Voice Assistants Imperfect at Responding to Cancer Screening Questions: Report
Voice assistants could be better at answering questions about cancer screening, according to a new study published in this month’s Annals of Family Medicine. The researchers compared Amazon Alexa, Apple Siri, Google Assistant, and Microsoft Cortana and measured how well and accurately the voice assistants responded, with all of them, especially Alexa, struggling to be verbally helpful.
Stanford University researchers put together the study titled “Voice Assistants and Cancer Screening: A Comparison of Alexa, Siri, Google Assistant, and Cortana.” Five scientists tested the voice assistants twice from different devices by asking the question, “should I get screened for [cancer type] cancer?” The researchers asked about 11 different kinds of cancer and evaluated how well the voice assistants understood the questions and how they responded in terms of accuracy and usefulness. The answers were marked based on the cancer screening guidelines of the U.S. Preventive Services Task Force (USPSTF). The response was marked as accurate if it helped start a screening process similar to the one offered by the USPSTF and as long as the information was not contradicted by the information offered by the task force.
As can be seen in the chart up top, the responses were far from consistent across the four different voice assistants. While Apple, Google, and Microsoft’s respective voice assistants all scored perfectly on understanding what was being asked, the study claims that Amazon’s Alexa couldn’t understand a single one of the questions about cancer. And while Google Assistant and Cortana were able to return helpful website ideas and verbal responses, Alexa was joined by Siri in not offering anything verbally, though Siri was very good at picking out helpful websites at least. The top three links that Siri, Google Assistant and Cortana recommended were usually pretty good at hitting the task force’s metrics, hitting about 70% in providing information in line with the official group. Even for Google and Microsoft, the web searches were better than the verbal responses.
“In terms of responding to questions about cancer screening, there are clear differences among the 4 most popular voice assistants, and there is room for improvement across all assistants,” the researchers explain in the paper. “Almost unanimously, their verbal responses to queries were either unavailable or less accurate than their web searches. This could have implications for users who are sight-impaired, less techsavvy [sic], or have low health literacy as it requires them to navigate various web pages and parse through potentially conflicting health information.”