From Crafting Text to Crafting Thought: Grounding AI Writing Support to Writing Center Pedagogy

Paper Title

From Crafting Text to Crafting Thought: Grounding AI Writing Support to Writing Center Pedagogy

Publication Info

• Topic area: AI-driven writing support systems informed by educational pedagogy.
• Keywords: AI writing tools, writing center pedagogy, non-directive feedback, writer agency, process-oriented learning, collaborative writing, large language models, educational technology, human-computer interaction, writing support systems.

Background and Problem

• Problem / challenge: Current AI writing tools often prioritize text generation and task efficiency, which can undermine writer agency, critical thinking, and ownership of the writing process. These tools risk becoming "fix-it shops" that displace human involvement in writing.
• Significance: Addressing these challenges is crucial to ensure AI tools support educational goals, such as fostering independent writing skills, preserving student voices, and aligning with pedagogical principles.
• Motivation and related work: Writing centers have historically transitioned from error correction to process-oriented, writer-centered support. Existing AI tools often focus on directive feedback or text generation, which can diminish creativity and ownership. This paper seeks to bridge writing center pedagogy with AI design to address these gaps.

Solution

• Proposed approach: Writor, an AI writing support system that provides non-directive, process-oriented, and collaborative feedback grounded in writing center pedagogy.
• Novelty:
  1. Introduces writing center pedagogy as a theoretical foundation for AI writing tools.
  2. Develops seven design guidelines for AI systems based on interviews with writing tutors.
  3. Implements these guidelines in Writor, emphasizing non-directive feedback and writer agency.
  4. Validates the approach through expert reviews with 30 domain experts, assessing pedagogical soundness and integration potential.
• Procedure and key techniques:
  1. Conducted interviews with 10 writing tutors to identify key pedagogical strategies.
  2. Developed Writor, which includes features like goal setting, reader-perspective feedback, and dialogic interaction.
  3. Evaluated Writor through expert reviews, including surveys and interviews with writing instructors, tutors, and AI researchers.

Results

• Concrete findings:
  • Writor’s feedback was rated as well-balanced, positive, and aligned with writing center pedagogy.
  • Internal audits showed high success rates for goal relevance (100%), critique accuracy (98.33%), and non-directive feedback (100%).
  • Compared to baseline AI tools, Writor produced 46.9% longer and 74.3% more specific critiques.
• Advantage over baselines:
  • Experts perceived Writor as superior to generic AI tools like ChatGPT in preserving writer agency and fostering critical thinking.
  • Writor avoided generating copyable text, addressing concerns about plagiarism and over-reliance on AI.
• Experiments / evaluation:
  • Expert review with 30 participants (writing instructors, tutors, AI researchers) using structured surveys and follow-up interviews.
  • Participants engaged with Writor using sample texts or their own writing and provided feedback on its pedagogical alignment and usability.
• Limitations and future work:
  • Focused on expert feedback rather than student experiences; future studies should include student usability and learning outcomes.
  • Feedback specificity varied based on user needs; future systems could incorporate adaptive granularity.
  • Current implementation relies on prompt-based LLMs, which may limit consistency; finetuning on curated datasets could enhance performance.

Summary

This paper introduces Writor, an AI writing support tool designed to align with writing center pedagogy by providing non-directive, process-oriented, and collaborative feedback. Through interviews with writing tutors, the authors developed seven design guidelines, which were operationalized in Writor. Expert reviews demonstrated that Writor effectively supports writer agency and critical thinking, outperforming generic AI tools in pedagogical alignment. The study highlights the potential of AI tools to complement human writing support, offering design insights for trust-building, adaptive feedback, and integration into educational ecosystems.