In an experiment, the four leading virtual digital assistants (VDAs) were queried aloud about first aid for a range of health situations. Even when the virtual assistant understood the question, the answers were often off the mark, researchers report in BMJ Innovations.
The technology is promising and improvements are being made daily, said coauthor Matthew Douma of the department of critical care medicine at the University of Alberta in Edmonton, Canada.
“The greatest potential would be for an elderly person who fell and is on the floor,” Douma said. “If they can speak out loud they could get help.”
Unfortunately, Douma said, the VDAs provided lifesaving information only about half of the time. Worse, Apple’s Siri and Microsoft’s Cortana often were unable to parse the words that were spoken to them.
When contacted for comment, Google, Microsoft and Amazon responded, noting that they strove to provide the best information possible. Microsoft and Amazon suggested the new study might spark improvements in their VDAs.
“The safety of our users is extremely important to us and we will evaluate the study and its findings and continue to inform our products from a number of valuable sources, Courtney Gehring, a spokesperson for Microsoft, said in an email.
“We’re always working to make Alexa more helpful for customers,” Shelby Delano a spokesperson for Amazon said in an email. “The ways customers want to use Alexa continue to evolve and we’ll continue to take customer feedback into account for our products and services.”
Google took a different tack. “Google Assistant was not designed for medical emergencies and we encourage people to use traditional emergency response channels,” Christina Peck, a spokesperson for Google said in an email.
Apple did not respond to a request for comments.
In the study, while Alexa and Google Assistant understood more than 90% of the queries, their advice, when compared to recommendations from the Canadian Red Cross Comprehensive Guide for First Aid, often fell short. Google Assistant’s advice agreed with the guide 56% of the time, while Alexa was on the money just 19% of the time.
Douma gives a striking example. “One trigger we used was ‘Google, I can’t breathe,’” he said. “And it would play the Faith Hill song ‘Breathe.’”
Siri and Microsoft’s Cortana fared worse, correctly parsing the queries just 17% and 5% of the time respectively.
One bright spot: All the VDAs recommended calling emergency services if the user said “I’m having a heart attack.”
Douma and his colleagues put the four VDAs to the test in March 2018. They queried the virtual assistants on 39 first aid topics ranging from nausea and vomiting to penetrating chest trauma.
The VDAs were prompted by remarks such as: “How do I know if someone is having a heart attack?” “What do I do for someone who is having a heart attack?” and “I’m having a heart attack.”
While the virtual assistants may have improved somewhat since the study was done in 2018, Douma ran an impromptu test of one of them for Reuters Health, saying “Alexa, I can’t breathe.” Alexa’s response: “Take deep breaths.”
The correct response would have been to alert the user to the fact that this could be an emergency situation and suggesting a call to 911 for help, Douma said.
Dr. Leonard Weiss agrees that VDAs have a lot of potential for helping people in a medical emergency and he hopes that feedback like the current study may spur companies to improve their products.
While it’s good that the VDAs instructed people with heart attack symptoms to call 911, it would be better if they also offered advice on what to do while waiting for the ambulance, Weiss said.
SOURCE: bit.ly/3aStc5D BMJ Innovations, online January 7, 2020.
https://www.reuters.com/article/us-health-emergencies-digital-assistants/if-you-feel-you-cant-breathe-dont-expect-virtual-assistants-to-call-for-help-idUSKBN1ZT2XW
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.