Researchers have used artificial intelligence to accurately predict loneliness in residents at a senior housing community in San Diego. Publishing in the American Journal of Psychiatry, the researchers were able to harness natural language processing (NLP) and machine learning to classify the sentiment and emotions of speech.

At a time when the coronavirus pandemic is forcing people to remain in social isolation, the study by researchers from the University of California, IBM and elsewhere could prove vital in helping society assess and address widespread loneliness. However, while technology is providing ever-more powerful means of identifying societal problems such as loneliness, it remains questionable as to whether such problems can be solved by tech alone.

For the purposes of the study, the researchers interviewed 80 residents at an independent living sector of a senior housing community in San Diego County. They asked questions intended to gauge various aspects of loneliness, with the answers being transcribed and then analysed using the IBM Watson NLU (natural language understanding) iv program, which could “quantify sentiment and expressed emotions.”

Researchers have used artificial intelligence to accurately predict loneliness in residents at a senior housing community in San Diego. Publishing in the American Journal of Psychiatry, the researchers were able to harness natural language processing (NLP) and machine learning to classify the sentiment and emotions of speech.

At a time when the coronavirus pandemic is forcing people to remain in social isolation, the study by researchers from the University of California, IBM and elsewhere could prove vital in helping society assess and address widespread loneliness. However, while technology is providing ever-more powerful means of identifying societal problems such as loneliness, it remains questionable as to whether such problems can be solved by tech alone.

For the purposes of the study, the researchers interviewed 80 residents at an independent living sector of a senior housing community in San Diego County. They asked questions intended to gauge various aspects of loneliness, with the answers being transcribed and then analysed using the IBM Watson NLU (natural language understanding) iv program, which could “quantify sentiment and expressed emotions.”

These analytic methods operate by scanning for the frequency of words and phrases used in responses, and by assigning scores for sentiment (from -1.0 to 1.0) and also for emotion (from 0.0 to 1.0). The scores assigned by artificial intelligence were compared against manual assessments of loneliness, in order to evaluate their accuracy.

In their discussion of the results, the authors found that the machine learning models they used were surprisingly accurate. These models could predict qualitative loneliness (based on transcribed interviews) with 94% precision and quantitative loneliness (based on self-assessment scores) with 76% precision.

In other words, artificial intelligence is almost as good as qualified clinicians in predicting loneliness and isolation. As the authors state in their conclusion, this could have very significant implications for future:

“NLP and ML techniques can be scaled up to handle hundreds or thousands of interviews and can provide consistent ratings that may not be possible with human raters,” they write.

The authors also envisage a future scenario where artificial intelligence-based services could provide help to individuals, without the direct involvement of humans.

“Eventually, complex AI systems could intervene in real-time to help individuals to reduce their loneliness by adopting in [sic] positive cognitions, managing social anxiety, and engaging in meaningful social activities,” they say.

However, while artificial intelligence obviously has a future in the large-scale detection of loneliness (and other emotional states) in people and populations, it’s questionable as to whether it can be a significant part of the cure.

The research paper itself states that the overall incidence of loneliness among participants was 45%, with many reporting a lack of emotional and instrumental support. This lack isn’t something that AI-based systems on their own can resolve. Indeed, loneliness is fundamentally a social problem, and it can be solved only with social solutions and changes.

Put simply, it’s great that AI could theoretically identify every single lonely or isolated person in America or on the planet. But what could a tech-based approach do to actually reduce such loneliness and isolation? Very little, it could be argued.

This point is important because we all too often see technological innovations in, say, mental health diagnoses–or physiological health diagnoses–championed as if they were almost the same thing as actually healing the associated conditions. But while AI, virtual reality, and other technologies can certainly be used to detect problems, we need to remember that most of our problems aren’t caused by a lack of tech.

Rather, most of our problems originate form a complex web of causes and factors. Most of these causes and factors are social, economic, and political in character. As such, they will admit only of solutions that are similarly social, economic, and political.

This applies to loneliness, which is not only on the rise but likely a symptom of life in the increasingly individualistic and competitive 21st Century. If we’re really serious about loneliness, we need to look carefully at what aspects of our age are causing loneliness and change them accordingly. Otherwise, simply using artificial intelligence-based methods to detect and diagnose loneliness will amount to little more than another money-making exercise.

Write A Comment