Narrix: Remixing Narrative Strategies from Examples for Story Writing

Paper Title

Narrix: Remixing Narrative Strategies from Examples for Story Writing

Publication Info

• Topic area: AI-assisted creative writing tools for novice writers
• Keywords: narrative strategies, story writing, creative support tools, AI-assisted writing, cognitive apprenticeship, storytelling, remixing, narrative arcs, generative AI, user study

Background and Problem

• Problem / challenge: Novice writers struggle to identify and reuse narrative strategies effectively, as these techniques are often tacit and subtly embedded in text. Existing systems focus on genre norms or linguistic patterns but lack support for recognizing and repurposing deeper creative strategies in storytelling.
• Significance: Addressing this gap can help novice writers improve their storytelling skills, enabling them to create compelling narratives and develop long-term creative abilities.
• Motivation and related work: Prior tools like IntroAssist and CorpusStudio surface genre conventions but do not focus on narrative strategies. Cognitive apprenticeship theory highlights the importance of learning through examples, yet existing systems fail to scaffold the interpretation and application of narrative strategies in creative writing.

Solution

• Proposed approach: Narrix, an interactive AI-assisted writing tool that helps novice writers discover, interpret, and remix narrative strategies from example stories into their own writing.
• Novelty:
  1. Surfacing narrative strategies through color-coded annotations, explanations, and lexical cues.
  2. Interactive story arc visualization to explore strategies based on emotional shifts and turning points.
  3. Track-based remixing interface to apply strategies across creative dimensions like plot, character, and linguistic style.
• Procedure and key techniques:
  • Example stories are segmented into blocks, and narrative strategies are extracted using GPT-based prompts.
  • Strategies are annotated with explanations and lexical cues, categorized into creative dimensions, and visualized on story arcs.
  • Users drag and drop strategies onto multi-dimensional tracks to guide AI-assisted revisions or story continuations.
  • Reflective comparison views contrast original examples with user revisions, highlighting differences and similarities.

Results

• Concrete findings:
  • Participants recalled significantly more narrative strategies (mean: 4.08 vs. 1.00) and demonstrated higher understanding quality (mean: 2.75 vs. 0.83) using Narrix compared to a baseline chat-based system.
  • Stories produced with Narrix were rated significantly higher in quality (74.3% preference) by an expert-trained model.
  • Participants applied more strategies (mean: 8.00 vs. 3.92) and reported greater confidence and satisfaction in adapting strategies.
• Advantage over baselines:
  • Narrix outperformed the baseline in usability (SUS score: 82.64 vs. 59.03), creativity support (e.g., exploration: 6.08 vs. 3.42), and perceived AI collaboration (e.g., transparency: 6.00 vs. 2.67).
  • Participants found Narrix more effective in helping them identify, understand, and remix narrative strategies.
• Experiments / evaluation:
  • Within-subjects study with 12 novice writers (ages 23–28, non-native English speakers).
  • Tasks involved writing micro-stories using Narrix and a baseline system, with counterbalanced conditions.
  • Measures included surveys (NASA-TLX, CSI, AI experience), recall tasks, usage logs, and interviews.
• Limitations and future work:
  • Small sample size (N=12) and culturally homogeneous example stories limit generalizability.
  • LLM errors in strategy extraction and explanation persist despite mitigations.
  • Limited support for global-level narrative coherence and parallel strategy comparisons.
  • Future work includes longitudinal studies, diverse participant recruitment, and enhanced mechanisms for long-form writing.

Summary

Narrix is an AI-assisted writing tool designed to help novice writers learn and apply narrative strategies from examples. Through features like strategy annotations, story arc visualizations, and track-based remixing, Narrix enables users to explore, understand, and adapt storytelling techniques in their drafts. A user study demonstrated significant improvements in strategy retention, application, and story quality compared to a baseline chat-based system. While promising, the tool faces challenges in scaling to long-form writing and addressing LLM errors. Future iterations aim to enhance global coherence, support diverse narratives, and foster long-term creative growth.