Using Open-Ended Stressor Responses to Predict Depressive Symptoms Across Demographics

Explore the complex relationship between stressors and depression in this 31-minute conference talk from the Center for Language & Speech Processing at Johns Hopkins University. Delve into a study that analyzes open-ended text responses about stressors and their correlation with depressive symptoms across different gender and racial/ethnic groups. Learn how open-ended responses in survey instruments can provide more nuanced insights compared to traditional multiple-choice questions, particularly in mental health contexts. Discover the potential of language models in automatically analyzing these responses, while also considering the associated risks such as biases. Examine the findings that demonstrate a relationship between stressors and depression, and how these trends differ across demographic groups. Gain insights into how these differences impact downstream performance across various demographics. This talk is based on research published in a paper available on arXiv.