An interdisciplinary team led by Columbia University researchers has received a National Science Foundation grant to study the online expression of grief in Black communities and to develop computational methods to automatically identify these expressions.
Computer science professor Kathleen McKeown, social work and sociology associate professor Desmond Upton Patton, psychiatrists, linguists, and community members will all collaborate to explore how emotions related to various triggering events, including COVID-19, police brutality, the economic downturn, or the loss of a loved one, are expressed online.
“Black grief is a phenomenon that is not well understood, especially when it occurs in a networked public,” said Patton, who also serves as associate director of diversity, equity, and inclusion for Columbia’s Data Science Institute (DSI). “The impact of co-occurring pandemics—COVID-19 and anti-Black racism—have elucidated new and historic ways individuals may cope with grief in physical and digital spaces.”
Given the frequency with which people post online, the ability to automatically flag posts that indicate that an author may need help would have profound implications for the fields of counseling and social work. It would certainly be more efficient than the current approach of manually scanning online spaces. This development could shift how social workers, mental health professionals, and outreach workers treat complex grief online and may inform new intervention and treatment programs that respond to an individual’s digital life.
Project results will include an annotated dataset that scholars and computational researchers may use to understand the online expression of Black grief. This dataset will also serve as a basis for computational models developed to automatically identify online posts related to expressions of grief.
McKeown, who was founding director of DSI and serves as principal investigator of the grant, noted that automatic methods for identifying expression of grief—and the events that trigger it—do not currently exist. “This project is unique in its focus on computational models developed specifically for African American English. The vast majority of [natural language processing] models are trained on data drawn from Standard American English,” she said.
The researchers will collect, analyze, and annotate social media posts containing expressions of grief to identify how people communicate about different types of loss. These posts will be analyzed with a focus on semantic interpretation and context, psychological interpretation of expressed emotion, and linguistic expression of grief, yielding a richly layered and annotated dataset.
Linguistic analyses by Jessie Grieser‘s team at the University of Tennessee, Knoxville will explore the significance of linguistic variation online and the role of specific digital language strategies in the creation of social meaning. Columbia internist and psychiatrist M. Katherine Shear will conduct focused, in-depth interactions with participants to fully understand the emotions they are experiencing and the depth of their grief.
The social work team will augment and qualify the automated results with a qualitative analysis of the data informed by an understanding of historical trauma, bias, and racism. Black Harlem residents will also serve as “experiential experts” and will engage in the evaluation of the tools and research methods.
“It is critically important that practitioners and community experts be front and center with any study of Black grief, especially one that involves [artificial intelligence], to ensure that we have the most accurate interpretation of culture, nuance, and culture expressed in text,” Patton explained.
Community experts will also work with researchers to decipher hyperlocal language. There will also be collaborations to better understand how the automated predictions may change when the computation model is applied to different demographic groups (e.g., age, socioeconomic status).
— Karina Alexanyan, Ph.D.