Pragmatic Inference for Moral Reasoning Acquisition: Generalization via Metapragmatic Links

The paper presents a pragmatic inference approach for enhancing moral reasoning in large language models (LLMs) by utilizing metapragmatic links and Moral Foundations Theory. This method was tested across three distinct moral reasoning tasks, demonstrating significant improvements in generalization capabilities for LLMs. The findings suggest that integrating pragmatic inference can advance the development of LLMs in understanding and applying moral reasoning, which is crucial for practitioners aiming to build more context-aware AI systems.