08 Aug An algorithms that detect people with depression
The researchers from the University of Vermont and Harvard claim their algorithm’s detection rate is seventieth, adding that it’s “more reliable than the forty second success rate of general-practice doctors identification depression in person”.
“This points toward a replacement technique for early screening of depression and different rising mental diseases,” says Chris Danforth, of the University of Vermont.
The research was done in two stages: the first was about identifying the clues on Instagram photos that suggests the user might be depressed, while the second stage involved teaching the computer to detect those people using machine learning algorithms.
The scientists asked for access to the Instagram feed and mental health history of 166 participants, half of whom were reported to have been clinically depressed in the last three years.
They analysed around 44,000 photos and used findings from well-established psychology research to look for clues in Instagram images that could reveal the user’s state of mind.
“Pixel analysis of the photos in our dataset revealed that depressed individuals in our sample tended to post photos that were, on average, bluer, darker and greyer than those posted by healthy individuals,” the researchers wrote in a blog post.
They also found that people depression algorithmwithout depression opted for warmer and brighter photo filters, like Valencia, while black-and-white filters, such as Inkwell, were more popular among depressed people.
“In other words, people suffering from depression were more likely to favour a filter that literally drained all the colour out of the images they wanted to share,” the scientists continued.
In addition, the researchers also looked at photos that included people and found that depressed individuals were more likely to post selfies when compared with non-depressed people.
However, these photos had fewer faces on average than the healthy people’s Instagram feeds which, the scientists say, corresponds to existing research linking depression to reduced social interaction
The researchers next developed a machine learning algorithm based on established studies and pitted it against volunteers to see who was better at identifying Instagram photos posted by depressed people.
The scientists claim that artificial intelligence performed better, with Danforth saying: “Obviously you know your friends better than a computer.
“But you might not, as a person casually flipping through Instagram, be as good at detecting depression as you think.”
The scientists say their new study also showed that the algorithm was able to detect signs of depression before a person’s date of diagnosis.
Danforth points out that while their research holds promise, the technology is still far from perfect.
“This study is not yet a diagnostic test, not by a long shot,” he said. “But it is a proof of concept of a new way to help people.”
Figures from the mental health charity Mind show one in six people reports experiencing a common mental health problem (such as anxiety and depression) in any given week, while one in four in the UK will experience a mental health problem each year.