Prompt strategies improve social media stress and emotion prediction
Identifying Effective Strategies in Prompt Evolution for Stress and Emotion Prediction
Computation and Language
Summary
Detecting stress and emotions early in social media posts can help provide timely support. The authors studied different ways to improve computer prompts that guide language models in predicting stress and emotions. They found particular strategies, like focusing on emotional clues and distinguishing recent from past feelings, help improve prediction accuracy. The study also showed how some improvements are seen early on while others vary based on the task and evaluation.
What this means in practice
- •For mental health app developers: Implement improved prompt strategies to enhance early detection of stress signals in user-generated social media content.
- •For sentiment analysis teams: Incorporate combined instruction prompts to boost emotion prediction accuracy on social media datasets.
Authors
SeungHyun Yi, Seungjun Yi, Sriram Sattiraju, Haeun Park, Joakim Nguyen, Terence Lim, Joseph Skrovan, Mehak Beri, Hitakshi Modi, Eloy Carrion, Andrew Well, Carlos M. Mery, Alexander Rasgon, Mia K Markey, Ying Ding
Abstract
Early stress detection and risk prediction are crucial for timely mental health assessment and support. While large language models have recently shown promise for social media stress prediction, relying on accuracy can obscure false positives. Also, prompt revisions yield inconsistent gains across models and tasks, with unclear links to strategy combinations and instruction changes. By comparing six methods on social media stress and emotion data across two model sizes, we found that GEPA and its extension combining instructions from two prompts (GEPA+merge) improved prediction over initial prompts in several settings. We also identified five prompting strategies associated with development gains. These strategies included directing attention to emotional evidence, distinguishing current from past distress, illustrating indirect cues, requiring emotional context, and guiding label selection. Moreover, we documented four observations across tasks and evaluation settings. In a separate set of searches, binary stress prediction improved early, whereas multilabel emotion prediction initially declined and partly recovered. The other three concerned incorrect or conflicting instructions, gains on small comparison samples that did not hold on full development data, and changes in selected prompts' development and reported test rankings. These findings identify effective prompt evolution methods, strategy combinations associated with gains, and task and evaluation limits, providing guidance for applying prompt evolution methods to social media stress detection.