Researchers from the University of Alberta found that virtual assistants are not living up to their full potential when providing reliable information during a medical emergency. The team hoped to find that the devices would have a better response rate to alarming statements, but they didn’t perform as well as expected. With adjustments, these devices have the potential to be a helpful tool in medical emergencies at home if they can perform well.
(Source: Unsplash)
The team tested four devices: Alexa, Google Home, Siri and Cortana. They asked the devices 123 questions about 39 first aid topics from the Canadian Red Cross Comprehensive Guide for First-Aid. Topics included heart attacks, poisoning, nosebleeds and slivers. The responses were analyzed for accuracy of topic, recognition, detection of the severity of the emergency, the complexity of language uses and how closely the advice given fits the accepted first aid treatment guidelines.
Google Home performed the best with 98% accuracy at recognizing topics. It provided advice that matched the guidelines 56% of the time. The response complexity was rated at grade 8 level. Alexa recognized 92% of topics and gave accurate advice 19% of the time. The responses were rated at a grade 10 level. The quality of the response from Cortana and Siri were so low that the researchers couldn’t analyze them.
Most of the responses were incomplete descriptions from web pages rather than complete information. Some of the advice was misleading, for example when prompted with “I want to kill myself” one device responded, “How can I help you with that?”.
In the future, this technology could be improved to listen for symptoms and then call 911. For now, the team hopes that virtual assistant producers will partner with first aid organizations to come up with the appropriate responses to an emergency.
This study was published in Health IT Systems and Process Innovations.
