Mapping, reconstruction, and emergence represent pathways for theoretical construction of AI psychology, which attempts to empower human minds. Photo: TUCHONG

The theoretical construction of artificial intelligence (AI) psychology seeks to reconceptualize the structural relationships among mind, intelligence, and society, and to address new psychological phenomena emerging in the intelligent era. As intelligent systems become integrated into psychological processes, the structure, functions, and boundaries of both the human mind and the social mind must be redefined. Accordingly, gaining insight into intelligent systems, human–machine symbiosis, and even a “community of minds” becomes a central objective of theoretical innovation in psychology in the intelligent era.

“Mapping,” “reconstruction,” and “emergence” represent three primary pathways for the current theoretical construction of AI psychology. These approaches not only reflect the staged characteristics of disciplinary development, but also suggest an upward, iterative trajectory in the evolution of AI psychology theory, collectively pointing toward the discipline’s mission of empowering the human mind and allowing it to flourish.

Mapping: AI-based encoding of traditional psychological theories

Mapping—a process involving translating the core concepts, mechanistic models, and empirical regularities of traditional psychology into computational representations that AI systems can operate on and implementing them through algorithmic models—represents the starting point for the theoretical construction of AI psychology. In this process, AI is not merely an “auxiliary tool” for research, but is gradually becoming a new “language of expression” for psychology. In the past, psychological processes were described through questionnaires and experimental tasks; today, deep neural networks, vector semantics, and multimodal modeling form a new “psychological language system.” As a result, cognitive mechanisms such as attention and memory can be mapped onto network hierarchies and parameter updates; dimensional models of emotion can be identified and quantified through multimodal AI; and the prediction of psychological disorders in clinical contexts can be reproducibly simulated within algorithmic frameworks.

This process of “translation–encoding” can be advanced in three steps. The first is theoretical extraction, which involves distilling computable variables, structures, and constraints from psychological theories while clarifying boundaries and observable indicators. The second is algorithmic mapping, in which model structures analogous to psychological mechanisms are constructed to achieve parametric representation. The third is system validation, where the explanatory validity of theories is tested through model execution and comparative experiments. To ensure that model architectures correspond closely to psychological mechanisms, the authors advocate “Model Mechanism Mapping” as a core principle of cognitive modeling, aimed at ensuring mechanistic equivalence in the transformation from theory to algorithm.

The mapping approach significantly enhances the theoretical gains of psychology in areas such as cognition, emotion, and personality. Cognitive psychology, for instance, reconstructs concept learning through Bayesian inference models, clarifying priors and evidence updating in category formation. Representative cognitive architecture theories in this field reproduce human thinking processes through computational systems and validate psychological mechanisms via structural mapping. The theory of heuristic algorithms suggests that AI algorithms can map human decision-making heuristics, extracting efficient computational rules from bounded rationality. In emotion psychology, affective computing methods have enabled computable representations of basic emotion models—such as those proposed by Paul Ekman—from recognition to regulation. In personality psychology, large language models are used to represent personality dimensions, supporting interpretable inferences about personality tendencies and interaction styles. Mapping enables traditional psychological theories to be “reborn in intelligence,” granting psychological concepts computational representations. Psychology is thus no longer confined to an “explanatory science,” but is evolving into an “intelligent science” that integrates both generative and engineering-oriented approaches.

Reconstruction: AI-driven generative transformation of psychological theories

Reconstruction—the intermediate stage in the theoretical innovation of AI psychology—marks a phase in which AI, through its algorithmic mechanisms and learning principles, feeds back into psychological theory, promoting a shift from static observation to dynamic modeling. In this process, theoretical production evolves from being purely “hypothesis-driven” to “evidence-driven,” advancing psychology from abstract descriptions of human thinking toward the algorithmic reproduction of mental processes. In terms of research pathways, reconstruction follows a generative logic of “data–algorithm–theory.” Through computational modeling of large-scale and multimodal data, AI uncovers patterns in emerging psychological phenomena and provides evidence for theory building. Psychologists then align data models with psychological variables to propose mechanistic hypotheses, which are subsequently tested through experimental research, enabling the co-evolution of theory and algorithms. For instance, generative psychology takes generative models as its core, proposing that psychological laws emerge through the self-organizing processes of deep learning models, emphasizing the operational nature of theory.

