Relational Role Attribution: A Unifying Construct in Human–AI Interaction

How users come to perceive an AI not merely as capable but as intentionally oriented toward them in a specific relational capacity remains undertheorized. We propose relational role attribution (RRA) as a discrete psychological construct addressing this gap, positioned as a precursor to downstream relational outcomes rather than a reframing of trust. A PRISMA-ScR scoping review charted 1223 reports at full text against a registered protocol, including 466. Of 179 records excluded for establishing no relational or social role for the agent, 122 (68.2%) nonetheless manipulated or measured anthropomorphism: the field reaches the cue and stops short of the attribution. RRA comprises two concurrent components, a relational inference of social orientation and a role assignment drawn from contextual and communicative cues, checked against the expectancy the user brings. Naming the attribution allows findings accumulated separately across marketing, teaming, and communication literatures to constrain one another, and scaffolds role-specific design principles.