Speaking of Psychology: Twitter and ADHD, with Sharath Guntuku, PhD

Looking at large numbers of social media postings in aggregate can tell us quite a bit about Americans’ mental state. Sharath Guntuku, PhD, has analyzed the language in tweets to identify regional variations in stress and well-being.

Guntuku is a postdoctoral researcher in multimedia computing working at the Penn Medicine Center for digital health. His research aims to leverage large scale social media image and text data to model social health outcomes and psychological traits. He also works closely with the World Well-Being Project and the Positive Psychology Center at the University of Pennsylvania. Before joining the Penn Medicine Center, Guntuku completed his PhD studies on modeling users’ personality and multimedia preferences at the School of Computer Science and Engineering at Nanyang Technological University in Singapore. He also worked on facial expression recognition while interning at Panasonic.

Kim Mills: Hello and welcome to Speaking of Psychology, a podcast produced by the American Psychological Association. I'm your host Kim Mills. Speaking of Psychology is a podcast for anyone with an interest in the science of psychology. We talk to psychological researchers, practitioners, and educators about any and every aspect of psychology and its application to the world around us. Dr. Sharath Guntuku is a postdoctoral researcher in multimedia computing working at the Penn Medicine Center for digital health. His research aims to leverage large scale social media image and text data to model social health outcomes and psychological traits. He also works closely with the World Well-Being Project and the Positive Psychology Center at the University of Pennsylvania. Before joining the Penn Medicine Center, Dr. Guntuku completed his PhD studies on modeling users’ personality and multimedia preferences at the School of Computer Science and Engineering at Nanyang Technological University in Singapore. He also worked on facial expression recognition while interning at Panasonic. Thank you for joining us today.

Sharath Guntuku: Thank you very much for inviting me.

Kim Mills: So, let's start by talking about the research that you presented here. Can you explain what you did? I know it had to do with Twitter and Attention Deficit Hyperactivity Disorder.

Sharath Guntuku: So, we were looking at people who post that they had a diagnosis of ADHD on Twitter, and we got something like 1,400 people with the self-diagnosis of ADHD, and we analyzed how their linguistic patterns vary with a control group. Are they posting more at night? Are they posting more about being anxious? Are they posting more about being stressed? And how does their personality vary when compared to the control group? And some of the findings that we get from Twitter, it's interesting and they sort of confirm existing clinical studies on ADHD, and there are others which sort of give new hypotheses, which can then be tested in an actual clinical study. For instance, a lot of the studies have been have shown that anxiety is a comorbid condition with ADHD and we find a lot of people using a lot of anxiety related words on Twitter. And we also see that people are posting a lot at night so people with ADHD are posting a lot at night, between 7:00 a.m. and 6:00 a.m. when people are supposed to be sleeping. And this sort of suggests that Twitter is giving them the instant gratification that people with ADHD want. So, you post a tweet and you get instant comments. Yeah, so this platform is giving you that opportunity or to trigger your different traits.

Kim Mills: And that's something that ADHD people would be looking for is they want. They want that instant answers or whatever it is. Is there's something particular about people with ADHD that they'd be awake at those hours? Did you know that going in?

Sharath Guntuku: No, we didn't. So, we had no clue what times they were posting. We had a general idea that maybe they'll be posting more when compared to the control group, and that's something we found. But we didn't really know when they'll be posting. For instance, it was completely new for us and even a collaborator on the paper who is who sees patients with ADHD, and it was like “oh this is interesting” because often patients who come in to him they never report, or they seldom report that they were up until 4:00 a.m. because we don't really keep track of time that well. But then if you have a way to summarize the activity on social media and give that to clinicians, that's where we are trying to take this forward in terms of providing clinicians the right feedback about patients.

Kim Mills: So then what? They go back and ask their patients with ADHD “Do you have trouble sleeping?” and they can then help treat that?

Sharath Guntuku: Yes.

