Creative Ways AI is Helping Aid Public Safety Operations

Software technology has come quite far indeed. I’ve been working in product management for over nine years now, with a bulk of that time being spent building platforms for software engineers and DevOps professionals. With all of that said, Even in my wildest dreams did I think that Artificial Intelligence would be this accessible to the common man before 2025, yet here we are. ChatGPT, closely followed by Google Bard and other AI vendors, is changing the technological landscape in front of our eyes.

While a lot has been written on how AI can improve the efficiency of various processes in the private sector, there’s more that could be discussed on how Natural Language Processing, Machine Learning, and AI could be utilized to drive positive change in public safety.

In this article, I want to think through the role of AI in helping 911, specifically in the Public Safety Answering Point (PSAP) realm. A PSAP is a call center where emergency calls are received and processed. When someone dials an emergency number, such as 911 in the United States, their call is routed to the nearest PSAP.

About 911

As of 2023, there are approximately 240 million 911 calls received each year in the United States alone. This does not include the millions of non-emergency calls received each year. These calls are processed by what is called a Public Safety Answering Point, or PSAP. The PSAP is where 911 call operators sit in, answer calls, emergency or non-emergency alike, and dispatch responses across law enforcement, fire, and EMS personnel.

Now let’s talk about some of the challenges faced by 911.

Staffing: One of the challenges faced by 911 call centers, or PSAPs, is staffing. With millions of emergency and non-emergency calls received each year in the United States alone, it can be difficult to ensure that there are enough trained operators available to answer every call in a timely manner. This problem can be exacerbated by factors such as high turnover rates and low wages for call center staff. According to a report by the National Emergency Number Association, in 2019, approximately 25% of 911 call centers had a staffing shortage. Furthermore, the report found that the turnover rate for 911 call center staff was around 20%. These staffing challenges can result in longer wait times for callers and a higher risk of burnout for call center staff.

Increasing calls: Over the years, there has been a noticeable increase in the number of emergency and non-emergency calls. According to recent statistics, the total volume of emergency calls has been rising steadily. In some regions, there has been an annual increase of approximately 3-5% in emergency calls. This upward trend indicates a growing dependence on emergency services and highlights the need for a robust system to handle emergencies effectively. Similarly, the number of non-emergency calls has also seen a significant surge. These calls typically pertain to community services, information queries, or non-urgent assistance. The exact increase varies by region, but on average, non-emergency calls have been growing by approximately 7-10%.

Mental health toll: Studies have shown that listening to tragedy can have a lasting impact. Dr. Michelle Lilly, a clinical psychology professor at Northern Illinois University, found that between 18% to 24% of 911 dispatchers show symptoms of post-traumatic stress disorder (PTSD). That’s on par with the rate of PTSD for police officers.

We’re going to explore just some of the interesting ways PSAPs are addressing some of these challenges with Natural Language Processing, Machine Learning, and AI.

AI for call routing

Handling non-emergencies

Non-emergency calls to 911 refer to situations in which immediate assistance from law enforcement, fire, or medical personnel is not required. These calls are made to request information, report minor incidents, or seek assistance with non-life-threatening issues.

Examples of non-emergency situations that people might call 911 for include noise complaints, minor traffic accidents without injuries, reports of non-dangerous disturbances, vandalism, or lost property.

While these situations are not urgent or life-threatening, they still require attention from appropriate authorities, albeit not at the same level of urgency as emergencies like crimes in progress, medical emergencies, fires, or situations with an immediate threat to safety. Estimates say that about 40% of calls to 911 are non-emergencies.

The challenge is in a severely understaffed PSAP, non-emergencies could come at the cost of handling more critical calls. Non-emergency calls can sometimes flood the call center during peak hours, especially in cases where there's a community event, local news coverage, or public interest. This increased call volume can strain the resources available for addressing emergency calls promptly. Call-takers need to switch their focus between non-emergency and emergency calls, potentially causing them to lose concentration or context when handling high-stress 911 calls. This can result in errors or inefficiencies when handling emergencies.

Technology can help here. One example of how AI could contribute to a solution is a virtual assistant. Imagine if every non-emergency call is picked up by an AI assistant, and with a few investigative questions for triaging, could both route the call to the right representative (or guide the caller to an online reporting portal in case a human is not required to assist them). With its ability to analyze and categorize calls, the AI system could accurately direct callers to the appropriate departments or resources, ensuring a quicker response to inquiries related to non-urgent matters.

This is exactly what Portland is doing by debuting an automated AI system. By automating the initial call-handling process for non-emergency calls, law enforcement agencies in Portland hope to optimize resource allocation and response times. The AI's capacity to swiftly process information and accurately identify the nature of the calls will contribute to more efficient utilization of personnel and departmental resources. This innovative move has generated both curiosity and optimism within the community. While concerns regarding potential challenges in understanding nuanced human speech patterns persist, authorities have emphasized that the AI system has undergone rigorous testing to ensure its effectiveness in comprehending a wide array of caller inputs. Another example of a county doing something similar is Charleston County in South Carolina.

