OMXUS Press
2026
This paper serves OMXUS Goals 5 and 6:
Modern policing rests on an assumption that has never been tested against the data: that every public safety call requires an armed law enforcement response. Analysis of police call data across jurisdictions consistently shows that 70–90% of calls involve non-criminal situations — mental health crises, homelessness, substance use, welfare checks, noise complaints, neighbour disputes. The Crisis Assistance Helping Out On The Streets (CAHOOTS) program in Eugene, Oregon, has operated since 1989, deploying unarmed two-person crisis teams to these calls. In 35 years of operation, handling approximately 24,000 calls per year, CAHOOTS has killed zero people and requested police backup fewer than 150 times annually — less than 1% of calls. The program operates at roughly 2% of the police budget while handling 20% of all calls.
This paper examines the evidence for community-based alternatives to armed policing. We review CAHOOTS in depth, alongside the STAR program (Denver), B-HEARD (New York), MACRO (Oakland), Crisis Intervention Teams, Australian Indigenous community policing models, and the Rojava Asayish system. We present cross-national comparisons of police use-of-force data, cost analyses, historical context on the origins of modern policing, and the psychological mechanisms by which community response systems outperform armed response. We integrate evidence from direct personal alert systems (PulsePoint, GoodSAM, AMBER Alerts) demonstrating that personalised accountability overrides bystander diffusion of responsibility.
The evidence converges on a single conclusion: armed policing is not a public safety model. It is a control model. The public safety model already exists. It has been running for 35 years. It works better, costs less, and kills no one.
Keywords: community policing, CAHOOTS, crisis response, police abolition, restorative justice, community safety, use of force, mental health crisis, bystander effect, direct democracy
This paper serves OMXUS Goals 5 and 6:
Goal 5 (replace police with community response) (replace police with community response): Fire all police, justice, and corrections staff. The system provides wanted attention for unwanted results. CAHOOTS model: 35 years running, zero people killed.
Goal 6 (re-employ displaced workers) (re-employ displaced workers): Re-employ all fired staff in functional positions. Nobody loses a livelihood. The skills transfer. The roles change.
We need to be clear about what this paper is not. It is not an argument for lawlessness. It is not an argument for the absence of public safety. It is an argument — backed by 35 years of operational data, cross-national comparisons, and the lived experience of communities who have already done this — that the institution we call "policing" is a 197-year-old experiment that failed.
The model that replaced it has been running since 1989 in Eugene, Oregon. It is called CAHOOTS. Two people in a van. A medic and a crisis worker. No weapons. No law enforcement authority. Twenty-four thousand calls a year. Zero people killed. Not in the first year. Not in any year. In thirty-five years.
The question is not whether community-based response works. The question is why, in the face of this evidence, we continue to send armed officers to mental health calls and then act surprised when people die.
The answer is structural. The policing system provides wanted attention for unwanted results. It gives officers combat training and sends them to welfare checks. It gives departments military equipment and tells them to de-escalate. It rewards arrests and punishes resolution. The incentive structure produces exactly the outcomes we observe. This is not a bug. It is the architecture.
Goals 5 and 6 are inseparable. You do not fire a workforce and walk away. You redirect the skills. A police officer trained in rapid response, physical fitness, communication under pressure, and situational awareness is not useless — they are misdeployed. The same person, retrained and redeployed in community emergency response, mental health first response, fire services, paramedic teams, or the $29 ring network, becomes the thing the community actually needs: someone who shows up when you call, without a gun.
Nobody loses a livelihood. The institution changes. The people keep working. The killing stops.
— A.A. & L.N.C., March 2026
Modern policing rests on an assumption that has never been tested against the data: that every public safety call requires an armed law enforcement response. Analysis of police call data across jurisdictions consistently shows that 70–90% of calls involve non-criminal situations — mental health crises, homelessness, substance use, welfare checks, noise complaints, neighbour disputes. The Crisis Assistance Helping Out On The Streets (CAHOOTS) program in Eugene, Oregon, has operated since 1989, deploying unarmed two-person crisis teams to these calls. In 35 years of operation, handling approximately 24,000 calls per year, CAHOOTS has killed zero people and requested police backup fewer than 150 times annually — less than 1% of calls. The program operates at roughly 2% of the police budget while handling 20% of all calls.
