Projection and Transference in Human–ai Interaction: Relational Schemas, Anthropomorphism, and Algorithmic Mirroring
Year: 2026
Type: preprint
Abstract: The growing integration of conversational artificial intelligence into everyday communication raises fundamental questions concerning how human relational expectations, previous interpersonal experiences, and psychological needs influence interactions with artificial systems. Generative AI can produce responsive, personalized, and emotionally appropriate language, potentially encouraging users to experience these systems as socially meaningful partners rather than merely computational tools. However, the psychological processes through which such relationships emerge remain theoretically heterogeneous. This paper examines projection, transference, anthropomorphism, relational schemas, and algorithmic mirroring as distinguishable but potentially interacting processes in human–AI interaction. Drawing upon psychoanalytic theory, social-cognitive models of transference, attachment research, anthropomorphism theory, human–computer interaction, and empirical studies of AI companionship, the paper develops the Relational Schema–Mirroring–Revision Model (RSMR). The proposed model describes four interconnected processes: activation of pre-existing human relational schemas, algorithmic adaptation to user behavior, psychological attribution of meaning to AI responses, and subsequent reinforcement or revision of relational expectations. A central distinction is established between human psychological experience and computational responsiveness. Although AI systems may adapt their language and behavior to user characteristics, such adaptation does not establish subjective emotional reciprocity, psychological attachment, or human-like countertransference. The framework differentiates anthropomorphic attribution from projection, transference, attachment, and parasocial engagement. It further distinguishes stylistic accommodation, emotional validation, perspective mirroring, and sycophantic reinforcement as forms of algorithmic response adaptation with potentially different psychological consequences. Eight research hypotheses are proposed concerning schema activation, perceived understanding, interaction history, algorithmic mirroring, expectancy reinforcement, relational rupture, individual differences, and the possible transfer of human–AI interaction patterns to human relationships. An empirical program is outlined involving controlled chatbot conditions, preinteraction relational-schema assessment, longitudinal measurement, experimentally manipulated personalization, and causal evaluation of relational feedback. The proposed framework does not assume that artificial intelligence possesses subjective experience or participates in psychologically reciprocal relationships equivalent to human interpersonal relations. Its principal theoretical contribution is a differentiated model of how human relational cognition and adaptive AI behavior may jointly shape the interpretation, development, reinforcement, and revision of human–AI relationships.
Author: Shaban, Roman
Language: English
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7237366271
Cites:
Cited by:
Related to:
