PULSE: AI System Predicts Emotional Needs of Cancer Survivors
A team of researchers has created PULSE, an AI platform that leverages smartphone sensing information to anticipate when cancer survivors may require emotional assistance. This system utilizes LLM agents, which are equipped with eight different tools to analyze sensing data, assess current behaviors against personal baselines, and access historical cases. In a study involving 50 cancer survivors, the agentic multimodal condition demonstrated a balanced accuracy of 0.743 in forecasting the desire for emotion regulation. This research tackles the 'diary paradox,' which occurs when self-reports are lacking during crucial times. The findings are published on arXiv (2605.17679).
Key facts
- PULSE is a system for agentic sensing investigation.
- It uses LLM agents with eight purpose-built tools.
- The system queries smartphone sensing data.
- It compares current behavior with personal baselines.
- It retrieves outcome-labeled historical cases.
- The study involved 50 cancer survivors.
- The agentic multimodal condition achieved balanced accuracy of 0.743.
- The paper is on arXiv with ID 2605.17679.
Entities
Institutions
- arXiv