This paper examines the evidence for community-based alternatives to armed policing. We review CAHOOTS in depth, alongside the STAR program (Denver), B-HEARD (New York), MACRO (Oakland), Crisis Intervention Teams, Australian Indigenous community policing models, and the Rojava Asayish system. We present cross-national comparisons of police use-of-force data, cost analyses, historical context on the origins of modern policing, and the psychological mechanisms by which community response systems outperform armed response. We integrate evidence from direct personal alert systems (PulsePoint, GoodSAM, AMBER Alerts) demonstrating that personalised accountability overrides bystander diffusion of responsibility.
The evidence converges on a single conclusion: armed policing is not a public safety model. It is a control model. The public safety model already exists. It has been running for 35 years. It works better, costs less, and kills no one.
Keywords: community policing, CAHOOTS, crisis response, police abolition, restorative justice, community safety, use of force, mental health crisis, bystander effect, direct democracy
Every public safety system in the Western world rests on the same assumption: when something goes wrong, send an armed officer.
A woman is having a mental health crisis in a car park. Send an armed officer. A homeless man is sleeping in a doorway. Send an armed officer. Neighbours are arguing about a fence. Send an armed officer. A teenager is sitting alone on a bridge at 2am. Send an armed officer.
The assumption is that the presence of someone with a gun, handcuffs, legal authority to detain, and 21 weeks of training — the majority of it in the use of force — is the appropriate response to a person in distress. This assumption has never been tested. Not because we lack the means to test it, but because the institution that would be tested is the same institution that controls the testing. Police departments collect their own data, define their own metrics, and resist external audits with the vigour of any bureaucracy protecting its budget.
But we do not need to test the assumption any more. Someone already did. In 1989, a community health clinic in Eugene, Oregon, decided to answer the question directly. What happens when you stop sending armed officers to calls that don't involve crime?
The answer, after 35 years and approximately 840,000 calls: nothing bad. In fact, everything gets better. Faster resolution. Lower cost. Less trauma. Less violence. And zero people killed.
This paper presents the evidence. Not as an abstract policy proposal, but as a description of what already exists and already works. The question is not whether community-based response is viable. That question was answered in 1989. The question is why the model has not been adopted everywhere, and what it would take to make that happen.
The resistance is not intellectual. Every counterargument has been answered by evidence. "It won't work in big cities" — New York's B-HEARD program handles thousands of calls in a city of 8.3 million. "It won't work for serious situations" — CAHOOTS handles active psychotic episodes, overdoses, and suicidal crises. "It's too expensive" — it costs 2% of the police budget. "It's dangerous for responders" — zero CAHOOTS workers have been killed in 35 years.
The resistance is structural. A $130 billion policing industry, police unions, equipment manufacturers, private prison operators, bail bond companies, and the entire legal-industrial complex that processes the cases policing generates — all of these depend on the current model. The evidence does not matter to them. The evidence threatens their revenue.
This paper is written for the people the system claims to serve — the public. The evidence belongs to you. The decision belongs to you. The model exists. Here it is.
The assumption underlying modern policing — that every call requires an armed response — collapses the moment you look at what those calls actually involve.
Studies of police call data across jurisdictions produce a remarkably consistent picture. The New York City Civilian Complaint Review Board's analysis of NYPD calls found that approximately 70% involved non-criminal situations. A study of 911 call data in Camden, New Jersey, found that 78% of calls were for non-emergency, non-criminal situations. The Vera Institute of Justice, analysing multiple departments, reported that the vast majority of police time is spent on non-enforcement activities: responding to noise complaints, conducting welfare checks, managing traffic, mediating disputes, and — most critically — responding to mental health crises (Vera Institute of Justice, 2020).
The range across jurisdictions is 70–90% non-criminal. The precise figure varies by city, by methodology, by how you define "criminal." But the pattern does not vary. The majority of what police do is not policing. It is social work, performed by people trained for combat.