Routing medical calls

Handling 911 medical calls presents a complex set of challenges for emergency response systems. The urgency of these calls demands swift and accurate assessment of medical conditions, which can be complicated by the caller's emotional state, limited medical knowledge, or the unpredictability of the situation. Call-takers must navigate through varying levels of urgency, from life-threatening emergencies to less critical situations, all while providing clear instructions and reassurance to often distressed callers.

The Support Team Assisted Response (STAR) Program in Colorado and other states in the US deploys Emergency Response Teams that include Emergency Medical Technicians and Behavioral Health Clinicians to engage individuals experiencing distress related to mental health issues, poverty, homelessness, and substance misuse. STAR responds to low-risk calls where individuals are not at imminent risk. There are many places that offer star programs in Denver, Colorado, including addiction treatment centers, mental health services, and non-profit organizations.

911 could use AI similar to how Denver, CO, is utilizing the technology to analyze calls for keywords and patterns and then flag them for automatic routing to the STAR program.

That could be used to help a dispatcher's ability to better identify an appropriate response since it's up to emergency dispatchers to determine whether to send police, fire, EMS, or STAR to calls. It could also help supervisors and administrators determine which calls could be STAR-eligible that aren’t being tagged as one to improve PSAP policy.

AI for language translation

If someone calls 911 and doesn't speak English, they will be transferred to a language line service where an interpreter will translate all questions and answers. The caller will need to tell the dispatcher what language they speak, and the call taker will conference in an interpreter to translate the conversation. The interpreter will be an over-the-phone (OPI) interpreter who is specially trained in dealing with emergency calls.

Addressing non-English calls within the context of 911 services poses unique challenges to effective emergency response. Language barriers can impede clear communication between callers and call-takers, potentially leading to misunderstandings about the nature of the emergency, the location, and the required assistance. This can result in delays in dispatching appropriate resources and potentially compromising the safety of individuals in need. Moreover, the absence of readily available interpreters or translation services can further exacerbate the problem. Ensuring accurate and timely assistance for non-English callers requires specialized language support, culturally sensitive communication strategies, and the integration of technology to bridge language gaps and enhance the overall effectiveness of emergency response systems.

AI-powered language translation utilizes advanced machine learning algorithms to automatically convert text or speech from one language into another, enabling effective cross-lingual communication. These systems leverage large datasets and neural networks to understand linguistic nuances, idiomatic expressions, and context, resulting in more accurate and contextually relevant translations. They continually improve over time by learning from vast amounts of multilingual content available on the internet. This technology could be leveraged in 911, and cities like Baltimore, MD, are leading the charge on this front.

AI to assist in note-taking

When hiring new 911 dispatchers and call-takers, a big emphasis is placed on the potential recruit’s ability to type both quickly and accurately. This is so they can distill all key insights said by a caller during an emergency as quickly as possible for first responders to take action on.

In a world where 911 is facing a staffing issue, expanding the talent pool could be a key way to attract more heads into the world of public safety.

AI can significantly reduce the emphasis on dispatchers taking extensive notes during a 911 call by automating the process of extracting critical information from the conversation through:

1. Automated Speech Recognition (ASR): AI-powered ASR technology can transcribe spoken words into text in real time. By accurately converting the caller's words into text, dispatchers can focus more on actively engaging with the caller and providing necessary guidance rather than struggling to jot down every detail.

2. Contextual Insights: Advanced AI systems can provide dispatchers with contextual insights by analyzing the call history of the caller or location. This enables dispatchers to make informed decisions based on past incidents, enhancing the quality of their responses. Companies like Prepared are pioneering technology in this field for public safety.

3. Assistance in Information Gathering: AI can prompt dispatchers with relevant questions based on the information gathered, ensuring that no critical details are missed during the call. This guidance streamlines the conversation and reduces the need for extensive note-taking.

   
   

By aiding in the process of capturing, transcribing, and extracting key information from 911 calls, AI technology allows dispatchers to focus on the human aspect of the interaction, providing empathy, reassurance, and vital instructions to callers in need while simultaneously allowing PSAPs to expand their talent pool of candidates.

AI for tagging mental-health-related calls

Mental health is a huge challenge in public safety, with public safety officials being exposed to some of the worst in humanity.

A 911 dispatcher has to take calls back to back. In large centers, particularly where their call volume is in the hundreds of thousands in a given year, a 911 telecommunicator could spend 30 minutes on a call talking to a suicidal person and follow that up immediately by answering another call without breaks.

Agencies like NCT911, which is responsible for supporting more than 40 emergency communications centers across 14 counties in the Dallas-Fort Worth area of Texas, are using services from AWS to pioneer such a system. The implementation of AI software that flags distressing calls is seen as a game changer in retaining employees and ensuring their well-being. If a telecommunicator has handled a certain amount of mentally taxing calls based on keywords and sentiments analyzed, the supervisors are notified to help provide backup and support to the telecommunicator. The program will monitor calls for keywords and the tone of the caller and keep tabs on whether a call was especially stressful.

Closing Notes

In conclusion, the evolving landscape of AI technology is transforming the field of emergency response and public safety in remarkable ways. From addressing mental health challenges among 911 operators to handling non-emergency calls more efficiently, AI's potential to enhance emergency services is substantial. The intersection of AI and emergency response holds the promise of safer, more efficient, and compassionate public safety services, ushering in a new era of technological advancements that benefit communities on a global scale.