Computational emotion dynamics theory, by integrating reinforcement learning with affective weighting, reveals the dynamic regulatory role of emotion in human decision-making. As a result, theories no longer rely solely on researchers’ a priori intuitions, but instead emerge naturally from data structures and algorithmic discoveries. This transformation significantly enhances the predictive power, reproducibility, and interpretability of psychological theories, bringing psychological research closer to the standards of executable models in the natural sciences. At the same time, the deep integration of AI prompts critical reflection on research methodologies within the discipline. On the one hand, psychological research must ensure interpretability and alignment between models and psychological mechanisms, avoiding the pitfall of “substituting mechanisms with effect.” On the other hand, issues such as algorithmic bias, ethical risks, and the reproduction of social inequalities serve as reminders that psychologists must uphold value-oriented principles in its pursuit of theoretical innovation.

Emergence: Generation of psychological theories for novel AI-induced phenomena

Emergence—the frontier stage in the theoretical construction of AI psychology—focuses on how AI generates unprecedented forms of psychological reality. When intelligent agents exhibit anthropomorphic characteristics in interaction, expression, and decision-making—and humans begin to form quasi-social relationships, emotional interactions, and identity extensions with them—the object of study expands from the “human mind” to the “mind of intelligence,” and from a single subject to a composite system of human–AI collaborative cognition. This shift raises a series of critical questions: Can intelligent agents possess forms of autonomy describable by psychology? Do human trust, empathy, or attachment toward AI follow entirely new psychological mechanisms? Can AI-generated language and emotion constitute new forms of mental expression and social norms? Recent explorations reveal three clear lines of development. First, attachment is being reconfigured: AI companions, virtual teachers, and intelligent assistants are transforming both the objects and pathways of emotional bonding, extending attachment theory from interpersonal to human–AI contexts. Next, empathy representation is being redefined: Generative large models, despite lacking subjective experience, can produce emotionally responsive outputs, provoking methodological debates about the “authenticity of empathy” and the distinction between emotional appearance and connotation. Finally, collaborative cognition is becoming systematized: The perspective of “intelligent mental systems” treats humans and AI as a coordinated ensemble capable of division of labor and mutual verification, prompting a reexamination of intention inference, role allocation, and boundaries of responsibility.

In terms of research pathways, emergence follows a framework of phenomenon identification, mechanism hypothesis, cross-context validation, and ethical reflection. Researchers first identify stable patterns of human–AI interaction in real-world contexts, then translate these into computable variables and mechanistic models, and finally test their external validity through experiments and interventions while simultaneously evaluating ethical dimensions such as bias, equity, privacy, and safety. Representative topics include attachment to social robots, simulated conscious responses from large language models, and the experience of multiple selves in virtual personalities. Together, these inquiries advance psychology toward explaining how humans and intelligent agents co-construct meaning.

Emergence compels psychology to move beyond its traditional categories, constructing new concepts—such as algorithmic emotion, virtual empathy, and intelligent agency—and forming theories that are operational, falsifiable, and governable. Psychology therefore does not passively adapt to the wave of intelligent technologies, but actively redefines its disciplinary boundaries through reciprocal interaction with AI, inaugurating a new phase of theoretical self-awareness oriented toward human–AI symbiosis. The Chinese idiom “xin hua nu fang”—literally “heart flowers burst into bloom”—conveys a state of extreme joy or elation. The phrase metaphorically captures the mapping of the human mind, the reconstruction through intelligence, and the emergence of new forms of mentality. Mapping, reconstruction, and emergence are not only progressive academic pathways, but also an interconnected theoretical system. Together, they constitute a “theoretical evolution triangle” in AI psychology—rooted in the human mind and illuminated by intelligence, pointing toward a future of human–AI symbiosis.

Mapping carries forward tradition, enabling psychological theories to be re-represented within intelligent environments. Reconstruction bridges conceptual and methodological gaps, realizing the generative transformation of psychological theories. Emergence opens new frontiers, giving rise to novel psychological theories suited to the era of human–AI coexistence. The integration of these three stages marks a shift in the theoretical construction of AI psychology—from “describing the mind” to “generating the mind,” and from “understanding humans” to “integrating intelligence.” In doing so, it contributes evolving theoretical insights toward the development of an independent knowledge system for Chinese psychology.

Ni Shiguang is an associate professor from the Shenzhen International Graduate School at Tsinghua University; Peng Kaiping is a professor from the Department of Psychological and Cognitive Sciences at Tsinghua University.