DECOMPOSING DEPRESSION: A COMPARATIVE STUDY ON SELF-DISCLOSURE USING WEB-SURVEYS AND CONVERSATIONAL AGENTS

Data quality has been a major challenge in precision healthcare. In this study, we used two data-gathering modalities, conversational agents (like chatbots) and web surveys, to explore how social structures, like genre rules associated with each modality, influence the content and structure of collected data, especially on sensitive topics such as depression. In general, these genre rules influence how participants interact and shape the information they disclose. Our results showed that the content and structure of self-disclosed information differ between modalities and that each modality also influences the depth and style of the information reported. We believe that these findings highlight the importance of carefully considering different data-gathering methods in precision healthcare and clinical research, particularly in experiment design.