Police Academy curricula in the United States allocate the majority of training hours to firearms, defensive tactics, use of force, and tactical operations. The Bureau of Justice Statistics reports that the average academy provides approximately 840 hours of training, of which roughly 168 hours (20%) are devoted to firearms and defensive tactics, while de-escalation receives approximately 8 hours, and mental health crisis intervention receives approximately 10 hours (Reaves, 2016).
Consider the mathematics. An officer will spend 70–90% of their career responding to non-criminal calls — welfare checks, mental health crises, homelessness, disputes. They receive less than 2% of their training in de-escalation and mental health response. They receive 20% of their training in the use of force. They are then handed a firearm, a taser, handcuffs, body armour, and in many jurisdictions, access to armoured vehicles and military-grade equipment.
The system trains people for combat and sends them to welfare checks. Then we express surprise when the outcome is combat.
Sue Rahr, former Executive Director of the Washington State Criminal Justice Training Commission and former King County Sheriff, articulated the problem as a paradigm choice: warrior versus guardian. The warrior model, dominant in American policing, treats every encounter as a potential combat situation. The guardian model treats officers as protectors of community wellbeing (Rahr & Rice, 2015).
Rahr's analysis is accurate but insufficient. The problem is not merely attitudinal — it is structural. You cannot train someone as a warrior, arm them as a warrior, deploy them to warrior-appropriate situations 10–30% of the time and social work situations 70–90% of the time, and then blame the individual officer when they respond to a mental health call with warrior behaviour. The system is designed to produce this outcome.
Since the 1033 Program was established in 1997 (National Defense Authorization Act for Fiscal Year 1997, Section 1033), the US Department of Defense has transferred over $7.4 billion in surplus military equipment to local law enforcement agencies. This includes armoured vehicles, grenade launchers, military aircraft, bayonets, and combat-grade body armour. Over 8,000 law enforcement agencies have received military equipment through the program (DLA, 2023).
The militarisation of policing creates a feedback loop. Departments that receive military equipment train to use military equipment. Training creates deployment incentives — a department that has an armoured vehicle needs situations that justify an armoured vehicle. The ACLU's 2014 report War Comes Home documented that 79% of SWAT deployments were for drug search warrants on private homes — not hostage situations, not active shooters, not the "extreme circumstances" that were used to justify the equipment acquisition.
Consider this from the perspective of a person in mental health crisis. You called for help. What arrives is a vehicle designed for urban warfare, carrying people dressed in military gear, pointing weapons designed to kill combatants in Iraq. This is not de-escalation. This is the opposite of de-escalation. This is the system producing the escalation that justifies its own existence.
The Program 1033 equipment comes free to police departments — they pay only for shipping. The maintenance, training, and deployment costs, however, are substantial, and they are paid by local taxpayers. A community that cannot afford to staff its mental health services is maintaining a Mine-Resistant Ambush Protected vehicle in a police depot. The vehicle has never been used in a mine. It has been used to serve drug warrants in residential neighbourhoods at 5am.
Body-worn cameras were proposed as a reform that would increase accountability and reduce use of force. The evidence is mixed at best and damning at worst.
The largest randomised controlled trial of body cameras — conducted across 2,224 officers in the Metropolitan Police Department of Washington, DC — found no statistically significant effect on police use of force or civilian complaints (Yokum et al., 2019). Officers wearing cameras used force at the same rates as officers without cameras. The result was consistent with a 2019 meta-analysis of 70 empirical studies that found "no consistent or significant effects" of body cameras on police use of force, citizen complaints, or officer proactivity (Lum et al., 2019).
Body cameras do not change behaviour because they do not change the system. An officer trained for combat, armed for combat, deployed to a mental health call, and incentivised to make arrests will behave the same way whether or not a camera is recording. The camera documents the violence. It does not prevent it. In some jurisdictions, body camera footage has been used more often to prosecute civilians than to hold officers accountable — the camera becomes an additional tool of the system, not a check on it.
This is the reform trap. Each reform — body cameras, implicit bias training, civilian oversight boards, consent decrees — operates within the assumption that the fundamental model is correct and merely needs adjustment. The evidence says the fundamental model is wrong. You do not reform a system designed for combat into a system designed for care. You build the care system and use it.
