Student AI Use During the Research Process: Scaffold or Crutch?

Polk Library, as part of the University of Wisconsin Oshkosh and the Universities of Wisconsin system, recognizes that AI tools—such as ChatGPT, Gemini, Copilot, and Perplexity—are increasingly integrated into academic work. When used responsibly, transparently, and in accordance with University of Wisconsin’s Generative AI policy and guidelines, these tools can help develop critical thinking, creativity, and analytical skills.

We are committed to:

• Following UW guidelines on academic integrity, privacy, intellectual property, and acceptable use of information technology resources.

• Providing instruction and guidance on the ethical and effective use of artificial intelligence, including when to cite AI use and how to protect personal data and student work.

• Encouraging critical evaluation of AI-generated content to maintain accuracy, minimize bias, and preserve student agency.

• Helping students and faculty distinguish when AI acts as a scaffold for learning versus when it risks becoming a crutch that undermines skill development (Han).

• Ensuring that no confidential, proprietary, or FERPA-protected information is entered into public AI tools.

Our ultimate goal is to prepare students to use generative AI tools thoughtfully and in compliance with UW policy—maximizing their educational benefits while safeguarding academic integrity, privacy, and the values of scholarship.

Librarians at Polk Library work daily with students to refine research questions and identify credible, relevant sources. While traditional library tools—such as databases and catalogs—help explore topics and scholarly conversations, AI tools are far more generative and increasingly built into these same platforms.

To better support students, faculty, and writing courses, we examined recent scholarly literature to see how AI tools are being integrated into the research and writing process. The literature reviewed in this guide was identified using the tool Elicit, which searched over 126 million academic papers and retrieved 500 of the most relevant to our query:

”How does the integration of AI writing tools influence college students' creativity and

critical thinking during the process of composing academic research papers?”

The papers that were ultimately screened into the following reivew by Elicit met each of the following criteria: 

•    Population Type: Does the study exclusively involve college/university students (undergraduate or graduate level)?

•    AI Technology: Does the study examine AI-powered writing assistance tools (e.g., GPT-based tools, automated writing

     feedback systems, AI writing companions)?

•    Writing Context: Does the study focus on academic research paper writing (e.g., literature reviews, term papers, thesis work)?

•    Outcome Measures: Does the study measure creativity and/or critical thinking outcomes using validated assessment tools

      or well defined evaluation criteria?

•    Study Design: Is the study an empirical research design (experimental, quasi-experimental, controlled observational study,

      systematic review, or meta-analysis) with quantitative or qualitative data?

•    Tool Type: Does the writing tool studied include AI-powered features (not just basic word processing or grammar checking)?

•    Research Focus: Does the study examine educational impacts rather than just technical aspects of the AI tool?

AI Writing Tools Investigated

While the studies screened and included in the literature review also mentioned writing tools such as Grammarly and Quillbot, this report focuses more on tools related to the research process. The most heavily used tools in the literature included: ChatGPT, Gemini, Copilot, and Perplexity.

Terms and Concepts identified in the literature

🎨 Creativity

The literature identified several key elements of creativity in academic writing with AI tools:

• Ideation and brainstorming capabilities (Karanjakwut and Charunsri)

• Original content generation (Han)

• Development of novel perspectives (Vicente-Yagüe-Jara et al.)

• Authentic voice in writing (Wang)

🧠 Critical Thinking

Based on the research report, critical thinking was understood and measured through these key dimensions. Each study measured 

the dimensions somewhat differently as indicated in the examples below.

• Analytical skills

• • • Measured through ability to construct cohesive arguments and integrate evidence (Rizkiani et al.)
    • Assessed via interpretation, analysis, and inference capabilities (Shen and Chen)
    • Evaluated through students' ability to produce and explain novel arguments (Pryma et al.)

