A Q&A with a UCIPT Postdoctoral Researcher
Would it be possible for college admissions personnel to use Twitter to find students who are most able to cope with college stress?
I believe that is possible. There was a recent study by Dr. DeChoudhury, and her team found that a person’s tweets over a one-year period could be used to detect depression with 70% accuracy. They used variables such as content of the tweets, time, and frequency of the tweets to build the depression prediction algorithm. We can apply these algorithms and tune it to predict other health outcomes such as stress coping. This is a very new field of study and we still have a lot to learn before we can accurately predict the ability of students to cope with college stress.
How can college faculty and staff spot warning signs on social media that a student might be at risk of harming him/herself or someone else?
A few studies have tried to use social media data to predict depression. From those studies, we learned that individuals with depression were more likely to post messages late at night compared with non-depressed individuals. Certain isolated words in a tweet (e.g., anxiety, suicidal, fatigue, additive, attacks) have also been found to be associated with depression. The overall volume of the tweets seems to change as well. Individuals with a sudden decrease in the frequency of posts were associated with a higher risk for developing depression. As I mentioned earlier, the accuracy of using social media data to predict depression is at about 70%. We are currently developing several new machine learning and data mining techniques that we hope can improve the accuracy of the prediction models.
Are students more likely to search for help on Twitter or other specific forums? Are there mental health agencies actively trying to sync up with people looking for help already on Twitter or in those other forums? If not, what kind of resources are needed?
We are not sure whether the students are more likely to search for help on Twitter or other specific forums, but we know that about 80% of Internet users actively seek health information online. Currently, there is a suicide prevention initiative between Facebook and the National Suicide Prevention Lifeline. I think partnerships like these are extremely important. However, the current technology relies on the users reaching out to the mental health agencies. Our research goal is to build an intelligent system that can help mental health agencies to automatically scan people’s Twitter feed and offer assistance if necessary.
If you were a psychologist, would you prescribe Twitter to an adolescent patient suffering from social anxiety?
It is estimated that about 19 million Americans are diagnosed with social anxiety. Currently, social anxiety is the third most common mental disorder in the United States. Several studies have reported that some online forums and Internet-based interventions can be effective in treating social anxiety. I do think psychologists can use Twitter to help those suffering from social anxiety, but it is important to develop a specific treatment plan that is clinically organized. Simply “using” Twitter may not be helpful to those suffering from social anxiety. I think this is a very exciting area of research since no studies so far have examined the effectiveness of prescribing Twitter to an adolescent patient suffering from social anxiety.
How does student Twitter behavior translate into real-world student behavior? Are there coping skills that they can gain from Twitter?
This is one of the main questions that we are trying to answer in our Freshman 200 study. We have been tracking 200 college students at UCLA over an entire semester, starting in October 2015. We asked the students to answer weekly surveys about their lifestyle behaviors and we asked them to wear a smart watch that tracked their physical activity, heart rate, and sleep quality. We will be comparing their Twitter behavior to their real-world behavior, and we are currently just starting to analyze the data. We will have an answer to this question fairly soon. So stay tuned!