Academic

From We to Me: Theory Informed Narrative Shift with Abductive Reasoning

arXiv:2603.03320v1 Announce Type: cross Abstract: Effective communication often relies on aligning a message with an audience's narrative and worldview. Narrative shift involves transforming text to reflect a different narrative framework while preserving its original core message--a task we demonstrate is significantly challenging for current Large Language Models (LLMs). To address this, we propose a neurosymbolic approach grounded in social science theory and abductive reasoning. Our method automatically extracts rules to abduce the specific story elements needed to guide an LLM through a consistent and targeted narrative transformation. Across multiple LLMs, abduction-guided transformed stories shifted the narrative while maintaining the fidelity with the original story. For example, with GPT-4o we outperform the zero-shot LLM baseline by 55.88% for collectivistic to individualistic narrative shift while maintaining superior semantic similarity with the original stories (40.4% imp

J
Jaikrishna Manojkumar Patil, Divyagna Bavikadi, Kaustuv Mukherji, Ashby Steward-Nolan, Peggy-Jean Allin, Tumininu Awonuga, Joshua Garland, Paulo Shakarian · · 1 min read · 2 views
#cs.CL #cs.AI

arXiv:2603.03320v1 Announce Type: cross Abstract: Effective communication often relies on aligning a message with an audience's narrative and worldview. Narrative shift involves transforming text to reflect a different narrative framework while preserving its original core message--a task we demonstrate is significantly challenging for current Large Language Models (LLMs). To address this, we propose a neurosymbolic approach grounded in social science theory and abductive reasoning. Our method automatically extracts rules to abduce the specific story elements needed to guide an LLM through a consistent and targeted narrative transformation. Across multiple LLMs, abduction-guided transformed stories shifted the narrative while maintaining the fidelity with the original story. For example, with GPT-4o we outperform the zero-shot LLM baseline by 55.88% for collectivistic to individualistic narrative shift while maintaining superior semantic similarity with the original stories (40.4% improvement in KL divergence). For individualistic to collectivistic transformation, we achieve comparable improvements. We show similar performance across both directions for Llama-4, and Grok-4 and competitive performance for Deepseek-R1.

Executive Summary

This study proposes a neurosymbolic approach to facilitate narrative shift in Large Language Models (LLMs) by leveraging abductive reasoning and social science theory. The authors demonstrate the effectiveness of their method in transforming stories to align with different narrative frameworks while preserving the original core message. The results show significant improvements in narrative shift and semantic similarity across multiple LLMs. The study contributes to the development of more sophisticated LLMs that can adapt to diverse audiences and narrative frameworks. The findings have implications for various applications, including content creation, marketing, and education.

Key Points

  • ▸ The study proposes a neurosymbolic approach to facilitate narrative shift in LLMs using abductive reasoning and social science theory.
  • ▸ The authors demonstrate the effectiveness of their method in transforming stories to align with different narrative frameworks.
  • ▸ The results show significant improvements in narrative shift and semantic similarity across multiple LLMs.

Merits

Strength in Theoretical Foundation

The study's use of abductive reasoning and social science theory provides a robust theoretical foundation for the proposed neurosymbolic approach, ensuring a more comprehensive understanding of narrative shift.

Improved Narrative Shift and Semantic Similarity

The study demonstrates significant improvements in narrative shift and semantic similarity across multiple LLMs, highlighting the effectiveness of the proposed method.

Demerits

Limited Generalizability

The study's results are based on a limited set of LLMs and narrative frameworks, which may limit the generalizability of the findings to other contexts.

Dependence on High-Performance LLMs

The study's results rely on the performance of high-performance LLMs, which may not be representative of all LLMs, particularly those with limited capabilities.

Expert Commentary

This study represents a significant contribution to the field of natural language processing by proposing a neurosymbolic approach to facilitate narrative shift in LLMs. The use of abductive reasoning and social science theory provides a robust theoretical foundation for the method, ensuring a more comprehensive understanding of narrative shift. While the study's results are promising, further research is needed to generalize the findings to other contexts and LLMs. The implications of the study are far-reaching, particularly in areas such as content creation, marketing, and education. The proposed approach can be applied to develop more effective communication strategies and has the potential to revolutionize the way we interact with LLMs.

Recommendations

  • ✓ Future studies should investigate the generalizability of the proposed method to other LLMs and narrative frameworks.
  • ✓ The study's findings should be applied to real-world applications, such as content creation and marketing, to develop more effective communication strategies.

Sources

Related Articles