01/11/2026
We submitted two different public records requests from about the same time frame in the city of Federal Way Washington from our Police Department.
Record Request Batch 1 – Violent incidents with legal constraints
“Please provide any case reports, CAD reports, referral slips, or receipts involving juveniles or youth with the following keywords from March 2024 to present: Threat with a Weapon, Domestic with a Weapon, Assault with a Weapon, Fight with a Weapon, Carjacking, Stabbing, or Shooting. We are not interested in water guns or sq**rt guns. If possible, please include any records where an interview or a search of a suspicious subject for a weapon was limited by current laws.”
This request was intended to capture situations where violent crimes or credible threats occurred, but officers were legally limited — including limits on searches, interviews, detention, or charging. This request surfaced unique situations to include behavioral health crisis response, legal restraint, victim non-cooperation, medical treatment and transport and information-only reports
Some of these incidents did not result in a crime being coded or resulted in a downgraded crime, even when violence risk or weapon involvement was present.
That does not mean the threat wasn’t real — it means the response pathway changed, and with it, the data footprint.
Batch 2 – Crime-coded incidents
“Please provide any case reports or CAD reports involving juveniles as suspects from March 2024 to present for the following offenses: Harassment, Assault 1, Assault 2, Carjacking, Burglary 1, Homicide, Prostitution, Possession of Stolen Property 1, Robbery 1, Theft of Vehicle, Weapons Offense, and VUCSA.”
This request was designed to capture officially coded crimes — incidents where probable cause thresholds were met and the response followed a criminal justice pathway. This is the data that appears in crime statistics, quarterly reports and public discussions about crime trends.
Neither batch comes first or last. They capture different scenarios, not different stages. What we found when we compared them:
Batch 1 includes many cases where threat behavior is explicit — guns, knives, threats to harm or shoot — but often embedded in narrative text instead of coded offenses. There is a much higher presence of mental health crisis, emotional escalation, legal restraints, medical transport, and victim non-cooperation.
Batch 2 mostly involves people clearly engaged in criminal conduct. Mental health still appears, but it is less dominant than violence history, with more focus on criminal justice outcomes. There are strong indicators of prior cases, repeat behavior, and cross-jurisdiction overlap — yet prior cases and other jurisdictions being disconnected often appear to reset accountability, without continuity of intervention.
Across both batches, the same patterns repeat: Family instability as a core driver, emotional escalation and crisis points, school discipline issues and school mobility, social media and peer pressure and prior incidents that don’t follow the person across systems.
Together, the data raises a hard question:
How many missed opportunities were there for earlier, stronger intervention before violence escalated? One of the biggest gaps is that repeat behavior and multi-system impact are not consistently tabulated across systems.
When prior incidents, school discipline, mental health responses, legal restraints, medical transport, and police contacts are all held in separate silos, decision-makers are forced to assess risk one incident or issue at a time, without seeing the full pattern.
Tabulating repeat contacts, cross-jurisdiction involvement, prior diversion attempts, prior crisis responses, and impacts on victims, schools, families, and the community would help determine what the next step of accountability should be — whether that is continued diversion, a higher level of intervention, or, when appropriate, arrest and booking.
Without that cumulative view risk can appear “new” when it is not, accountability resets instead of escalating appropriately and systems struggle to justify stronger responses even when patterns exist, matching the response to the actual level of risk, informed by the full history.
A major takeaway is how diversion and legal limitations change what gets counted. When firearm-related incidents are routed into crisis response, behavioral health framing, or information-only pathways — or when officers are legally constrained from searching, interviewing, or charging — many never become firearm offenses in crime statistics, even when a gun or credible threat was present.
This produces underreporting of firearm-related behavior, artificially lower weapon-related crime statistics, and provides a false sense that “interventions are working”.
This doesn’t mean diversion or legal safeguards are wrong. People want the mental-health-sensitive response reflected in Batch 1. But it does mean we should be honest about how these pathways affect the data — and acknowledge that Batch 2 often reflects what happens after the same pattern can no longer be downgraded or legally constrained.
Some Batch 1 cases may have been more appropriately handled through a Batch 2-type response if more information from other systems had been available earlier. The biggest issue is that our systems work in silos which create information vacuums. Prior responses aren’t visible, successes aren’t shared, accountability resets instead of continuing, efforts and resources get duplicated and it becomes difficult to determine next steps or who all is impacted.
In both Batch 1 and Batch 2, more information from other systems could sometimes help put events in context. That context could clarify whether actions were aggressive or defensive, whether someone was responding to ongoing threats, whether self-defense should be ruled out or identified, and whether earlier interventions already occurred. Right now, that context is often missing.
Police reports are publicly disclosable, audited, statistically counted and politically scrutinized.
Other systems are confidential, shielded, voluntary and rarely audited against outcomes.
Systems operate in isolation, guessing what “worked” and funding persists based on narratives and partial data. These two datasets don’t contradict each other, they complete each other. One shows what gets counted. The other shows what occurred but could not always be fully acted on or counted. Without shared context, risk doesn’t disappear — it just moves. And without continuity, the system keeps relearning the same lessons too late.
Note: Click on each image and you can zoom in on the content or try looking at the content on a desktop.
