Human health coaching has been established as an effective intervention for improving clients’ health, but it is restricted in scale due to the availability of coaches and finances of the clients. We aim to build a scalable, automated system for physical activity coaching that is similarly grounded in behavior change theories. In this paper, we present our initial steps toward building a flexible system that is capable of carrying out a natural dialogue for goal setting as a health coach would while also offering additional support through just-in-time adaptive interventions. We outline our modular system design and approach to gathering and analyzing data to incrementally implement such a system.

We envisioned responsive generic hierarchical text summarization with summaries organized by section and paragraph based on hierarchical structure topic models. But we had to be sure that topic models were stable for the sampled corpora. To that end we developed a methodology for aligning multiple hierarchical structure topic models run over the same corpus under similar conditions, calculating a representative centroid model, and reporting stability of the centroid model. We ran stability experiments for standard corpora and a development corpus of Global Warming articles. We found flat and hierarchical structures of two levels plus the root offer stable centroid models, but hierarchical structures of three levels plus the root didn’t seem stable enough for use in hierarchical summarization.