Socioduality: A New Framework for Human-AI Interaction

A recent article on arXiv (2608.11322) presents the concept of 'socioduality,' a framework aimed at examining interactions between humans and AI. The researchers contend that conventional assessments of individual skills, combined outputs, or overall performance overlook the interactive, back-and-forth nature of these exchanges. Socioduality is characterized as a process where one participant's reaction influences the conditions for the other participant's next input. This framework is tailored for human-AI pairs and includes nested components: moves, confirmed sociodual episodes, linked pathways, and the overarching interaction context. A basic episode (A1-B1-A2) necessitates proof of response contingency and return contingency. Candidate episodes are categorized as confirmed, non-sociodual, or indeterminate prior to further coding of response orientation and subs. This paper also serves as a cross-type announcement, suggesting prior publication. The findings enhance the expanding domain of human-AI interaction research, providing a more detailed model for assessing dynamic exchanges.