Jessica W. Gillooly is a PhD candidate in sociology and public policy at the University of Michigan.

The news about the white Yale University student who called 911 on a black classmate who fell asleep in the common room of a dorm didn’t surprise me. Nor was I surprised by the white woman in Colorado who called 911 on two Native American brothers taking a campus tour because they joined the tour late and were acting “just really odd,” or the Philadelphia Starbucks employee who called 911 on two black men for not making a purchase.

As a 911 call-taker, I process countless requests from citizens and rarely deny police services, no matter how benign a situation appears to be. I have handled a call from someone who found it suspicious that an elderly Asian man was walking on the side of the road. I have handled a call to settle a dispute over a pet peacock defecating on a neighbor’s front lawn. I have handled a call from a man who felt uncomfortable at the bus station because a black teenager’s jeans were hanging too low.

Each of these calls received a police dispatch because of the caller’s expectations about 911 and training policies on public safety answering points (PSAPs) that emphasize customer service. Calls for service are the public’s most common form of interaction with law enforcement. According to the Bureau of Justice Statistics, in 2011, of the estimated 62.9 million U.S. residents who had one or more contacts with the police, 51 percent of those contacts were because a citizen requested police services. Requests, however, come from citizens who are not trained in criminal law, have their own biases about people and places, and maintain complex personal relationships. These factors can lead to excessive, unjustified and racially disparate police-citizen contacts.

Of late, prominent solutions to racially driven police encounters have focused on retraining citizens in their use of 911 services. Starbucks’s planned “implicit bias” training program is one such attempt. But scaling up implicit bias training to entire communities is impractical, and evidence of its effectiveness is unconvincing. A 2017 meta-analysis of 494 studies on implicit bias by psychologist Patrick Forscher and colleagues found that while small changes in implicit bias are possible, these changes often do not translate into behavioral actions.

A more realistic solution lies in the call-taker, not in the call-maker. A primary duty of a 911 operator is to interpret a caller’s request and determine if it merits police attention. They are extensively trained to do so. Policing requires effective gatekeepers to filter out requests that should be handled by alternative agencies or service providers. But gatekeeping is hard work in an environment of uncertainty and confusion. Call-takers cannot see what is happening on-scene and understandably rely on the motto “When in doubt, send them out” to avoid liability.

Yet activating the police, however innocent the callers’ motives, has unpredictable, sometimes serious, results. The call at Yale meant the African American student had to justify her presence to the police for 20 minutes, the call from the woman in Colorado ruined a college visit for two teens, and the call from the Starbucks employee resulted in the arrest of two men. In their book, “In Context,” police detective Nick Selby and co-authors Ben Singleton and Ed Flosi wrote that in 2015, 83 of the 153 police killings of unarmed civilians began with a 911 call. Although most calls do not end like this, erring on the side of caution by indiscriminately sending the police overlooks the associated risks to families, communities and police officers.

PSAPs and police departments can adopt several policies to mitigate the problems posed by ambiguous calls. First, 911 operators need clearer protocols on how to handle ambiguous calls, be it a suspicious person or a neighbor dispute. For example, a man walking down the street trying to open car doors may meet the necessary criteria for a suspicious-person call, whereas a man sitting on a curb would not. Second, call-takers need to be trained in pressing callers to articulate their underlying suspicion if they report that someone “doesn’t belong here.” If the caller fails to articulate their suspicion, operators should record and pass along that information to appropriately prime responding officers. Third, 911 public awareness campaigns should align expectations between callers and call-takers as to the types of questions callers will face when reporting suspicious activity, including asking for the caller’s name and call back number to ensure greater accountability. Armed with such training, I could have gathered more information from the caller who complained about the teenager’s pants “hanging too low” to distinguish indecent exposure, which is a police matter, from a dress-code issue, which is not, and thus avoided a possibly unnecessary police-citizen interaction.

While greater public awareness of implicit bias can’t hurt, there is no way to ensure that callers will always make justified requests. For that reason, policymakers must consider reforms for call-takers, not just call-makers. Some may not agree with these ideas, including others in my own agency, but it is time for conversations about policing to include the role of dispatchers.

Read more: