, including Cheryl Fiandaca, chief of public information for Boston Police at the time of the bombings and now a reporter for WHDH television, and Dustin Fitch, of the Massachusetts State Police Office of Media Relations.

“Information is power. The most important thing is to let people know what is happening and what they should do or shouldn’t do,” Fiandaca said during the conference. She recalled the incidents of April in saying that “just like on Sept. 11, the phone service was not available. The only way to communicate – because it was a holiday, I wasn’t in my office – was through Twitter. That just in and of itself, letting people know what was happening, was really very valuable.”

As in Cambridge, Twitter followers for area police agencies soared during the weeklong crisis. The State Police account grew “exponentially,” in Fitch’s words. “The public was looking for real facts, real-time information. A lot of misinformation was coming out, and they were looking to us for the actual facts.”

Riviello explained the importance of the automated tweets to police officials from around the country by illustrating the difference in the social media approach of the Massachusetts Institute of Technology, where on Feb. 23, 2013, only a week after the automated tweets began, there was a Saturday morning report of an “active shooter” on campus.

Proof of concept

Police responded to the MIT campus Feb. 23, 2013, for what turned out to be a hoax report of a shooter. (Photo: Bree Sison via Twitter)

Police responded to the MIT campus Feb. 23, 2013, for what turned out to be a hoax report of a shooter. (Photo: Bree Sison via Twitter)

“This is what we consider our proof of concept,” Riviello said. “We fully mobilized our [special reaction] team and we went to the building – and about seven minutes after we got that initial call our automated tweet went out to people who were following us. So if you were following us on Twitter and you wanted to know what was going on, you’d see that we were responding to a report of a person with a gun. [Followers] can really know that we’re responding to this in real time and whether they need to shelter in place or take some sort of action.”

MIT, meanwhile, had “a completely human process. Somebody has to be called on the phone, they have to type that information in, they have to get it approved by administration,” Riviello said. “It took MIT over an hour to send an alert to their students. So in the follow-up to this they basically got hammered in the media for being so slow to respond.”

The report, which came through a lesser-used 911 relay for the deaf, wound up being a hoax.

Less than two months later, though, Cambridge faced an actual attack at the MIT campus when institute officer Sean Collier was shot multiple times in his cruiser near 32 Vassar St., in what some have conjectured was an attempt by the fleeing bombing suspects to get a second gun.

In this actual shooting, the automated system being down meant Cambridge police took 55 minutes, or nearly an hour, to tweet an alert to citizens. The shooting took place at 10:25 p.m. and Cambridge police first tweeted about it at 11:20 p.m., then retweeted an alert from MIT at 11:21 p.m.; The Tech, the campus newspaper, had tweeted first at 10:51 p.m., and the school had followed a minute later.

Other weaknesses

The system revealed some other weaknesses as well.

In the days after the April 15 bombing at the marathon, people were on guard and taking no chances when they encountered abandoned boxes or bags. Police were getting a surge in reports of suspicious packages around the city, which the automated system reported in turn.

Each suspicious package had to be investigated to be ruled out as a bomb, and Riviello said police tried to follow up on each report to let residents know they were safe. But those tweets had to be done manually, and the sheer numbers made it hard to keep up. In the eight days after the Boston Marathon, there were 70 automated tweets reporting suspicious packages, but a third of them – 24 out of 70 – never got follow-up tweets to give the all-clear signal.

042913i CPD tweets b

In addition, the ability of official police tweets to combat misinformation stumbled. The department’s second automated tweet of April 19, as the system recovered, was of a possible armed robbery on Memorial Drive about 11 minutes earlier, at 12:19 a.m. But the location and timestamp marks that as the origin of a 911 call in which a victim escaped and reported a carjacking that had taken place an hour and a half earlier in Allston.

The tweet made it harder to piece together an already muddled timeline of the night’s events, and reports that the Tsarnaevs had robbed a 7-Eleven added to the confusion. Margaret Chabris, a spokeswoman for the minimart chain, said she had to repeatedly knock down the rumor, insisting that the robber seen on security camera footage at the robbed store looked nothing like either bombing suspect.

“Some of them say they got the information from a credible source,” she told Maclean’s magazine, referring to other reporters. “And I say, yeah, they may be a credible source, but they’re incorrect.”

Muddled timeline

The robbery took place at 10:28 p.m. in Central Square, and the shooting of Collier in Kendall Square took place some three minutes earlier and a mile away in Kendall Square. Cambridge has more than one 7-Eleven, and there was no automated tweet or follow-up from police to say which had been robbed.

On the other hand, the tweet from Memorial Drive conflicted with scanner reports and postings from journalists and other observers, resulting in mounting confusion even as the action shifted definitively to Watertown. The media tried to set down a timeline of events, but the tangle of facts led outlets including NPR to talk about “reports of an armed carjacking in Cambridge.”

Attempts to straighten out the timeline failed. Cambridge Police were called the morning of April 19 (as the action had quieted somewhat and a search of Watertown streets and homes was underway) with requests to identify the time and address of the 7-Eleven robbery and carjacking. Officers said they did not know where Riviello was – in fact he was in Toronto, working from afar via a smartphone with social media apps – or if he would be available as the business day began.

They would not answer the question about the 7-Eleven or carjacking themselves. One refusal to answer came from commanding officer Superintendent Steven A. Williams.

Policy for media, public information

This is against department policy as laid out by Riviello, Haas and written guidelines sent as a result of the Freedom of Information request. In Riviello’s view, “if there’s a specific crime spree or something we need to address, we’re putting a uniformed member of the department in front of a camera – because they don’t want to see me in a suit, right? They want to see a police officer in uniform. Our command step is also trained and responsible for interacting with the media when necessary. That’s all of our deputies and our superintendents, and of course the commissioner.”

Haas said, “The Cambridge Police Department is committed to sharing information openly with the public in a timely manner, whether that be during a crisis or otherwise,” and that “the entire command staff is trained and able to release information as well” as a communications specialist such as Riviello.

The written policy for command staff authorized and trained to answer questions is found in the department’s Media/Public Information Guidelines, where were last updated in 2011. “Members of the Cambridge Police Department who are authorized to release public information shall make every effort to … be available for on-call responses to the news media,” it says, as well as that:

The department’s goal in sharing public information is to appropriately inform the community as to those events that are of interest and to maintain a positive relationship of mutual trust, cooperation and respect with the news media. This can be best accomplished by providing the community with accurate and timely information on events that affect the lives of citizens in the community and the department’s administration and operations. Authorized personnel will be expected to release public information with openness and candor while maintaining the privacy rights of individuals and the integrity of criminal investigations. The department is committed to informing the community and the news media of events within the public domain that are handled by or involve the police department.

Thinking back on the madness following the bombings, Riviello said in September that “We were just trying to share information so that people knew what we were doing, how we were responding and what we were asking them to do … There was so much misinformation. Everybody was tweeting about this. They were in their homes watching TV 24 hours a day trying to see what the next little piece of news would be. They knew the could turn to us for what was actually true.”

“We always remind people that we believe an informed community is a safer community,” he said.