SendHelp

SendHelp is an AI messaging platform that directs residents to non-police first responders in non-emergency situations, such as homelessness and mental health episodes. By reallocating police resources for "quality of life" violations, we reduce at-risk citizens' exposure to police and consequently, police violence.

A Berkeley shop owner hears some commotion outside, and there is someone yelling random things and causing a disturbance at the front of their store. The owner does not want to call the police and bring them into the community, but they do need help so people still feel comfortable entering the store. On SendHelp, the shop owner declares a non-emergency, selects mental health crisis in Berkeley, and deems the situation potentially dangerous. SendHelp directly links the owner to the relevant resources and organization to call.

Beginnings

With the resurgence of the Black Lives Movement in the summer of 2020, I began to think more deeply about what I could do besides going to protests, signing petitions, and educating myself. I wanted to leverage the skills and resources I had as a Berkeley student to do more.

Over the course of a year (May 2020 - June 2021), I worked closely with my colleague Marcel Tan to find a way to address police brutality and reduce police incidents. We conducted research to understand the institutional and systemic racism in several interacting government systems (public welfare, criminal justice, educational, police); interviewed stakeholders and actors within the Berkeley city and campus communities; figured out what specific root causes and needs we could address and how; and implemented a solution under the guidance of UC Berkeley's Big Ideas Program.

Big Ideas

Pitch Video for SendHelp (prev Not the Police)

While we conducted the majority of the research on our own, in Fall 2020 were given the opportunity to compete in the Big Ideas Contest, which focuses on student-led innovation on a social impact track. With our pre-proposal, we were selected as finalists and paired with an amazing mentor to bring our project to life. She provided indispensable guidance as we collaborated with government officials, first responders, and other important actors in local public services. Our final proposal became an award winner ($10,000!) and we became grand prize finalists. Soon after, with graduation and several other factors our work came to a close, but the chatbot is still live and ready to SendHelp on Facebook Messenger. Feel free to try it out below!

SendHelp Bot

Final Proposal

Process

1. The Big Picture

Police violence is intimately connected with the criminal justice system, mass incarceration, systemic racism, education, and more; and there is a long history in the policies, systems, and laws that perpetuate police violence. We needed to educate ourselves and prepare ourselves with as much context as possible before diving into problem solving.

4. Creating the Solution

There are police alternatives, but people don't know about them, so how do we enable them to call the right services to make all residents feel secure? How do we make it as easy as dialing 911? We complied a database of police alternatives in Berkeley, SF, & Oakland and created a chatbot to guide people to the appropriate service for a non-violent situation, and reduce unnecessary police exposure.

2. In-Depth Research

We dug into reports, scanned crime statistics, listened to dozens of podcasts, and interviewed 4 UC Berkeley professors to gain a better understand of the scope and context of the problem. We mapped out how several governmental and public systems interacted and affected each other to see what levers we could pull to create the most impact.

5. Validation & Testing

Besides continually testing on our peers, we were already in conversation with UC Berkeley's homeless outreach coordinator, the City of Berkeley's mobile crisis team, and the UCPD; we kept them updated with our progress and received good feedback as we continued to present our work to UC Berkeley's mental health response team and City of Berkeley council member Rigel Robinson.

3. Defining the Problem

Police violence has many "why's" - implicit bias, arrest quotas, lack of public welfare support for low income and Black communities, just to name a few. How might we reduce/eliminate police violence without revamping the entire interconnected system? The answer ended up being simple: reduce' exposure to the police, and police violence decreases.

6. Implementation & Impact

Though SendHelp was not fully implemented due to personal factors, we planned to continue user testing, partner with campus mental health, run a digital marketing campaign, and run a pilot with the City of Berkeley. We'd measure our impact via # of subscribers, # of calls routed from the app, and # of unhoused who feel safer via a survey performed before and after the release.

Problem Statement

Police killings of Black Americans and at-risk citizens are often linked to an initial 911 call reporting a non-violent incident. These tragic cases include allegedly using counterfeit notes (George Floyd), sleeping intoxicated in a parked car (Rayshard Brooks), or suffering from a mental breakdown (Tanisha Anderson). Their deaths were preventable.