• • Metacognitive components

• • • Assessed through self-regulation and reflection on writing process (Shen and Chen)
    • Measured via students' ability to recognize and address limitations in their writing ("GPT3 and Artificial Intelligence in College Composition,")
    • Evaluated through decline in metacognitive engagement when over-relying on AI (Han)

• Evaluation skills

• • • Measured through students' ability to assess writing quality (Hikmah and Walida)
    • Assessed via critical evaluation of AI-generated content (Shen and Chen)
    • Evaluated through ability to identify different text genres and assess writing quality (Hikmah and Walida)

• Writing-specific elements

• • • Assessed through cohesion, coherence, and argumentation quality (Villacreses Sarzoza et al.)
    • Measured via organization and content development metrics (Zhang)
    • Evaluated through syntactic complexity and academic voice (Khampusaen)

• Independent thinking

• • • Assessed through students' ability to maintain autonomous writing abilities (Molinari and Molinari)
    • Measured via ability to critically engage with AI suggestions rather than passive acceptance (Wang)
    • Evaluated through preservation of student voice and originality (Bista and Bista)

📚 Impact on Creative and Cognitive Processes

💡 Ideation and Brainstorming

• AI-driven brainstorming enhanced creativity, increased idea fluency, and supported novel perspectives (Karanjakwut and Charunsri)

• Overreliance led to risks of homogenized or formulaic outputs (Sudrajad et al.)

• Repeated concerns noted regarding diminished student agency in idea development

✍️ Original Content Generation

• AI tools helped structure drafts and organize thoughts (Villacreses Sarzoza et al.)

• Heavy reliance reduced originality and authentic voice (Han et al.)

• Used as a supplement, AI exposed students to new rhetorical strategies and creative techniques (Wang)

🧠 Critical Thinking Development

🔎 Analytical Skills Enhancement

• Integration with explicit instruction improved analytical thinking and argument construction (Shen and Chen)

• Notable improvements in logical reasoning and evidence integration (Villacreses Sarzoza et al.)

• Passive use of AI sometimes undermined metacognitive engagement (Pryma et al.)

🧱 Argument Construction

• AI-supported writing enhanced argumentation skills when paired with critical revision strategies (Rizkiani et al.)

• Excessive or unguided use led to formulaic and shallow arguments (Han et al.)

✨ Student Agency and AI Tool Interaction

🖊️ Autonomy in the Writing Process

• Reflective, transparent AI use fostered student autonomy and self-regulation (To Hoai An et al.)

• Students using AI as a drafting aid—not a replacement—gained confidence and independence

⚖️ Balancing AI Assistance with Independent Thought

• Maintaining student voice while leveraging AI benefits was a central challenge (Molinari and Molinari)

• Instructional strategies emphasizing ethical awareness and critical thinking preserved originality (Bista and Bista et al.)

📚Key Recommendations for Implementing AI Tools in Academic Writing

Based on the research findings, several key recommendations emerge for thoughtful and effective integration of AI tools in academic writing and research contexts:

✅ Implement Structured Guidance

Use structured, guided approaches rather than open-ended implementation

• Explicit instruction and critical reflection enhance analytical skills and argument construction (Zamorano)

• Unguided or excessive use leads to formulaic arguments and reduced depth (Han)

🧠 Incorporate Reflective Practice

Include reflective components to promote critical engagement

• Reflective and transparent AI use fosters autonomy and self-regulation (To Hoai An)

• This helps maintain student voice while leveraging AI benefits (Villacreses Sarzoza)

🕒 Use Phased Implementation

Introduce AI tools gradually, in phases

• Positive outcomes come from systematic integration with clear strategies (Han)

• Helps prevent overreliance and promotes critical engagement (Rizkiani)

🚨 Monitor and Address Overreliance

Actively monitor student use to prevent dependency on AI tools

• Overreliance is a key challenge across multiple studies (Molinari and Molinari)

• Strategies should maintain independent thought and authentic voice (Wang)

🧰 Combine Multiple Tools Strategically

Use different AI tools for specific writing tasks when appropriate

• Successful cases involve using multiple tools with clear guidance (Karanjakwut and Charunsri)

• Define appropriate use for each tool to prevent misuse (Hutson)

🔍 Focus on Metacognitive Development

Emphasize metacognition and higher-order thinking

• Successful integrations boost critical reflection and self-awareness (Han)

• Passive use can undermine these essential academic skills

📏 Establish Clear Assessment Criteria

Develop rubrics and criteria specific to AI-assisted writing

• Clear evaluation frameworks maintain academic integrity (Rizkiani)

• Supports transparency and proper tool usage (Molinari and Molinari)

📚What markers or behaviors indicate that a student is going from Using an AI tool as support to over 

relying on the AI tool?

