DISCOVER: a Data-driven Interactive System for Comprehensive Observation, Visualization, and ExploRation of human behavior.

Abstract

Understanding human behavior is a fundamental goal of social sciences, yet conventional methodologies are often limited by labor-intensive data collection and complex analyses. Computational models offer a promising alternative for analyzing large datasets and identifying key behavioral indicators, but their adoption is hindered by technical complexity and substantial computational requirements. To address these barriers, we introduce <i>DISCOVER</i>, a modular and user-friendly software framework designed to streamline computational data exploration for human behavior analysis. <i>DISCOVER</i> democratizes access to state-of-the-art models, enabling researchers across disciplines to conduct detailed behavioral analyses without extensive technical expertise. In this paper, we are showcasing <i>DISCOVER</i> using four modular data exploration workflows that build on each other: Semantic Content Exploration, Visual Inspection, Aided Annotation, and Multimodal Scene Search. Finally, we report initial findings from a user study. The study examined <i>DISCOVER</i>'s potential to support prospective psychotherapists in structuring information for treatment planning, i.e. case conceptualizations.