Emma, a virtual receptionist equipped with artificial intelligence that operates in several family doctor clinics in South Yorkshire, has encountered an unexpected problem; the system is having difficulty understanding some strong local accents, and as a result, a number of patients have hung up and gone to clinics in person to receive services.
The Guardian has reported that Healthwatch Rotherham, which monitors the region’s health, care and social services, has received complaints from patients who said the system was unable to understand their questions because of their local accent.
Kim Gleeson, the director of the agency, referring to the variety of accents and speech patterns across South Yorkshire, said this diversity has posed a challenge for the system. According to her, accents in the area vary greatly from one another and the system cannot always correctly recognize them; an issue that is frustrating for patients.
According to the organisation, one patient also said they could not convince the system to understand what they meant and ultimately hung up and did not attempt to book an appointment again.
AI instead of long waiting queues
An increasing number of family doctor clinics in the UK are using Emma to answer patient calls immediately. The purpose of using this system is to reduce patients’ waiting time on the phone before connecting with reception staff.
However, Gleeson said the agency has also received complaints from elderly people and veterans about difficulty using the system and navigating its options. She said some patients, after finding the telephone system difficult or unpleasant to use, have been forced to attend their clinics in person to receive services.
She added that this issue can create greater concern for people who have less confidence using digital technologies or who live with disabilities.
Healthwatch has emphasised that family doctor clinics in the UK are legally required to make "reasonable adjustments" for patients who need assistance. The organisation has also urged patients who experience problems interacting with AI-powered reception staff to ask their clinic about alternative ways to access services.
Developer defends Emma
In response, QuantumLoopAI, the company that developed Emma, has defended the technology and stated that the system is designed to handle a wide range of accents and speech patterns.
The company said Emma, in addition to English, supports 17 other languages and does not make any medical decisions. It also said that if the system cannot understand or handle a patient’s request, the call will be transferred to a human reception team.
QuantumLoopAI also emphasised that callers can request to speak to a human staff member at any stage and are not obliged to continue the conversation with Emma.
According to The Guardian, the experience of using Emma in South Yorkshire illustrates one of the practical challenges AI systems face when moving from the laboratory to everyday use. A system’s ability to support multiple languages and dialects does not necessarily mean it can correctly understand all the variety of accents and speech patterns within a single language.