Title: Social Media-Powered Artificial Intelligence for Patient-Clinician Collaboration in Mental Health Consultations: A Human-Centered Design Approach

Date: Tuesday, April 12th, 2022

Time: 5:00 - 7:00 PM ET

Location (in-person): CODA C1215

Location (remote): click here to join via Zoom 

Dong Whi Yoo

PhD student in Human-Centered Computing

School of Interactive Computing

Georgia Institute of Technology

Committee:

Dr. Munmun De Choudhury (co-advisor), School of Interactive Computing, Georgia Institute of Technology 

Dr. Gregory D. Abowd (co-advisor), School of Interactive Computing, Georgia Institute of Technology & College of Engineering, Northeastern University

Dr. Jennifer Gahee Kim, School of Interactive Computing, Georgia Institute of Technology

Dr. Andrea Grimes Parker, School of Interactive Computing, Georgia Institute of Technology

Dr. Mary Czerwinski, Human Understanding and Empathy group, Microsoft Research Redmond

Dr. Madhu Reddy, Donald Bren School of Information and Computer Sciences, University of California, Irvine

Abstract:

Patient-clinician collaboration during mental health consultations is crucial for improving clinical outcomes. During collaboration, a patient and a clinician unpack and interpret the patient’s illness experiences and daily lives to make clinical decisions. However, it is challenging to communicate those experiences in sufficient detail due to issues of recall bias, limited time, and the subjectivity of illness experiences. To address this challenge, social media data has been touted as a promising vehicle to support this collaboration given that it provides a rich context regarding patients’ feelings, thoughts, and social functioning recorded close to eventful moments. To concretize the potential of social media data in mental health, researchers have utilized AI technologies to process large amounts of patient social media data to provide clinically relevant information. I refer to this as Social Media-Powered Artificial Intelligence (SMPAI).

Incorporating SMPAI into a context as sensitive as mental health consultations could present several issues and risks to both clinicians and patients. Clinicians may suffer from information overload and choose not to adopt the technology if SMPAI is incompatible with their workflows. Patients may feel uncomfortable and be less willing to provide their social media data if SMPAI fails to provide transparency and control over the sharing process. To address those challenges, I have been working with mental health patients and clinicians to design and examine SMPAI that can support patient narratives and clinical decisions while also assuaging ethical concerns.

In this thesis proposal, I first explored the opportunities and challenges of SMPAI by examining how mental health patients discuss their social media activities to augment their narratives during consultations. To understand if and how SMPAI could be incorporated into clinical workflows as collateral information, I designed and evaluated a SMPAI prototype with clinicians to explore the potential usages of computationally derived insights from patient social media. In my current work, I have been examining the potential of SMPAI as Clinical Decision Support Systems (CDSS). I have been working with patients, clinical researchers, and AI researchers to understand how they collaborate to develop predictive models, which can be used in CDSS in the future. In my proposed work, I plan on taking these insights further by co-designing SMPAI that can be used as CDSS for schizophrenia patients and clinicians regarding relapse-related decisions. I will then conduct a formative evaluation of my SMPAI prototypes to learn design implications for useful and implementable SMPAI. These design implications from the co-design and the formative evaluation will contribute to digital mental health, human-AI interaction, and computer-supported cooperative work.