Kim Mills: Is there anything about the medication that somebody with ADHD might be taking that is going to keep them awake at night?

Sharath Guntuku: Frankly, I don't know the answer to that because I'm not an expert in ADHD, but we did see a lot of language related to weed-- smoking weed, and we were surprised. Why are people posting about smoking weed when they have ADHD? And then this collaborator of ours said “Oh, this is something that I always see. People are using weed to self-medicate”.

Kim Mills: Oh okay, that makes sense. So how did you get this cohort? I mean how would you know that just because somebody is on Twitter and says he or she has ADHD that that's true?

Sharath Guntuku: So, there is no way of formally confirming if they had an official diagnosis of ADHD, but there are a few studies which looked at depression related tweets. For instance, I was diagnosed with depression and then they asked people to, or rather, two independent assessors to go through these tweets and then give their assessment of whether these tweets look genuine or not. And those studies report high inter-rater agree ability, but then that's about the confirmation that we have so far about the data that we have collected. But going forward, especially with depression, we have data from multiple sources. For instance, we have a lot of people who have self-reported their depression diagnosis on Twitter and we have also asked a lot of people to take depression screening questionnaires like CESD or BDI and they have shared their social media data with us. So, they are consenting to share them with us and we also have clinical records of depression at Penn Medicine Center for Digital Health. So, we have these huge variety of data sources and now we have a way to compare which data is reliable in terms of whether the data that we collect based on the self-reports. If we build a model on that and test it on the clinical diagnosis, does it really work? And the note here is that even the clinical diagnosis themselves are not really robust because the idea of diagnosing somebody for these mental health conditions takes at least two or three hours, and often the diagnosis in the clinical setting in the primary care are not that rigid or that accurate so there is some fuzziness in the ground truth.

Kim Mills: Interesting. So, did you get permission from everybody to follow them on Twitter?

Sharatha Guntuku: So, we're not following anyone on Twitter. We were just using the Twitter API to look at public tweets. So, anyone who posts on Twitter, the data is public unless they set it to be private and we don't go into the private tweets because we don't have access. But we have IRB at the University which controls what we do with the data and how we store the data and so on. So, we are using mostly public tweets and we don't follow or we don't communicate or engage with the participants that we're actually recruiting.

Kim Mills: So, what are the next steps now based on what you learned? How does this get applied?

Sharatha Guntuku: So here, it's multifold actually. The first thing we're looking at is if we can develop some feedback apps so that you take our survey and then we give you, as the person who is taking the survey. Can we give some feedback about how they have been using social media the last week or the last month? “Here are the words that you're posting on social media last week and you've been swearing a lot. Perhaps you've not noticed, and you have been posting a lot at night. It looks like you should keep your phones away before you sleep or at least one hour before you sleep.” So, the we are looking at some of these feedback mechanisms to the user of social media and we are also working with clinicians to find out a way to summarize this data and give them in an actionable format.

Kim Mills: What made you choose ADHD out of all the possible syndromes that you could be looking at?

Sharath Guntuku: So, ADHD, unlike depression unlike say anxiety, is not really well studied in the in the literature and some of its manifestations are still being discovered. So, we thought social media being the in the moment feel for users, so we thought social media will give us some of these manifestations which are not yet discovered in the clinical setting and it will help us generate some of the hypothesis which we can go and test back.

Kim Mills: Great! Well that's all very interesting. Thank you so much for being with us today.

Sharath Guntuku: Thanks a lot! It was great!

Kim Mills: Speaking of Psychology is part of the APA podcast network, which includes other great podcasts such as APA Journals Dialogue, about the latest and most exciting psychological research, and Progress Notes, which discusses the practice of psychology. You can find all APA podcasts on iTunes, Stitcher or wherever you get your podcasts. You can also go to our website, www.speakingofpsychology.org, to listen to more episodes and see more resources on the topics we discuss. I'm Kim ills, with the American Psychological Association, and this is Speaking of Psychology.