Your Instagram filter may reveal more than you realize about your mental health.
Researchers from Harvard and the University of Vermont have found that Instagram photos can be analyzed to screen for depression. The scientists used the photos' attributes, including brightness and color, to correctly identify which participants suffered from depression at a better rate than the typical physician.
Photos with decreased brightness, decreased saturation and increased hue indicated depression. The computer correctly identified 70% of the instances of depression.
Valencia was the most popular filter with users who were not depressed, whereas depressed users were most likely to use Inkwell. Valencia lightens photos whereas Inkwell converts images to black and white.
The findings point to an unobtrusive, inexpensive way to use social media to detect health issues.
"It's all a good thing," said Michael Thase, the director of the Mood and Anxiety Program at the University of Pennsylvania. "It's only not a good thing when the information would be sold to somebody who might make a buck from interfacing with depressed people."
Thase suggested that in the future we may opt in to receive screening feedback from social media. A user who gave consent would have their photos scanned for signs of potential illnesses.
"Did you know your choice of hues and colors go along with people who are prone to depression. Would you like to know more?" Thase suggested as a possible private message that could be sent to users who appeared depressed.
To reach their conclusions, the researchers analyzed roughly 13,000 photos from 166 Instagram users, some of whom were clinically depressed.
A computer system was trained by looking at the photos' brightness, vividness, hue and whether an Instagram filter was used. The computer learned to make predictions of depression after comparing the photos of depressed individuals, and those who are not depressed.
The research is the latest example of how useful insights can be gleaned from the digital footprints we leave online. Tech companies will increasingly have opportunities to assist patients and their health care providers, according to Thase.
So how would such a program appear in the real world? It would likely start with a tech company inviting users to participate, and encouraging the traditional medical community to endorse it. Information could potentially be automatically shared with a patient's physician, provided the patient opts in.
Earlier this year, Microsoft (Tech30) researchers showed that a person's Web search history could sometimes predict an upcoming diagnosis of pancreatic cancer. ,