The Data We Cannot Know —Nikitha Donekal Chandrashekar, Virginia Tech

The Data We Cannot Know: Privacy, Cybersecurity, and the Future of Clinical Interaction Research High-stakes clinical environments depend on rapid interpretation, coordinated action, and continual adaptation among clinicians, yet the structure of this activity remains difficult to study at scale. Existing approaches typically rely on qualitative fieldwork, simulation, or retrospective video review. Although these methods provide valuable insight, they may lack ecological fidelity, require labor intensive annotation, or collect identifiable audiovisual data that create substantial privacy, security, and governance concerns. This paper argues that a central methodological challenge for future HCI research in clinical environments is to collect clinically meaningful interaction data without introducing unacceptable privacy risk. We conceptualize clinical activity through three interaction classes: clinician–clinician coordination, clinician–patient interaction, and clinician–tool interaction. We then propose a low-identifiability multimodal methodology that combines full-body skeletal tracking, RFID-based tool tracking, and clinician localization to characterize team behavior without directly retaining conventional audiovisual identifiers. Privacy is treated not as a downstream de-identification task, but as a property designed into the sensing pipeline. We further outline a phased study and evaluation plan, grounded in simulated traumatic brain injury resuscitation, that pairs quantitative multimodal sensing with structured expert debriefs. The long-term objective is to support AI-assisted analysis of clinical activity sequences, distinguish meaningful adaptations from potentially harmful deviations, and develop real-time systems that augment clinical teamwork.