Dive Brief:
- DoorDash is adding an artificial intelligence tool, called SafeChat+, to its chat program to detect potentially abusive or harassing speech in app-based communication between customers and workers, the delivery giant announced Tuesday.
- The press release offers little clarity on the technical processes or kind of analysis SafeChat+ performs to review messages.
- SafeChat+ supersedes the original SafeChat, which DoorDash launched in 2022 and was based on a less powerful technology relying primarily on keyword matching. The new tool can detect subtle nuance and problematic content within messages without relying on specific keyword matching, meaning it can flag a wider range of potentially hostile communications, DoorDash wrote in an email to Restaurant Dive.
Dive Insight:
SafeChat+ will review upwards of 1,400 messages a minute to keep pace with the overall volume of messages sent through DoorDash’s chat system, but human employees are still required to oversee the tool’s findings.
According to the press release, DoorDash’s Trust & Safety team will review all messages flagged by SafeChat+ “and take appropriate actions to enforce our policies, which strictly prohibit any verbal abuse or harassment.” The company didn’t respond to a request for comment about how many workers will be tasked with reviewing all the conversations flagged by SafeChat+.
When DoorDash’s tool flags a conversation for potential verbal abuse between a customer and a worker, the worker “will be given the option to quickly cancel the order without impacting ratings. If the order is already completed, the feature will automatically end any further chat to help prevent the situation from escalating.” When a worker uses inappropriate language, the customer “can reach out to support via chat or phone to report the incident and receive assistance.”
The tool itself does not review any consumer information, only the text included in a chat. DoorDash didn’t respond to a request for comment as to how SafeChat+ determines a message is harassing or abusive.
According to the press release, SafeChat+ can analyze dozens of languages, including English, Mandarin, French and Spanish.
DoorDash emphasized that while the vast majority of its deliveries occur without any safety issues, verbal harassment is the most common safety problem faced by its workers. SafeChat+ follows a number of other safety features DoorDash has added to improve driver safety, after unsafe working conditions helped spur worker organizing in New York City. In 2021, DoorDash added SafeDash, a safety tool developed in concert with ADT, which gave workers an emergency assistance button. The company added porch light reminders, safety check-ins, and and real-time safety alerts in 2022 around the time it deployed an earlier version of SafeChat.
