19:30 - 20:30 | Tue 6 Mar | Caribbean ABC | TuPO.8
Biometric sensors and mobile devices enable the monitoring of individuals over long periods of time providing new insights into diagnostic and therapeutic means. We demonstrate an application of such technology in the Family Studies domain, where we monitor couples over a one-day period. Through passive ambulatory data collection, we aim to automatically detect instances of interpersonal conflict in everyday life, since conflict is long recognized as having deleterious effects on couples' relationship quality and outcomes. We design a multimodal set of features and develop machine learning systems to classify between the presence or absence of conflict within 3-min time intervals. Our preliminary results indicate up to 0.17 and 0.80 F1-scores for the conflict and non-conflict class and up to 53% unweighted recall for our task.