Investigating student perceptions of creativity and generative ai in computational physics

This study analyzes interviews with six upper-division physics majors to reveal that while they view generative AI as a useful learning resource, they remain skeptical of its accuracy and primarily conceptualize their own creativity within the "mini-c" and "little-c" frameworks of the Four C Model.

The Gist

Imagine a classroom where students are learning to build complex machines (computational physics). Suddenly, a new, incredibly fast, but sometimes hallucinating robot assistant (Generative AI) walks in. It can write code, solve math problems, and draft essays in seconds.

This paper is like a detective story where two researchers from Oregon State University try to figure out: How do these student mechanics feel about their new robot helper? Do they think it's a genius partner, a dangerous crutch, or just a fancy search engine?

Here is the breakdown of their investigation, translated into everyday language.

1. The Setting: The "Physics Lab"

The researchers studied six senior physics students. These aren't beginners; they are like master mechanics who know how to program computers to simulate how the universe works. They were asked to write essays about two things:

• Creativity: What does it mean to be "creative" in physics?
• The Robot (Gen-AI): How does this AI tool fit into their creative process?

2. The Lens: The "Four C" Creativity Scale

To understand the students' answers, the researchers used a special map called the Four C Model. Think of creativity like a ladder with four rungs:

• Mini-C (The "Aha!" Moment): This is the creativity that happens inside your own head. It's when you finally understand a concept or solve a problem in a way that makes sense to you. It's personal and private.
• Little-C (The "Show and Tell"): This is when you share your idea with others, and it's recognized as new and useful by your friends or classmates.
• Pro-C (The Professional): This is for experts who make a living creating new things.
• Big-C (The Legend): This is for geniuses like Einstein or Picasso who change the world.

The Finding: The students mostly lived on the bottom two rungs (Mini-C and Little-C). They weren't trying to change the world yet; they were trying to understand the material and share solutions with their study group.

3. The Students' Verdict: The "Trust Issues"

The researchers found two main stories in the students' essays:

Story A: The "Personal Tutor" (Mini-C)

Most students saw Gen-AI as a personal tutor for their "Aha!" moments.

• The Analogy: Imagine you are trying to assemble a giant IKEA shelf, and you are stuck. You ask the robot assistant, "How do I do this?" It gives you a hint, a diagram, or a different way to look at the instructions.
• The Result: Students used AI to brainstorm ideas, outline their code, or explain confusing concepts. It helped them learn how to think.
• The Catch: They treated the robot like a skeptical friend. They would say, "Thanks for the idea, but I'm going to double-check your math because you might be lying." They knew the AI could be wrong, so they used it as a starting point, not the final answer.

Story B: The "Uninspired Copycat" (Little-C)

When it came to sharing their work with others (Little-C), the students were much less impressed.

• The Analogy: Imagine asking the robot to write a poem for your best friend's birthday. It writes something that is grammatically perfect but feels generic, boring, and lacks "soul."
• The Result: One student tried using AI to write an essay and found the ideas "unoriginal and uninspiring." They felt the AI couldn't truly create something new; it just rearranged old things.
• The Exception: One student (Robin) had a unique take. They thought the AI was actually creative because it could mix old "experiences" (data) to make something new. The researchers wondered: Is this student seeing the AI as a human, or is the AI just a mirror?

4. The Big Picture: What Does This Mean?

The paper concludes that these physics students are smart users. They aren't blindly trusting the robot, nor are they throwing it in the trash.

• They use it as a "Training Wheels" tool: They use AI to help them learn and get started (Mini-C), but they know they have to pedal the bike themselves to get anywhere.
• They are the "Editors": They see AI as a first draft generator that needs a human editor to fix the mistakes and add the "human touch."
• The Warning: The researchers worry that if students start thinking the AI is "creative" in the same way humans are, they might stop valuing their own unique problem-solving skills.

The Takeaway

Think of Gen-AI in physics class like a super-powered calculator.

• If you use it to do the thinking for you, you'll fail the test.
• If you use it to check your work, brainstorm ideas, or learn a new trick, it's a superpower.

The students in this study are learning to be the pilots, with the AI as their co-pilot. They know the co-pilot has a great database, but they know the co-pilot can sometimes fly into a cloud that isn't there. So, they keep their hands on the controls.

Technical Summary

Here is a detailed technical summary of the paper "Investigating student perceptions of creativity and generative AI in computational physics" by Pachi Her and Patti Hamerski.

1. Problem Statement

The integration of Generative Artificial Intelligence (Gen-AI) into education is rapid, yet research within the Physics Education Research (PER) community has largely focused on high school or lower-division undergraduate contexts. There is a significant gap in understanding how upper-division undergraduate physics majors perceive and utilize Gen-AI, particularly within computational physics.

Computational physics is inherently problem-solving and modeling-oriented, aligning closely with Gen-AI capabilities. However, the intersection of Gen-AI and the development of creativity in this specific domain remains unexplored. The authors aim to address the research question: How are students' conceptual framings of their creativity connected to their perceptions of gen-AI in a computational physics education context?

