Pranika Jain and Siddhartha Roy Nandi, two graduate students in Syracuse University’s College of Engineering and Computer Science, developed an app that allows psychiatrists to see their patient’s social media posts and alert the doctor of concerning activities. The Daily Orange sat down with Jain and Roy Nandi to discuss the app and its development.

The D.O.: What’s the purpose of the app, and how does it work?

Siddhartha Roy Nandi: … The purpose of the app is authenticating doctors as well as patients, registering them and collecting the data. There is another section, which is the learning algorithm, which is running in on the back end, but is actually working with the app. The algorithm takes a look at the patient’s social media activities, and if any kind of anomalies happen, then it generates an alert. And because of that alert, a corresponding alert goes to the corresponding doctor.

The D.O.: Where did you get the idea to incorporate social media?

S.R.: Social media is very popular nowadays.

Pranika Jain: It’s the most expressive platform.

S.R.: Many researchers are going on … social media to look at the sentiments of the individuals over social media. So the idea came to us as whether we can see the behavioral change in social media, also.

The D.O.: How is language used as an indicator?

S.R.: Language might be an indicator, but apart from that there are several other indicators. For example, if you post more than your average postings, that means … something must have happened. And then you can take a look at the sentiment analysis of those posts. If your average behavior is negative, then it is more likely that you are a negative person or you are in a depressed mindset, but if the frequency of the posts is very high and all the posts are negative, then it is more like you are in trouble and indirectly you’re crying for help. There are starting words you can look for. So when I mention negative, there has been a purpose of negative words like “bad,” “sad,” “die,” “kill.” So starting words can be used to find out the behavior of the negativity. For that, we are consulting with a psychiatrist to build the corpus of what words we should (use to) consider that post as negative.

The D.O.: How will the doctors be able to see the posts for themselves?

P.J.: How this works is, whenever something indicating of self-harm comes across these feeds, then an email is being sent to (the) doctor telling them that there is some data which is anomalous in behavior.

S.R.: The alert rate will tell the doctor that there is anomaly to the corresponding patient, and if the doctor wants to see the post which caused the behavior, the doctor can go to that post.

P.J.: Suppose a doctor logs into his account, you can see four feeds. First is inconsistency. This comes up whenever there is anomalous behavior found in the feed of the patient. If a doctor clicks, it’ll show patients who are having some anomalous behavior or are considered to be mentally sick by the algorithm. After that, if you click on “show feeds,” a doctor will be able to see the feeds, which will be a single feed of that patient. Which says, “Because of this feed, I am indicating for you that some problem occurred.”

The D.O.: Would the doctors be having the conversation with the patient beforehand?

P.J.: Suppose a patient comes to a doctor. The doctor would ask, “Would you like to share your Facebook data?” Then the doctor will hand over his phone to that patient, and (the) patient will log in with Facebook credentials. After that, the complete details of the patient — which will include his date of birth, his feeds and some of the other features like his address — will be recorded in a database. This will be a one-time login, and at the back end, we’ll be getting the complete data.

The D.O.: How far along are you with the app?

S.R.: The app part is almost done, because the purpose of the app is registering the patients and the doctors, and that part has already been developed. The purpose of the app is (also) to collect the data. We’re almost done with the app. The main objective is we’re developing the algorithm on the back end. And based on the more data we will see, the more accurate the algorithm will become. We will keep on working on the algorithm more in the future.

The D.O.: Are you excited to really be able to launch it?

S.R.: Of course. The sooner I launch, the sooner I will get more data for analyzing.

P.J.: We need to get our first version out to see how the reaction is.
S.R.: There are certain legal issues to allow an app in the market that captures users’ personal data. Unless you can get those permissions, you cannot launch the app, so we’re waiting for that permission.

Published on February 21, 2018 at 9:21 pm