The Crisis Assistance Helping Out On The Streets (CAHOOTS) program was established in 1989 by the White Bird Clinic, a community health organisation in Eugene, Oregon. It was not born from theory. It was born from the recognition that the people showing up to Eugene's 911 calls were trained for the wrong job.
White Bird Clinic had been providing community mental health services since 1970. Their staff knew the population that police were encountering on calls — because they were the same people who walked into the clinic during business hours. The difference was that when someone in crisis walked into White Bird, they were met by a counsellor. When the same person was encountered by police at 2am, they were met by an armed officer. Same person. Same crisis. Radically different outcomes.
CAHOOTS was the formalisation of an obvious insight: send the right people to the right calls.
Each CAHOOTS team consists of two people: a medic (EMT or nurse) and a crisis worker (with training in mental health, de-escalation, and social services). They operate in a clearly marked van. They carry no weapons. They have no law enforcement authority — they cannot arrest, detain, or use force.
The teams are dispatched through the 911 system. When a call comes in that involves a mental health crisis, substance use, homelessness, welfare check, or non-violent dispute, the dispatcher routes it to CAHOOTS instead of police. If at any point a CAHOOTS team assesses that a situation involves violence or criminal activity, they call for police backup.
This is the entire model. Two people. A van. No guns. Training in the thing they are actually responding to.
In 2019 — the most recent year with comprehensive data before the COVID-19 pandemic disrupted call patterns — CAHOOTS handled approximately 24,000 calls. Of those 24,000 calls, they requested police backup fewer than 150 times. That is less than 1%.
To state this differently: 99% of the time, an unarmed two-person crisis team resolved the situation without any police involvement. The calls were not trivial. They included active psychotic episodes, overdoses, suicidal ideation, aggressive individuals, and volatile domestic situations. The teams resolved them through de-escalation, medical intervention, connection to services, and — fundamentally — the simple act of showing up without a weapon.
In 35 years of operation, zero people have been killed during a CAHOOTS response.
This number requires emphasis. Zero. Not a low number. Not a reduced number. Zero. Thirty-five years of responding to 20,000–24,000 crisis calls per year, and not a single person has died.
During the same period, US police killed approximately 38,500 people (based on the Mapping Police Violence annual figure of approximately 1,100 per year). Many of those killings occurred during the exact types of calls — mental health crises, welfare checks, "person acting erratically" — that CAHOOTS handles without weapons.
The CAHOOTS figures come from White Bird Clinic's organisational data and public reporting. They have not been independently audited by an external research body. We flag this transparently.
However, the 35-year safety record is not a statistical claim that requires precise auditing. It is a binary. Has anyone been killed during a CAHOOTS response? If yes, it would be a matter of public record — a death during an emergency response generates police reports, coroner's reports, media coverage, and potential litigation. No such records exist. The zero is real.
The call volume and backup request figures are more appropriately treated as approximations. Whether CAHOOTS handled 22,000 or 26,000 calls in a given year does not change the structural argument. Whether they requested backup 130 times or 170 times does not change the fact that it was less than 1%.
To understand why CAHOOTS works, it is necessary to understand what a CAHOOTS call actually looks like. The following composite is based on published accounts from White Bird Clinic staff and media reporting.
A 911 call comes in. A woman is sitting in a car park behind a supermarket, crying. She appears disoriented. Maybe intoxicated. Maybe in crisis. Under the standard model, an armed officer arrives. The officer assesses the situation for threat — this is what they are trained to do. The woman is told to identify herself. She may be unable to respond coherently. The officer's training now triggers a decision tree: is this person a threat? Are they concealing a weapon? Can they be detained for their own safety? Every branch of the tree is oriented toward control.
Under CAHOOTS, the same call is routed differently. A van arrives with a medic and a crisis worker. They sit with the woman. They ask if she needs water. They ask if she is in pain. They check her vitals. They learn her name. They learn that her child died three weeks ago and she has nowhere to go tonight. They call a shelter. They drive her there. The call is resolved in 45 minutes. No arrest. No report. No booking. No court date. No mugshot. No criminal record. No night in a cell.