One impactful lever to reduce the amount of police violence is reducing people’s exposure to the police - NYPD’s stricter oversight of Stop & Frisk in 2013 led to a rapid decrease in police stops, leading to 37 fewer police stops with use of force toward Black Americans daily.

Non police services exist, but our interviews with homeless outreach and mental health crisis workers in Berkeley revealed that residents and merchants do not know who to call in a non-emergency and resort to calling 911 for help. When police respond in non-violent situations, unhoused individuals feel unsafe and Black Americans face a higher risk of experiencing police violence.

SendHelp directs residents experiencing non-emergencies to non-police first responders. Our mission is to make non-emergency calls as easy and intuitive as calling 911.

Discovery

There's no doubt that police violence is an ongoing problem in the United States and across the world. Breaking down this complex issue was the most difficult and daunting part of starting SendHelp. We needed to gain context on the overarching problems and institutions that feed into police brutality. The policing methods and relationships within low-income communities (especially Black communities) is strained because of underlying systemic and racialized problems such as poverty, discrimination, and criminalization - so where can we provide effective support with technology?

Research Plan

Interviews

Goals (varies for each interviewee and their field/expertise):

Understand how policy and reform exacerbates or reduces police violence
Understand linkages between crime and police violence
- Implicit bias in police because of communities’ self-policing
- When did the cycle start?
Figure out what’s in our control or out of our control
Understand how police interact with their community, if there are any problems, what their relationship is
Understand police's role, the most common issues they encounter
Understand the interacting players in the community ecosystem and how to move towards defunding, what the police are doing that they should not be doing

Interviewees

Professors specializing in criminal justice and interpersonal/state violence
- Jonathan Simon - Criminal Justice Law Professor at the UC Berkeley School of Law
- Gabriel Lenz - Political Science Professor at UC Berkeley
- Laurel Eckhouse - quantitative social scientist focused on racial inequality and the criminal legal system at Code for America, former assistant professor of political science at the University of Denver
3 UCPD officers in Community Engagement Unit
Ari Neulight, Homeless Outreach Coordinator, UC Berkeley
Jeff Buell, Program Supervisor, City of Berkeley’s Mental Health Crisis Services
Databases manager for SF public services

Crime Statistics and Data

Research Papers and Podcasts

Problem Definition

Potential Goals

figure out an effective way to complement new community measures (such as defunding the police) whether that be digital, physical, or organizational
reduce police incidents involving excessive use of force
higher trust between lower-income communities and police departments
direct mental health crises to specialized professionals rather than police officers

Understanding

We charted out the expansive systemic issues and patterns across our research and interviews to try to see which levers we could pull and where we would be able contribute with our current resources and skillsets. We were trying to pinpoint deeper causes, asking why certain issues prevailed. We are residents of the Berkeley community so we could tackle this locally and the touch points were the police force, community organizations, government services, and us, people who could call the police.

VIEW FULL BOARD

Prioritizing

Continuing on the miro board, we wrote down the most common problem threads that were coming up in our research and placed them on a 2x2, with the X axis representing the severity of the problem and the Y axis representing the feasibility for us to address it. The problem we investigated further was "mental health, homelessness, other social problems not addressed by specialized first-responders."

Further Research + Refined Problem Statement

After we narrowed down a specific problem space, we complied more specific stats, touched base with people we interviewed previously, and investigated existing solutions in the space. We also surveyed Berkeley students about their awareness of local non-police alternatives.

Key Findings

1. Black Americans are 2.3 times as likely as white Americans to be killed by police today, partly due to disparate exposure to the police. We analyzed all 1.05 million LAPD police dispatches in 2019 and found that 66.9% of them concerned non-violent or non-criminal incidents.

2. Many police brutality cases have stemmed from 911 calls reporting such non-violent incidents, including allegedly using counterfeit notes (#GeorgeFloyd), sleeping intoxicated in a parked car (#RayshardBrooks), and suffering from a mental breakdown (#MichelleCusseaux). These police killings could’ve been avoided had the complainants been aware of a non-police avenue to resolve the incident.

3. 1 in 4 Berkeley Police Department (BPD) calls for service in a 6 months span in 2020 involved “quality of life” violations, such as vandalism, disorderly conduct, and liquor law violations.

