DS@GT at eRisk 2025: From prompts to predictions, benchmarking early depression detection with conversational agent based assessments and temporal attention models
This work addresses depression screening through conversational AI, but it is incremental as it applies existing methods to a new benchmark without ground-truth labels.
The paper tackled early depression detection by using prompt-engineered LLMs to conduct BDI-II-based assessments from conversations, achieving second place on the leaderboard with metrics like DCHR=0.50 and ADODL=0.89.
This Working Note summarizes the participation of the DS@GT team in two eRisk 2025 challenges. For the Pilot Task on conversational depression detection with large language-models (LLMs), we adopted a prompt-engineering strategy in which diverse LLMs conducted BDI-II-based assessments and produced structured JSON outputs. Because ground-truth labels were unavailable, we evaluated cross-model agreement and internal consistency. Our prompt design methodology aligned model outputs with BDI-II criteria and enabled the analysis of conversational cues that influenced the prediction of symptoms. Our best submission, second on the official leaderboard, achieved DCHR = 0.50, ADODL = 0.89, and ASHR = 0.27.