Causal Autoencoder-like Generation of Feedback Fuzzy Cognitive Maps with an LLM Agent

2025-09-29

AI Summary

Traditional autoencoders (AEs) yield uninterpretable latent representations and fail to preserve underlying causal structures. Method: We propose a causal autoencoder architecture leveraging large language model (LLM) agents to establish a bidirectional, interpretable mapping between feedback fuzzy cognitive maps (FCMs) and natural language text. An LLM agent encodes FCMs into causally rich, human-readable textual descriptions and reconstructs FCMs from such text—achieving identity-mapping approximation without input-output supervision. Causal edge weights are explicitly modeled and weak connections pruned to enhance semantic naturalness and structural fidelity of encoded text. Contribution/Results: This work pioneers end-to-end, causal-aware, human-interpretable bidirectional translation between FCMs and natural language—replacing opaque latent variables with semantically grounded, linguistically transparent representations. It preserves core causal topology while enabling direct, intuitive semantic understanding.

Technology Category

• Machine Learning: Causal Learning
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Agentic search

Abstract

A large language model (LLM) can map a feedback causal fuzzy cognitive map (FCM) into text and then reconstruct the FCM from the text. This explainable AI system approximates an identity map from the FCM to itself and resembles the operation of an autoencoder (AE). Both the encoder and the decoder explain their decisions in contrast to black-box AEs. Humans can read and interpret the encoded text in contrast to the hidden variables and synaptic webs in AEs. The LLM agent approximates the identity map through a sequence of system instructions that does not compare the output to the input. The reconstruction is lossy because it removes weak causal edges or rules while it preserves strong causal edges. The encoder preserves the strong causal edges even when it trades off some details about the FCM to make the text sound more natural.

Problem

• Generating text from feedback fuzzy cognitive maps
• Reconstructing cognitive maps from generated text
• Creating explainable AI systems using autoencoder principles

Innovation

• LLM maps feedback causal fuzzy cognitive maps to text
• Reconstructs FCM from text with lossy autoencoder-like process
• Preserves strong causal edges while removing weak ones