SPD Risk Analysis Recategorized 911 Calls More than Half the Time

Pie chart showing that SPD changed the analysis of the risk of 911 calls 54% of the timeBy Erica C. Barnett

The Seattle Police Department presented its initial “risk managed demand” analysis of 911 calls to the city council’s public safety committee earlier this week, a long-awaited presentation that was cut short because council members needed to get to Mayor Bruce Harrell’s budget speech. The eventual goal of the analysis, which looked at 356 call types and categorized them by the risk of harm they posed to callers, is to come up with a system for routing some calls to non-police responders or co-response teams, in which police serve as backup to service providers; this initial report is just a first step toward that eventual goal.

As we’ve reported, Mayor Bruce Harrell has adopted a go-slow approach to implementing alternatives to police, frustrating some council members who have been pushing for years to implement a pilot for responding to low-risk calls, along the lines of programs already in place in cities across the country, and see how it works. Earlier this month, Harrell’s office indicated they were open to a small pilot along the lines of Eugene’s CAHOOTS program or the Star program in Denver; his proposed budget for 2023 includes about $2 million for this purpose.

First, SPD gave each call type a risk number based on how the city currently responds to calls—an “all-hazards” response that always includes police. Then they applied a “mitigation” factor, essentially asking: What would this call look like if non-police responders were available?

SPD’s analysis looked at 356 different types of calls and categorized them by risk, taking two passes at the question. First, they gave each call type a risk number based on how the city currently responds to calls—an “all-hazards” response that always includes police. Then they applied a “mitigation” factor, essentially asking: What would this call look like if non-police responders were available to go out either alone or in tandem with police?

Between the two rounds, SPD data crunchers manually “recoded” more than half of all calls; 31 percent were “upgraded” to a higher risk level (meaning SPD believes police need to be present) and 23 percent were downgraded to a lower risk, based on the assumption that non-police responders would de-escalate a situation. SPD Chief Operating Officer Brian Maxey told the council that, for example, a welfare check might be categorized as Tier 1 (the highest-risk situation, requiring police), but further analysis would downgrade it to a Tier 3, which a social service provider could respond to with police backup.

In response to questions from committee chair Lisa Herbold and Councilmember Teresa Mosqueda, Maxey said SPD looked at the risk to people on the civilian end of 911 calls, rather than police, because the estimate of risk to police would have to be based on the current all-hazards response, which created the risk of “bootstrapping” the assumption that the risk to officers is generally higher than it is. “It seems like a completely different exercise to me,” Herbold said. “If you include use of force, then we are skewing the data in favor of a police response,” Maxey responded.

Herbold told PubliCola she was also concerned about the fact that SPD’s modeling required so much intervention by a human being who reversed the initial finding more than half the time. Since SPD’s model “only gets it right about half the time,” she said, “is this one performing the way SPD wants it to perform? Should we think about revising the formula?” It may be, ultimately, that there is no truly objective formula for pinpointing how much risk every 911 call poses to anyone, which means the best course of action could be moving forward with alternative responders based on the imperfect information we do have, rather than perfect information that will never be possible.

During the next phase of the analysis, SPD will begin setting up a “call triage system” in which bots that can process and categorize natural language to help 911 operators categorize calls based on their level of risk, using the “objective” measure of word frequency to augment call-takers’ human instincts. Accenture, a multinational management consulting and data analysis firm that has had a blanket contract with SPD since 2015, is developing that system now.

The public safety committee won’t meet again until December, after adopting the 2023 city budget.

3 thoughts on “SPD Risk Analysis Recategorized 911 Calls More than Half the Time”

  1. Getting it right half the time seems pretty good to me, considering the only information they get is from 911 calls.

  2. Ah, I’ve been to Eugene and although it much smaller community, it has the same levels (if not higher) of drug use, homelessness, petty crime and urban filth as any other NW town. CAHOOTS isn’t a solution any more than hiring more cops is a solution.

    I’d guess greater Seattle has over well over 5,000 people who need to be committed to a mental hospital for at least a year and spend the rest of lives in some sort highly supervised group home setting. We’re talking about people who have decades of untreated mental illness and drug abuse. This sort of long term intervention isn’t cheap, but it’s the only way forward. Hiring quasi-trained mental health workers for say, $22 a hour isn’t a solution… if you could even hire them. There’s no healthcare and no housing, so what tools would the “rent-a-cops” have to work with?

  3. I have a better idea. Since Eugene Oregon has been doing non-officer crisis response successfully for more than 30 years, and using what were then called police aides as an alternative for many non-crime-in-progress calls was studied and demonstrated successful in 1973, let’s just follow those models and cut the tendentious BS that’s a waste of time and taxpayer dollars.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.