Inferring Mood Instability on Social Media by Leveraging Ecological Momentary Assessments

Abstract

Active and passive sensing technologies are providing powerful mechanisms to track, model, and understand a range of health behaviors and well-being states. Despite yielding rich, dense and high fidelity data, current sensing technologies often require highly engineered study designs and persistent participant compliance, making them difficult to scale to large populations and to data acquisition tasks spanning extended time periods. This paper situates social media as a new passive, unobtrusive sensing technology. We propose a semi-supervised machine learning framework to combine small samples of data gathered through active sensing, with large-scale social media data to infer mood instability (MI) in individuals. Starting from a theoretically-grounded measure of MI obtained from mobile ecological momentary assessments (EMAs), we show that our model is able to infer MI in a large population of Twitter users with 96% accuracy and F-1 score. Additionally, we show that, our model predicts self-identifying Twitter users with bipolar and borderline personality disorder to exhibit twice the likelihood of high MI, compared to that in a suitable control. We discuss the implications and the potential for integrating complementary sensing capabilities to address complex research challenges in precision medicine.

A computational study of mental health awareness campaigns on social media

“…This is essential due to the fact that in many real-world scenarios involving automatic categorization of content, like ours, and because labeled data are both expensive and timeconsuming to gather as well as scarce (e.g., needs expert involvement and manual labor), although unlabeled data are comparatively huge and easier to gather. To build robust classification models that can generalize across datasets, settings, and mental health campaigns, we use semi-supervised learning [35][36][37] in this second phase, that is able to leverage both labeled and unlabeled data in unison, thereby is to cover a better diversity of training examples [38].…”

The Tesserae Project

“…Further, researchers from different time zones could cover more sessions, allowing participants to enroll within 8 am and 10 pm (local times). We had a combination of student and postdoc researcher proctors (6) and externally paid proctors (3) to conduct these remote enrollment sessions. However, remote enrollment had its downsides, including a wide range of technological failures, e.g.…”

“…At the same time, researchers have been discovering the potential of wearable devices and other sensors for use as research tools to understand individuals in their natural environments at-scale. Some early efforts include projects such as NetHealth [4], StudentLife [7], CampusLife [6], and WorkSense [2] that share several key characteristics. First, they have used multiple sensors to measure different individual attributes such as mood, movement, and technology use.…”

Designing a Clinician-Facing Tool for Using Insights From Patients’ Social Media Activity: Iterative Co-Design Approach

“…In recent years, patient-generated data, machine learning, and digital health technologies have been recognized to offer unprecedented opportunities to help consumers, clinicians, researchers, and other stakeholders measure, manage, and improve mental health [ 10 - 12 ]. In particular, digital records of people’s activities on social media, such as Facebook and Twitter, have been shown to enable the development of machine learning algorithms that can assess and even predict risk to a variety of mental health challenges [ 10 ], such as major depression [ 13 ], postpartum depression [ 14 ], posttraumatic stress [ 15 , 16 ], and schizophrenia [ 17 ].…”