Enabling self-identification in intelligent agent: insights from computational psychoanalysis

Lingyu Li, Chunbo LiarXiv 2024 · DOI · arXiv

Abstract

Building upon prior framework of computational Lacanian psychoanalysis with the theory of active inference, this paper aims to further explore the concept of self-identification and its potential applications. Beginning with two classic paradigms in psychology, mirror self-recognition and rubber hand illusion, we suggest that imaginary identification is characterized by an integrated body schema with minimal free energy. Next, we briefly survey three dimensions of symbolic identification (sociological, psychoanalytic, and linguistical) and corresponding active inference accounts. To provide intuition, we respectively employ a convolutional neural network (CNN) and a multi-layer perceptron (MLP) supervised by ChatGPT to showcase optimization of free energy during motor skill and language mastery underlying identification formation. We then introduce Lacan's Graph II of desire, unifying imaginary and symbolic identification, and propose an illustrative model called FreeAgent. In concluding remarks, we discuss some key issues in the potential of computational Lacanian psychoanalysis to advance mental health and artificial intelligence, including digital twin mind, large language models as avatars of the Lacanian Other, and the feasibility of human-level artificial general intelligence with self-awareness in the context of post-structuralism.

Introduction: The "I" in AI

Current Artificial Intelligence excels at pattern recognition and task execution but lacks a fundamental "sense of self." While we have sophisticated agents, we do not yet have subjects. For AI to achieve human-level General Intelligence (AGI), it requires a mechanism for self-identification—the process by which an entity differentiates itself from its environment and adopts a social identity. This paper bridges Jacques Lacan’s psychoanalytic theories with the Free Energy Principle (Active Inference) to model this transition.

Theoretical Framework: Imaginary and Symbolic Identification

Self-identification is explored through two primary lenses:

  1. Imaginary Identification (The Body): Rooted in the "Mirror Stage," this involves forming a unified body schema (the ego) by recognizing oneself from an external perspective.
  2. Symbolic Identification (Language): This involves entering the "Symbolic Order," where identity is defined through language and social interactions with the "Other."
Conceptualizing the Mirror Stage: (A) The physical setup for mirror self-recognition; (B) A state transition diagram representing the agent's internal model.

Methods: Simulating Self-Formation

The researchers utilized Active Inference, where systems survive by minimizing "free energy"—the discrepancy between internal models and external reality. Two simulations were conducted:

  • Motor Mastery: A Convolutional Neural Network (CNN) trained on a "stickman" dataset learned to recognize its own limb configurations, mimicking the Mirror Stage.
  • Language Acquisition: A Multi-Layer Perceptron (MLP) interacted with ChatGPT (acting as the Lacanian "Other"). ChatGPT interpreted the agent's random inputs as specific needs, forcing the agent to master language to minimize free energy.

Results and Discussion

The simulations demonstrated that self-identification can be modeled as an optimization process. In the motor task, as the CNN improved its recognition of limb angles, variational free energy decreased and "pleasure" (defined as the stabilization of certainty) emerged. In the language task, the agent successfully reduced free energy by aligning its vocalizations with the feedback from the "mOther" (ChatGPT).

Experimental results: (A) Correlation between predicted and actual limb angles; (B, C) Stabilization of measure and free energy; (D) The MLP-ChatGPT interaction loop; (E) Convergence of average loss and free energy components.

The proposed FreeAgent model unifies these processes using Lacan’s "Graph II of Desire." A provocative insight is that Large Language Models (LLMs) function as avatars of the "Other," representing the collective symbolic order that defines a user's identity.

Conclusion

This framework suggests that machine self-awareness is a social and linguistic construct rather than an inherent "soul." Future work may lead to "digital twin minds" capable of predicting mental health risks by simulating an individual's internal symbolic world.

Summary updated

Abstract

Building upon prior framework of computational Lacanian psychoanalysis with the theory of active inference, this paper aims to further explore the concept of self-identification and its potential applications. Beginning with two classic paradigms in psychology, mirror self-recognition and rubber hand illusion, we suggest that imaginary identification is characterized by an integrated body schema with minimal free energy. Next, we briefly survey three dimensions of symbolic identification (sociological, psychoanalytic, and linguistical) and corresponding active inference accounts. To provide intuition, we respectively employ a convolutional neural network (CNN) and a multi-layer perceptron (MLP) supervised by ChatGPT to showcase optimization of free energy during motor skill and language mastery underlying identification formation. We then introduce Lacan's Graph II of desire, unifying imaginary and symbolic identification, and propose an illustrative model called FreeAgent. In concluding remarks, we discuss some key issues in the potential of computational Lacanian psychoanalysis to advance mental health and artificial intelligence, including digital twin mind, large language models as avatars of the Lacanian Other, and the feasibility of human-level artificial general intelligence with self-awareness in the context of post-structuralism.
BibTeX
@misc{li2024enabling,
  author = {Lingyu Li and Chunbo Li},
  title = {{Enabling self-identification in intelligent agent: insights from computational psychoanalysis}},
  year = {2024},
  eprint = {2403.07664},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2403.07664},
}
Figure