This review examines how love, attachment, and parasocial relationships are operationalized in empirical studies of interactions with conversational artificial intelligence. Despite the rapid growth of research in this area, the ways in which emotional bonds with these systems are defined and measured remain fragmented across disciplines. The reviewed literature was synthesized across four methodological traditions: psychometric, qualitative, computational, and behavioral, each capturing distinct dimensions of relational engagement while systematically overlooking others. The findings indicate that such bonds represent a hybrid form of engagement that is simultaneously real in its emotional consequences and simulated in its underlying mechanisms. Existing frameworks, largely adapted from interpersonal or media contexts, provide only partial accounts of this phenomenon. To address this limitation, the review proposes a grounded distinction between simulated reciprocity as a property of conversational systems and projective bonding as a property of user cognition, and considers them as two complementary constructs that may serve as a basis for future operationalization. This distinction emerges from systematic discrepancies between system design properties and user-reported experiences observed across the reviewed studies. Together, these mechanisms provide a framework for understanding the structure of these relationships and help explain inconsistencies across existing measurement approaches. The review highlights key measurement gaps and outlines directions for future research, including the integration of psychometric, computational, and behavioral methods within preregistered, multimodal study designs that can examine whether these mechanisms operate as separable construct.
Malejka Patryk (Thu,) studied this question.