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

Yijun Liu, Tal AugustNAACL 2025

TL;DR

This paper explores how to design AI writing tools that mimic the supportive, writer-centered, and collaborative approach of university writing centers. By interviewing writing tutors and synthesizing their practices, the authors developed design guidelines and a prototype tool called "Writor" to foster deeper student engagement with their writing processes, rather than just generating text.

What Problem Does This Paper Address?

The rapid advancement of AI, particularly large language models (LLMs), has led to writing tools that can now collaborate and even co-create text with users. While this offers new possibilities, it also raises concerns, especially for student writers who are still developing their skills. There's a risk that fluent AI-generated text could diminish student ownership of their work, encourage plagiarism, or dilute the learning process. This paper addresses the challenge of designing AI writing support that avoids these pitfalls by drawing inspiration from the pedagogical principles of university writing centers, which have historically shifted from simply "fixing" student papers to fostering a more holistic, writer-focused approach.

Methods

The authors conducted a qualitative study involving semi-structured interviews with 10 writing tutors from three US universities. The interviews aimed to understand their strategies for supporting students and their perspectives on how AI could integrate into writing center practices. A reflexive thematic analysis was used to identify recurring themes and strategies. These insights were then synthesized into six core design principles for AI writing support tools. Based on these principles, the authors developed a prototype AI writing tool called "Writor," which operates in two stages: preparation and goal-setting, followed by an editing stage with AI-generated feedback.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Writing Centers as a Model for AI: The paper argues that writing centers offer a valuable pedagogical framework for designing AI writing tools. Their evolution from "fix-up shops" to collaborative, process-oriented spaces provides a blueprint for creating AI that supports, rather than replaces, the writer.
  • Beyond Text Generation: The core idea is to move AI writing support from mere text generation to facilitating the thought process behind writing. This involves helping students reflect, plan, and revise, mirroring the dialogic and scaffolding approaches used by human tutors.
  • Writer-Centered AI: AI tools should prioritize the student's voice, goals, and ownership. This means providing feedback that guides and prompts, rather than dictates, and ensuring the student remains in control of their writing.
  • Process Over Product: The focus should be on developing the student's writing skills and understanding of the writing process, not just on producing a polished final product. AI can support this by offering feedback on higher-order concerns like argumentation and organization, and by encouraging self-reflection.
  • The Importance of Dialogue: Collaborative and dialogic interactions are crucial. AI tools should be designed to engage writers in conversations about their work, prompting them to articulate their ideas and justify their choices.

Key Results

The interview study identified three key themes in writing center pedagogy: writer-centered, process-oriented, and collaborative. These themes translated into six design guidelines for AI writing support:

  • G1: Empathetic and confidence-building feedback.
  • G2: Non-directive feedback that helps writers refine their work.
  • G3: Support for writing process development through examples, analogies, and reader-perspective feedback.
  • G4: Assistance with understanding writing prompts and goals.
  • G5: Dialogic feedback that encourages conversation.
  • G6: Responsiveness to what the writer wants to work on.

The prototype tool, "Writor," incorporates these guidelines by offering features like goal setting, interactive highlighting of feedback, and a feedback sidebar that provides non-directive suggestions.

Limitations and Future Work

  • Limitation: The study primarily focused on the perspectives of writing tutors. The authors acknowledge that student perspectives are also crucial for a comprehensive understanding of effective writing support and plan to incorporate them in future work.
  • Future Work:
    • Evaluation of Writor: The authors plan to conduct controlled experimental studies to compare Writor with a basic chat interface, evaluating its effectiveness in improving writing and revision processes.
  • Integration with Writing Centers: They aim to explore how Writor can be integrated synergistically with existing writing center services, potentially offering it as a first step for students seeking support.
  • Refinements to Writor: Future development includes adding import/export functionalities for easier workflow integration and further evaluating the quality and specificity of the AI-generated feedback.
  • In-situ Signals: Brainstorming a more free-form deployment where students can provide lightweight, in-situ feedback on AI suggestions during assignments.
  • Tutor Evaluation of AI Critiques: Evaluating the AI's generated critiques and praises in isolation with writing center tutors to assess their accuracy, specificity, and overall quality.

Summary updated

BibTeX
@inproceedings{liu2025from,
  author = {Yijun Liu and Tal August},
  title = {{From Crafting Text to Crafting Thought: Grounding Intelligent Writing Support to Writing Center Pedagogy}},
  booktitle = {Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2025},
}
Figure