Closing the Parenting Intention-Behavior Gap: Automated Observational Coaching and the Observed Self in Everyday Family Life

Coaching from observation is common in sports: a swim coach sees a stroke break down and corrects it in the moment. For most parents, however, coaching is non-observational: they read a book, work with a counselor based on what they recall and report, or, more recently, a chatbot with the same limitation. Experiments with automated observation of parenting at home have been carried out, but relied on rudimentary counted words and turns, or sorted a few kinds of utterance, because that was all a machine could read from audio; language models change that. I describe an AI coach in which a wearable records ordinary family interaction, a nightly analysis finds moments that match or diverge from principles chosen by the parent, and then a morning report presents and coaches the previous day's unaligned moments, much as an athlete reviews game film. The AI coach keeps a record of the parent's observed behavior, noticing trends and challenges, and aims to help them gradually reduce the gap between the parent they intend to be and the one who appears at the end of a long day. I ask what this new approach affords in terms of self-understanding and also, crucially, the challenges the approach entails across four areas: the coach reads moments literally where the parent has context; the principles it applies must stay the parent's own; its understanding and coaching should adapt as the parent changes; and the implications of recording a household, not only the app user. For each I propose a design response, illustrated with a production system in pre-beta use that has been used by the author for four months.