Getty Images
- Researchers used AI to detect whether someone had a cold based on the tone of their voice.
- The findings are intended to diagnose illness over the phone to minimize medical appointments.
- The technology could also make it harder for people to call in sick when they're not ill.
While fears rise about the impact of AI on work, the technology may be able to help stop employees calling in sick when they're actually fine.
Researchers at the Sardar Vallabhbhai National Institute of Technology in Surat, India analyzed the vocal patterns of 630 German people — 111 of whom had a cold — in an attempt to identify those who really were ill.
The study, as reported by The Economist, made use of the vocal rhythms detected in human speech, which can be characterized as harmonics that diminish in amplitude as their frequency increases.
The researchers, led by electronic engineer Pankaj Warule, hypothesized that a cold could upset this regular pattern.
They then placed these components into machine-learning algorithms to identify and studied the differences in amplification to determine those who had a cold.
According to The Economist, the subjects were asked to count from one to 40 before describing what they did at the weekend. They were then told to recite Aesop's fable "The North Wind and the Sun," a text that has previously proved popular for phonetics researchers.
The study was able to detect cold correctly about 70% of the time.
The primary application of the findings, according to the researchers, was to be able to identify a cold without the need for costly and time-consuming visits to a doctor.
However, it might also interest employers keen to catch out staff pretending to be sick. However, that could involve asking a staff member to read a fable.
While there isn't much rigorous research on faking sickness to avoid work, Americans don't take many days off even if they are ill. Nearly a quarter of private sector employees don't get paid sick leave, unlike many other countries where paid sick leave is mandated by the government.
The researchers' development in speech recognition is one of the latest in AI's continued application in the medical field.
Last year, a research group was able to use AI to detect Parkinson's through the breathing patterns of patients. Depression and several forms of cancer may also be detected with the use of AI-trained analysis of vocal patterns.