2. Methodology

The study employed a qualitative, interpretive approach to capture detailed student narratives regarding their attitudes and usage of Gen-AI.

• Participants: Six junior-year physics majors enrolled in the second term (PH 366) of a three-term computational physics sequence at Oregon State University. Participants had varying levels of prior programming experience (one with formal CS training, one with informal training, and all having completed the introductory course).
• Data Collection: Data was generated via narrative essays based on three specific prompts:
  1. Views and opinions on Gen-AI/ChatGPT.
  2. Definitions of creativity (and intuition) with examples from physics courses.
  3. The relationship between Gen-AI and creativity, and whether Gen-AI affects their creative process.
  • Note: Essays were chosen over interviews to allow students time for reflection and re-evaluation of their thoughts.
• Theoretical Framework: The study utilized the Four C Model of Creativity (Kaufman & Beghetto), which categorizes creativity into:
  • Mini-c: Novel, personally meaningful interpretations of experiences (learning process).
  • Little-c: Everyday creativity producing novel and useful products within a social context.
  • Pro-C: Professional-level creativity.
  • Big-C: Eminent, historically significant creativity.
  • Adaptation: The authors expanded the definitions of Mini-c and Little-c to explicitly incorporate the role of Gen-AI as a tool within the learning and social processes.
• Data Analysis: An iterative coding process was used. Initial coding applied standard Four C definitions. Subsequent rounds expanded these definitions to characterize student responses regarding Gen-AI, resulting in a revised codebook (Table 1 in the paper) that maps Gen-AI usage to specific creativity levels.

3. Key Contributions

• Contextual Gap Filling: This is one of the first studies to examine Gen-AI specifically within upper-division computational physics, a domain where AI's potential for code generation and debugging is highly relevant.
• Theoretical Application: The paper successfully adapts the Four C Model to the Gen-AI context, demonstrating how AI tools function differently across levels of creativity (e.g., as a brainstorming aid for Mini-c vs. a collaborative tool for Little-c).
• Nuanced Perception Analysis: Rather than a binary "AI is good/bad" conclusion, the study provides a granular view of how students distinguish between AI as a learning scaffold versus AI as a creative generator.

4. Key Results

The analysis of the six participants yielded two primary observations:

A. Conceptualization of Creativity (Mini-c and Little-c)

• Students predominantly framed their creativity within Mini-c and Little-c levels.
• Mini-c: Students viewed creativity as the process of finding personally meaningful solutions to problems, even if the solution wasn't globally novel. For example, one student defined creativity as creating a solution "new to us" rather than new to the world.
• Little-c: Creativity was also defined through social interaction, such as collaborating with peers, comparing different coding approaches to solve the same problem, and sharing novel solutions within the classroom community.
• Pro-C/Big-C: Very few responses reflected professional or eminent creativity, consistent with the students' status as early-career physicists.

B. Perceptions of Gen-AI and Creativity

• Gen-AI as a Mini-c Tool: Students widely accepted Gen-AI as a resource for independent learning tasks, such as brainstorming, outlining, debugging, and gathering information. They viewed it as a tool to facilitate the process of learning (Mini-c).
• Skepticism and Distrust: Despite using Gen-AI, students expressed significant skepticism regarding its accuracy and "creativity."
  • Reliability: Students noted that AI outputs often conflicted with their intuition or contained errors, requiring them to "take it with a grain of salt."
  • Originality: Students felt Gen-AI outputs were often "unoriginal and uninspiring." One student noted that while AI was useful for clarifying sentence structure (a mechanical task), it failed to provide genuine creative inspiration.
• Little-c Limitations: While students used Gen-AI for feedback (e.g., reviewing code), they were hesitant to rely on it for the core creative product, fearing a lack of novelty or social validity.
• Exception (Humanization): One student (Robin) exhibited a unique perspective, "humanizing" AI by suggesting it has "prior experiences" and can be creative. This suggests a potential for higher acceptance but also raises concerns about conflating human and machine contributions.

5. Significance and Implications

• Educational Strategy: The findings suggest that educators should position Gen-AI as a scaffold for the learning process (Mini-c) rather than a replacement for creative problem-solving. Students are already critically evaluating AI outputs, indicating a need for instruction on how to effectively verify and integrate AI tools.
• Critical Thinking: The observed skepticism is a positive sign of critical engagement. Students are not passively accepting AI; they are actively distinguishing between AI's utility in mechanical tasks (debugging, formatting) and its limitations in creative generation.
• Future Research: The study highlights the need for further investigation into:
  • The "humanization" of AI and its impact on student acceptance.
  • How students prompt Gen-AI and how prompt engineering efficacy correlates with their perception of creativity.
  • Larger-scale studies to generalize these findings beyond the small sample size.

In conclusion, the paper argues that while Gen-AI is a valuable tool for the process of learning physics (Mini-c), students currently view it with caution regarding its ability to generate genuine, novel, and socially validated creative products (Little-c/Pro-C).