Memory Continuity as a Foundation of User Trust in Emotional AI Avatars: A Literature Review

Abstract

Emotional AI avatars are increasingly used for companionship and emotional support, where users may expect continuity across interactions, including being remembered over time. This makes memory continuity relevant to user trust. Prior research shows that people respond to digital systems in social ways, that repeated interaction can support relationship-like expectations, and that trust in emotional support AI involves more than technical performance alone. Recent work also suggests that memory can influence how users judge AI systems, including perceived intelligence, likeability, and trustworthiness. However, existing literature has not clearly explained how memory continuity functions as a condition of user trust in emotional AI avatars. This review addresses this gap by synthesizing research on social responses to AI, long-term human-AI relationships, emotional support AI, and AI memory through a user trust lens. It argues that continuity in remembered context helps shape whether interaction feels stable, coherent, and personally responsive over time. When continuity is maintained, trust may be supported; when it breaks, trust may weaken. Taken together, the reviewed literature suggests a trust-centred way to understand how continuity and forgetting shape emotionally supportive AI interaction.