Affective personal informatics (PI) is the practice of capturing, tracking, analyzing, and reflecting on personal data to support emotional self-awareness and mental well-being. A new preprint on arXiv surveys how large language models (LLMs) are entering this space.
The review maps the current landscape of LLM-based affective PI, examining how these models can process and interpret emotional data in ways that earlier tools could not. It also reflects on the implications for users and designers.
The authors outline future directions for research, suggesting where LLM-based systems might improve emotional reflection and well-being support. As a scoping review, the paper aims to consolidate existing work and identify gaps.