4. Ari Neulight, Cal’s Homeless Outreach Coordinator, Jeff Buell, Supervisor of City of Berkeley’s Mental Health Crisis services, mentioned that Berkeley residents and merchants often call the police or an inappropriate agency for “quality of life” violations.

5. Unhoused residents in Berkeley feel unsafe when the police respond. Berkeley stakeholders have formed work groups to deal with this problem but they are admittedly slow-moving.

6. NYPD’s stricter oversight of Stop & Frisk in 2013 led to a rapid decrease in police stops, leading to 37 fewer police stops with use of force toward Black Americans daily.

7. 73% of young people aged 18 to 25 are not familiar with the 311 non-emergency helpline/app.

8. 72% of Californians support the reallocation of police responsibilities and funding for responding to non-violent situations to social and mental health services.

9. Non police services (mental health responders, homeless outreach coordinators) exist, but residents are unaware of who to call for non-violent situations and when.

Reduce exposure to the police and reduce police violence.

Potential Solutions

After our research, we determined we could approach reducing police violence and police exposure either from the community side or the police institution side. Policy change would work and was in motion, but we needed something more feasible that could be implemented sooner. How might we get the right services to be called to make all residents feel secure?

Landscape Analysis

Grassroots movements and city legislature.

Critical Resistance and the Anti-Police Terror Project (APTP).
Oakland public safety task force, headed by APTP, recently recommended that the City Council transfer the 911 call center out of the police department, and invest in gender-based violence prevention and violence interrupters (community members who deescalate situations before they become violent).
APTP launched Mental Health (MH) First Oakland, a non-police alternative that dispatches volunteers to respond to psychiatric emergencies, substance abuse, and domestic violence.
Berkeley City Council has passed BerkDOT to reduce police violence, shifting traffic enforcement responsibilities from BPD to a newly-created Department of Transportation.
Oakland Power Project is a grassroot organization that works with residents to create policing alternatives. They’ve created workshops to educate people on non-police options to crises

How we can build on it

We want to approach police violence from a similar angle, namely by reducing the number of encounters that residents have with the police. While BerkDOT has reallocated funding to non-police enforcement for traffic stops, we want to also fill the gap for cases involving mental health crises, the unhoused population, and other non-violent situations. Although the City of Berkeley and organizations such as APTP provide non-police alternatives, the public is largely unaware of these services due to their low digital presence. We’ll build on their amazing work by creating a scalable, community-centric solution.

Website aggregators.

Dontcallthepolice.com consolidates contact information for non-police alternatives. Users can view a list of organizations and their phone numbers, filtered by location.
Independently look up the phone number to call on their own.

How we can build on it

Sifting through information on a website aggregator to determine which organization best fits the situation, or looking for contact info on a service's website takes time. In a non-emergency, the time spent navigating these websites to find the right phone number could allow a situation to escalate. Additionally, city services occasionally change their staff and contact information – as the City of Berkeley did recently – so online sources would not reflect these changes in a timely manner.

311 helpline/app.

311 helplines/mobile apps enable residents to report non-emergencies to a municipal agency, serving as a non-police alternative.

How we can build on it

311 responses are limited to government services; they do not utilize the resources of influential grassroots/nonprofit organizations. Furthermore, a majority of younger residents have not heard of 311 services. Our survey of 90 respondents (18 to 25 years old) revealed that 73% of people were not familiar with the 311 concept. 311 apps also face high friction in user adoption because residents have to install and learn to use a separate app. We can an easier way for these resources to be integrated into people's daily lives.

Proposed Solution

It’s easy to scrape contact numbers and put them on a website. That’s why our biggest external challenge is making the process of reaching alternative first responders so intuitive that it can be done in one step. We must overcome residents’ ingrained habit of calling 911 for non-violent “disturbances”. How do we make sure people are contacting the right person and how do we direct people to the right phone number for their situation?

SendHelp is an AI messaging platform that enables residents to easily call non-police first responders (e.g., MH First Oakland, Berkeley’s Mobile Crisis Team) for non-emergencies. It analyzes users’ messages to provide fast, accurate responses to questions and requests for service.

