Millions of people undergo traumatic experiences annually. While acute distress is a normative response to trauma, a small percent of the people who have undergone traumatic experiences continue to exhibit severe stress reactions long after the trauma. The posttraumatic reactions may include intrusive memories, hypervig- ilant arousal, impaired concentration, depression, emotional detachment from others, and disengagement from aspects of life that provide meaning and self-fulfillment. In functional assessments these recurrent reactions seriously impair intrapersonal, interpersonal, and occupational functioning. Despite these ramifications, post-traumatic stress disorders are under-reported, and in many cases go completely undetected. Barriers to help-seeking include: lack of knowledge about predisposition symptoms, the unavailability of appropriate remedial/prevention services, the fear of the social stigma of mental illness, lack of social support, or assumptions that the mood changes are a part of the overwhelming nature of post-crisis life. In recent times, many crisis rehabilitation efforts have recognized these challenges, and have incorporated access to mental health services in their intervention and support programs. However, treatment efforts, that typically comprise medication, therapy, or both, are more successful with early intervention;in fact, the likelihood of achieving full recovery decline as the illness lengthens. Social media such as Twitter and Facebook are increasingly serving an important role in crisis situations: aggregating and disseminating information, while providing opportunities for reflection and discussion of collective grief and trauma. These platforms thus provide an opportunity to investigate if post-traumatic stress and anxiety can be detected on a macro scale by studying the affective responses of crisis-inflicted populations. Due to social media's archival record, reactions to societal crisis and traumatic happenings can be tracked longitudinally. We propose to mine behavioral data from these platforms to better understand the range of responses of people to crises. By applying state-of-the-art techniques from text and social network analytics, we will go be- yond population-level estimates of crisis behavior, to gauge the social, psychological, emotional, and linguistic attributes of specific crisis-laden communities, and the relationship of these behavioral attributes to key mental and public health outcomes in a crisis context. Specifically, we will accomplish the following research aims: 1. Design data mining techniques that can intelligently filter social media posts for crisis-relevant content, incorporating statistical correction methods that can extract population representative samples. 2. Use linguistic signals and social network metadata to identify key communities in the crisis-laden population. 3. Develop behavioral measures from the activities of these communities that reflect the extent of their risk to stress disorders from the crisis situation. Ths will include measures that automatically expand and refine existing technologies to match community-specific language and dialects, which can vary dramatically in social media writing, as well as factor in the unique context of crisis events. 4. Perform a longitudinal study of the behavior of crisis-embroiled communities, identifying the factors that indicate especially high-ris groups. If successful, this research will (a) bring to the fore variables related to the exacerbaton of (or even predisposition to) PTSD, (b) enable new mechanisms to identify at-risk communities in a near real-time fashion, and (c) lend a complementary perspective on current research around trauma diagnosis, which relies typically on laboratory studies and surveys. On a practical note, a potential link between behavior manifested in social media in the context of a crisis, and anxiety and post-traumatic stress symptomatology can also augment traditional efforts in providing valuable interventions as a part of disaster response. This work is a close collaboration with computational linguist and faculty of School of Interactive Computing at Georgia Institute of Technology, Dr. Jacob Eisenstein, who is an expert in the discovery of social relationships latent in linguistic data, and Dr. James Pennebaker, a social psychologist and faculty in the Department of Psychology at the University of Texas, Austin, whose expertise is in the area of psychological interpretation of language cues.

