New Orleans Implements AI Triage for 911 Calls to Manage High Call Volume

The Orleans Parish Communication District (OPCD) is putting an artificial intelligence system in place to help manage its heavy load of 911 calls. This AI, called Carbyne’s Emergency Call Triage system, will analyze and prioritize incoming calls, especially during surges linked to a single event. Its main job is to route callers and give them information. This lets human dispatchers focus on true emergencies.

Key takeaways

• New Orleans is using an AI system to sort 911 calls, managing a daily load that often passes a thousand calls.
• The AI automatically directs callers during a surge, asking if their call concerns a specific incident to offer updates or transfer them.
• The system will not manage emergency calls directly but will send them to human dispatchers after an initial check.
• Possible issues involve algorithmic bias and speech recognition problems with accents or dialects.
• This move comes after a similar AI system was used for the city’s 311 non-emergency call line.

Implementation and Function of the AI System

The Orleans Parish Communication District (OPCD) is introducing an AI system to answer some 911 calls in New Orleans. The district chose Carbyne’s AI Emergency Call Triage system, built to analyze and prioritize incoming emergency calls.

This system evaluates calls and gives immediate feedback. Its chief role is to manage spikes in calls about one incident. During such a surge, callers go straight to an AI agent. This agent asks if the call is about the specific event. If the answer is yes, the AI provides information or updates. If the answer is no, the call moves to a human dispatcher.

The OPCD says the AI will not directly handle emergency calls, only route them to a person. The tool should cut the number of calls human dispatchers take, a load that tops a thousand emergency calls daily. The system learns to spot repeated patterns, like floods of calls during rush hour accidents, to forecast and handle similar surges.

Purpose and Goals of the AI Integration

The main aim of using an AI system for 911 calls in New Orleans is to lower the volume of calls human dispatchers must field. This daily load exceeds a thousand emergency calls. The Orleans Parish Communication District (OPCD) is testing this new tool specifically to fight call pile-ups and shorten the time to answer potential emergencies during busy periods.

How the AI Triage Functions

The system, Carbyne’s AI Emergency Call Triage, evaluates incoming calls. It manages increases in calls tied to one incident by automatically sending callers to an AI agent. The agent asks if the call is about that event; if yes, the caller gets information, and if no, they transfer to a human dispatcher. The OPCD makes clear the AI will not handle emergency calls directly but will only send such calls to a human dispatcher.

This integration follows a prior use in April where AI answered 311 non-emergency calls. That system was launched because the OPCD states 50% of 311 calls ask for information.

Potential Risks and Reliability Concerns

While used to manage high call volume, depending on AI for 911 triage brings serious risks. The system’s success relies on proper monitoring, comprehensive training data, solid cybersecurity, and algorithm rules. Without these protections, results are not assured, stressing the vital need for constant human oversight to make sure the system performs as planned.

Bias and Speech Recognition Challenges

There is a chance the AI could develop unseen biases. If applied to predictive policing, analyzing when and where crimes might happen, results could be warped by historical crime data from over-policed neighborhoods. This could effectively strengthen systemic racism and socioeconomic bias.

Also, the AI’s automatic speech recognition systems might fail to understand people with strong accents, dialects, or differences in pitch and articulation. This makes the system possibly unreliable when faced with speech patterns different from those it learned to recognize.