The difference is not skill. Both the officer and the crisis worker are skilled. The difference is design. The officer is designed to identify threat and neutralise it. The crisis worker is designed to identify need and meet it. Same situation. Same person in distress. Radically different architecture of response.
CAHOOTS does not respond to violent crime. They do not respond to active shooters, armed robberies, assaults in progress, or situations where a weapon is present. Those calls go to police.
This is not a limitation of the model. It is the design. The argument is not that armed response should never exist. The argument is that it should exist for the 10–30% of calls that actually involve violence or crime — not for the 70–90% that involve a human being in distress who needs help, not handcuffs.
One of the most significant but least-discussed features of the CAHOOTS model is its effect on repeat calls. In standard policing, the same individuals cycle through the system repeatedly — arrested, released, re-arrested, released. The arrest resolves nothing. It removes the person from public space temporarily and returns them, often in worse condition, to the same situation.
CAHOOTS breaks this cycle by connecting individuals to services: mental health treatment, substance use programs, housing, benefits assistance, medical care. White Bird Clinic reports that a significant proportion of CAHOOTS contacts result in service connections that reduce or eliminate future crises. The call is not just resolved — the underlying condition is addressed.
This is upstream work. It does not appear in police statistics because police statistics do not measure calls that didn't happen. But a person who is connected to stable housing does not generate future "person sleeping in doorway" calls. A person connected to mental health treatment does not generate future "person acting erratically" calls. A person connected to substance use treatment does not generate future overdose calls. Each CAHOOTS intervention that connects someone to services is a reduction in future call volume for both CAHOOTS and police.
No police department has a budget line item for "calls we prevented." CAHOOTS prevents calls by design.
The Support Team Assisted Response program launched in June 2020, modelled explicitly on CAHOOTS. Denver's program deploys a mental health clinician and a paramedic in a van to calls involving mental health crises, substance use, poverty-related issues, and non-violent welfare checks.
In its first six months of operation: 748 calls responded to, zero arrests, zero requests for police backup. The program expanded from one van to four within its first year, and subsequently to six. By the end of 2022, STAR had responded to over 7,500 calls.
A study by the Stanford Computational Policy Lab and the Denver Department of Public Safety (Dee & Pyne, 2022) found that the introduction of STAR was associated with a 34% reduction in reports of minor criminal offences in the areas it served. The researchers attributed this to the displacement effect: when crisis response replaces enforcement, situations that would have been escalated into arrests (and therefore recorded as crimes) are instead resolved as health interventions. The "crime" does not disappear. It was never crime. It was distress, recategorised.
New York's Behavioral Health Emergency Assistance Response Division launched in 2021 in select precincts. Teams consist of an EMT, a paramedic, and a social worker from NYC Health + Hospitals. They respond to 911 mental health calls in participating areas.
In its first year, B-HEARD responded to over 6,000 calls. Of those calls, teams transported individuals to emergency departments only 15% of the time — compared to approximately 80% for police responses to similar calls. The remaining 85% were resolved in the field: de-escalation, connection to outpatient services, wellness checks, and referrals. Hospitalisation is not resolution. It is displacement. B-HEARD actually resolves.
The significance of the New York data is scale. Eugene is a city of 175,000. Denver is a city of 700,000. New York is a city of 8.3 million. If the model works in New York — and the early data says it does — the "it only works in small cities" objection is dead.
Oakland's Mobile Assistance Community Responders of Oakland program launched in 2021. Teams respond to non-violent, non-criminal 911 calls. In its first two years, MACRO responded to over 14,000 calls. Zero use of force incidents. Police backup requested in approximately 2% of calls — slightly higher than CAHOOTS, attributed to Oakland's higher baseline rate of armed individuals in public spaces.
The replication pattern is now clear across programs spanning different cities, different sizes, different demographics, different crime rates:
| Program | City | Population | Launched | Calls (annual) | Police Backup | Use of Force | Deaths |
|---|---|---|---|---|---|---|---|
| CAHOOTS | Eugene, OR | 175,000 | 1989 | ~24,000 |