By redirecting “quality of life” calls to alternative first responders, Not the Police lowers the exposure that at-risk residents have to the police. Decreased police exposure leads to lower levels of police brutality, saving lives.

How it works

Initial Contact

With the Facebook Messenger app/website, users will tell the chatbot (via text or guided buttons) that they are experiencing a non-emergency.

Guide to Service

The user will answer a short list of questions about the situation and the platform takes in the inputs and provides the user with a list of suitable first responder services, along with their respective hours of operation

Call the correct service

Once decided on a service, the user can call the chosen first responder service at the tap of the “Call Now” button – all without leaving the chat window on the Messenger app.

Value Proposition

More efficient than existing methods. Currently, finding the right non-emergency service is difficult given the plethora of service websites and lack of a central source of truth. The platform increases the “fit” between a caller’s situation and the service picking up the call, saving time for both residents and city services. Efficient dispatch is valuable because speed is of the essence even in responding to non-emergencies.

SendHelp serves a pre-dispatch role of educating residents about alternative first responders and streamlining the process of getting help. Jeff Buell, Mental Health Program Supervisor for the City of Berkeley, informed us in our discovery call that our product would be a useful “marketing tool” for city services.

Highly accessible, even in its prototype phase, because it is built on top of Messenger, a free and popular instant messaging platform. Additionally, our ideal first user is the Gen Z/Millennial social media user who volunteers for and seeks out activist causes. People in this segment not only resist bringing police presence into their communities, but also prefer using tech-enabled services like chatbots to meet their everyday needs.

Integration with city services. Our team has been in contact with university, nonprofit, and city stakeholders in the Bay Area (such as Berkeley City Councilmember Rigel Robinson and Berkeley Fire Department Assistant Chief David Sprague) from the get-go. Partnering with these stakeholders empowers us to consistently provide accurate, up-to-date information that is lacking in online sources such as crowdsourced spreadsheets/aggregators and antiquated government websites. We have the opportunity to share data two-way with our partner services to optimize our platform’s online-to-offline impact.

Benefits police departments by freeing up their time spent on non-emergencies to focus on handling serious, life-threatening situations. Police officers spent a significant amount of time managing minor, non-violent incidents. We interviewed three UCPD officers – Lt. Sabrina Reich, Sgt. James Jenkins, and Cpl. DiMarco Hoskins. They support educating UC students on non-police alternatives and reallocating police resources for cases involving homelessness and mental health.

Risks

Guiding people to an inappropriate service or to a non-police alternative when the situation calls for the police. To overcome this, our team has called and emailed services in our database to vet them and verify their information. We partner with organizations that we list, so that they are still operating and everything is up to date. Moreover, our platform has a built-in safety net to ensure that residents experiencing serious emergencies are immediately directed to 911. Our chatbot recognizes keywords such as “attack”, “bleed”, “chase”, etc., and auto-responds with a message that includes a button for the user to call 911 immediately.

Overstepping. I am not a target of police violence, and I need to be aware of my privilege in this space, always actively listen, and continue to amplify the voices I am really building this platform for. I need make sure people I'm working with are comfortable with these topics, and that sure they can speak freely and openly - one can never speak on behalf of other people when you don't have their lived experience. We've interviewed not only academics and police officers, but more importantly social workers and people who work directly with those affected by police violence. It's also important that we partner with the grassroots organizations in Berkeley/Oakland and have their support too.

Learnings

Partnerships

While SendHelp achieves tremendous potential for social impact, there are few ways to monetize in the social/public services space - we cannot pollute the bot with ads and detract from the flow to get people to the right service. As a result, partnerships with other organizations and municipalities is the best way to get funding. We were fortunate enough to start conversations with several UC Berkeley campus groups and our local city council member to start government digital transformation in addressing police violence.

Dissecting Social Issues

The hardest part of creating SendHelp was the very beginning. Understanding police violence in the context of the institutional and system racism throughout the government and community was heavy, complicated, and a lot all at once. People dedicate their careers to specific parts of that, and we understood just enough to move forward with SendHelp. The key was finding a very specific touchpoint that could be tackled, and continuing to narrow down solutions from there. In our case, once we decided to focus on reducing exposure so we could think more clearly about how police exposure comes about.