AI system detects loneliness in natural speech patterns

A new proof of concept study, led by researchers at University of California San Diego School of Medicine, has demonstrated how speech-analyzing artificial intelligence tools can effectively predict the level of loneliness in older adults.

Natural language processing (NLP) is an umbrella term encompassing a variety of techniques that process or analyze large volumes of unstructured natural speech and text. As artificial intelligence and machine learning systems have advanced, a number of fascinating preliminary studies have begun to suggest conditions such as psychosis, PTSD, bipolar disorder and depression may all be detected just by analyzing a person’s natural speech.

Now, a team of researchers is investigating whether these NLP tools can detect loneliness, a growing health concern that has been described as a bigger factor on premature mortality than obesity. Ellen Lee, senior author on the new research, suggests loneliness is a particularly difficult psychiatric condition to measure and because doctors generally struggle to quantify loneliness in patients there is a pressing need for some kind of objective measure.

"Most studies use either a direct question of 'how often do you feel lonely,' which can lead to biased responses due to stigma associated with loneliness, or the UCLA Loneliness Scale, which does not explicitly use the word 'lonely,'" explains Lee. "For this project, we used natural language processing or NLP, an unbiased quantitative assessment of expressed emotion and sentiment, in concert with the usual loneliness measurement tools."

The new study recruited 80 older adults. Each subject was evaluated using conventional loneliness assessments as well as completing a longer, more conversational, semi-structured interview lasting up to 90 minutes.

The interviews were transcribed and then analyzed with the help of a natural language system developed by IBM. As well as detecting loneliness in subjects not picked up by conventional assessments, the system uncovered differences in the way men and women talk about loneliness.

"NLP and machine learning allow us to systematically examine long interviews from many individuals and explore how subtle speech features like emotions may indicate loneliness,” says first author Varsha Badal. “Similar emotion analyses by humans would be open to bias, lack consistency, and require extensive training to standardize.”

The AI system reportedly could qualitatively predict a subject’s loneliness with 94 percent accuracy. The more lonely a person was feeling, the longer their responses were to direct questions regarding loneliness. The researchers even suggest the presence of a kind of “lonely speech” pattern could be used in the future to monitor the well-being of older subjects.

The men in the study were found to use more fearful and joyful words in their longer conversational interviews, while the women studied were more likely to explicitly vocalize feelings of loneliness. Even without incorporating these kinds of NLP tools in current practice, the researchers suggest the study’s findings offer clinicians important insights into the different ways men and women express loneliness.

The next stage of the research will be to combine other sensor data into the assessments (such as GPS tracking and sleep data) to personalize each individual finding. Plus, the system will need to be tested on larger, more diverse, populations to fine-tune its accuracy.

“Eventually, complex AI systems could intervene in real-time to help individuals to reduce their loneliness by adopting in positive cognitions, managing social anxiety, and engaging in meaningful social activities,” the researchers boldly conclude in the new study.

The study was published in The American Journal of Geriatric Psychiatry.

Source: UC San Diego Health