Public Health Relevance

An important area of investigation in behavioral health research is to understand the implications of crisis and disaster experiences on people's cognitive, affective, and behavioral processes. Crisis, especially if prolonged, can have several negative health implications. It is known to lead to declines in cognitive performance and attentional impairment, and is a known stressor of the onset of PTSD (post-traumatic stress disorder), an anxiety disorder widely recognized as having physiological concomitants. These challenges make the study of the relationship between protracted crisis exposure and social, psychological, and affective reactions in people vital from a public health perspective. Recognizing these needs, there have been efforts in the recent past where crisis responders and non- government agencies have incorporated mental health services into disaster response. However, it is well-known that survivors benefit best from early diagnosis, wherein treatment may range from brief interventions in the weeks and months after disaster exposure, to referral to mental health treatment if symptoms do not remit with brief interventions. In fact, the likelihood of achieving full recovery declines as the illness lengthens. Additionally, the diagnosis process has challenges as well: due to the inaccessibility of disaster survivors to mental health services, broad community-level diagnoses rely on government or agency administered surveys. However, typically there are large temporal gaps in which these surveys are administered and due to their observations being mostly retrospective, the surveys lack the contextualization that is needed to associate the manifested affective concerns to actual circumstances people face. This also makes it difficult for concerned agencies to track and identify risk factors that may be associated with post-traumatic mental illness, or to develop effective intervention programs. Complementing these surveys with near real-time information, such as from social media, about people's behavioral reactions seems to be an effective way of empowering people with tools and platforms that can provide them with psychological and emotional support in the context of catastrophic disasters. In the light of the above public healt relevance, the goal of this project is to examine the potential of online social networks and socia media as an unobtrusive lens on the behavior of crisis-laden communities, and identify factors indicative of their vulnerability or predisposition to post-traumatic stress and anxiety. Page 2

National Institute of Health (NIH)
National Institute of General Medical Sciences (NIGMS)
Research Project (R01)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1-HDM-Q (55))
Program Officer
Marcus, Stephen
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Georgia Institute of Technology
Biostatistics & Other Math Sci
Other Domestic Higher Education
United States
Zip Code
Baumel, Amit; Baker, Justin; Birnbaum, Michael L et al. (2018) Summary of Key Issues Raised in the Technology for Early Awareness of Addiction and Mental Illness (TEAAM-I) Meeting. Psychiatr Serv 69:590-592
Blackburn, Kate G; Yilmaz, Gamze; Boyd, Ryan L (2018) Food for thought: Exploring how people think and talk about food online. Appetite 123:390-401
Saha, Koustuv; Weber, Ingmar; De Choudhury, Munmun (2018) A Social Media Based Examination of the Effects of Counseling Recommendations After Student Deaths on College Campuses. Proc Int AAAI Conf Weblogs Soc Media 2018:320-329
Birnbaum, Michael L; Ernala, Sindhu Kiranmai; Rizvi, Asra F et al. (2017) A Collaborative Approach to Identifying Social Media Markers of Schizophrenia by Employing Machine Learning and Clinical Appraisals. J Med Internet Res 19:e289
De Choudhury, Munmun; Kumar, Mrinal; Weber, Ingmar (2017) Computational Approaches Toward Integrating Quantified Self Sensing and Social Media. CSCW Conf Comput Support Coop Work 2017:1334-1349
De Choudhury, Munmun; K?c?man, Emre (2017) The Language of Social Support in Social Media and its Effect on Suicidal Ideation Risk. Proc Int AAAI Conf Weblogs Soc Media 2017:32-41
Saha, Koustuv; Weber, Ingmar; Birnbaum, Michael L et al. (2017) Characterizing Awareness of Schizophrenia Among Facebook Users by Leveraging Facebook Advertisement Estimates. J Med Internet Res 19:e156
Bagroy, Shrey; Kumaraguru, Ponnurangam; De Choudhury, Munmun (2017) A Social Media Based Index of Mental Well-Being in College Campuses. Proc SIGCHI Conf Hum Factor Comput Syst 2017:1634-1646
Chancellor, Stevie; Mitra, Tanushree; De Choudhury, Munmun (2016) Recovery Amid Pro-Anorexia: Analysis of Recovery in Social Media. Proc SIGCHI Conf Hum Factor Comput Syst 2016:2111-2123
De Choudhury, Munmun; Kiciman, Emre; Dredze, Mark et al. (2016) Discovering Shifts to Suicidal Ideation from Mental Health Content in Social Media. Proc SIGCHI Conf Hum Factor Comput Syst 2016:2098-2110

Showing the most recent 10 out of 14 publications