Based on the research report, several key markers indicate a shift from supportive AI tool use to overreliance:

Changes in Writing Behavior:

• Passive acceptance of AI suggestions without critical evaluation (Pryma)

• Difficulty explaining reasoning behind written arguments (Pryma)

• Reduced ability to produce novel arguments independently (Han)

• Loss of authentic narrative voice (Han)

Impact on Critical Thinking:

• Decline in metacognitive skills (15% decrease noted in one study) (Han)

• Reduced depth in analytical development (Pryma)

• Struggle with independent critical thinking (Rizkiani)

• Diminished metacognitive engagement (Zamorano)

Changes in Creative Output:

• Homogenization of writing styles (Sudrajad et al.)

• Formulaic or superficial content (Pryma)

• Loss of originality in arguments (Han)

• Reduced creative depth (Rizkiani)

Student Engagement Patterns:

• Heavy reliance on AI for idea generation rather than as a brainstorming aid (Karanjakwut and Charunsri)

• Reduced autonomous decision-making in the writing process (Molinari and Molinari)

• Tendency to use AI as a replacement rather than a supplement (Wang)

📚 It’s difficult to police the guidelines and require students to document AI use. There is often no way to tell. What is a more collaborative option?

The research suggests these collaborative approaches work better than policing:

Open Discussion Model:

• Create class-wide conversations about AI use cases (Wang)

• Share examples of both good and problematic AI assistance (“GPT-3 and Artificial Intelligence in College Composition”)

• Normalize AI as a learning tool rather than a cheating risk (Shen and Chen)

Group Learning Activities:

• Have students work together to evaluate AI outputs (Karanjakwut and Charunsri)

• Use peer review sessions that include discussion of AI contributions

• Create group projects where teams openly use and critique AI tools (Wang)

Shared Learning Environment:

• Make AI use part of in-class activities rather than hidden homework (Bista and Bista)

• Practice collaborative prompting and editing of AI responses (Karanjakwut and Charunsri)

• Build a classroom culture of “critical chatting” instead of “casual cheating” (Shen and Chen)

This approach shifts focus from catching misuse to building shared understanding of effective AI use.

📚 How can Professors and Librarians work with students to use AI for brainstorming and initial ideation without encouraging overreliance?

Use AI as a Conversation Partner:

• Start with your own initial ideas first (Karanjakwut and Charunsri)

• Ask AI to expand or challenge your thoughts (“GPT-3 and Artificial Intelligence in College Composition”)

• Use AI responses as prompts for further thinking, not final answers (Wang)

Build in Reflection Steps:

• Write down why you agree or disagree with AI suggestions

• Mix AI-generated ideas with your own original thoughts (Rizkiani)

• Use AI to explore different angles, then develop your favorite ones independently (Karanjakwut and Charunsri)

Set Clear Boundaries:

• Use AI early in brainstorming, not during final development (Zamorano)

• Treat AI as a “thought partner” rather than an answer machine (Wang)

• Always build on AI suggestions rather than using them as-is—spark thought, not replace thinking (Karanjakwut and Charunsri)

"Tools like ChatGPT, Gemini, and Microsoft Bing assist students by offering prompts, vocabulary options, and varied perspectives,

helping to overcome common brainstorming limitations and stimulate creative thought. Integrating AI tools into brainstorming

activities supports students in the initial stages of process writing, promoting diverse, structured ideas that can enhance the

overall quality of their writing (Karanjakwut and Charunsri)."

Table was adapted from the prompting and ideation scaffolding framework in (Karanjakwut and Charunsri). Specific student examples and prompts were developed by ChatGPT for instructional use.