Whose agent are you? Relational norms shape expectation from algorithmic and human advisors in social decisions

As tech companies develop AI companions designed to function as friends, therapists, and personal advisors, a fundamental question emerges: can algorithms truly occupy these intimate social roles? Our research examines these expectations in social decision-making contexts. Across two experiments (N = 492) involving monetary allocation decisions, participants reported their expectations of advice from both human and algorithmic advisors. Results showed that participants expected algorithmic advisors to recommend more equitable distributions, even when this resulted in smaller gains for the advisee. In contrast, they anticipated that human advisors would prioritize the advisee's welfare, consistent with established relational norms. In a second experiment we framed advisors as either "Institutional" or "Personal". Participants expected both human and algorithmic advisors to demonstrate greater impartiality when framed as Institutional. However, when framed as Personal, human advisors were expected to prioritize the advisee's interests, in line with personal relational norms, while algorithmic advisors remained still expected to be impartial. These findings contribute to relational models theory by demonstrating how advisor type interacts with relational context. They also extend literature on algorithm perception by revealing that people apply different normative standards to algorithmic versus human advisors in social dilemmas. These insights have implications for AI system design. The persistent perception of algorithms as impartial suggests that attempts to position AI as replacements for human relationships may encounter psychological barriers. Instead, algorithmic advisors may be most effective when their perceived impartiality is leveraged to complement human advisors in contexts where balancing relationships with objective guidance is valuable.