Physics Education under the Application of Artificial Intelligence: Bibliometric Analysis Based on Web of Science Core Library (2021-2025)

This bibliometric analysis of 138 Web of Science publications from 2021 to 2025 reveals that physics education is undergoing an AI-driven transformation characterized by exponential growth since 2023, a shift from traditional data analysis to generative AI and interdisciplinary applications, and a future trajectory focused on adaptive learning ecosystems and ethical AI integration.

The Gist

Here is an explanation of the paper, translated from academic jargon into everyday language with some creative metaphors.

🚀 The Big Picture: Physics Class Gets a Superpower Upgrade

Imagine physics education as a classic, old-school library. For decades, students learned by reading dusty books, doing experiments with beakers, and solving equations on chalkboards. It worked, but it was slow and rigid.

Now, Artificial Intelligence (AI) has walked into that library and turned it into a futuristic, interactive video game. This paper is like a map of the construction site, showing us exactly how this transformation is happening, who is building it, and what the finished product might look like.

The authors looked at 138 research papers published between 2021 and 2025 to see what's going on. Here is what they found:

---

📈 The Explosion: From a Spark to a Firework

The Trend:
Until recently, people were just dipping their toes in the water. But starting in 2023, the number of studies exploded.

• The Metaphor: Think of it like a firework. For a few years, there were just a few sparks (a few papers). Then, in 2023, the fuse lit up, and by 2025, the sky is full of colorful explosions.
• Who is lighting the fuse? The United States is holding the biggest sparkler, followed closely by China and Germany. They are the main architects of this new era.

---

🧩 The Five Main Ways AI is Changing Physics Class

The researchers grouped all the new ideas into five "clusters" or themes. Here is what they are:

1. The "Super-Tutor" (Generative AI)

• What it is: Tools like ChatGPT are being tested as virtual physics teachers.
• The Reality Check: Researchers asked, "Can an AI pass a physics exam?" The answer is a shaky "Yes, but..."
• The Metaphor: Imagine an AI student who has memorized the entire textbook but doesn't actually understand the concepts. It can solve easy problems perfectly, but if you ask it a tricky question about how gravity works, it might confidently give you a wrong answer that sounds like nonsense (a "hallucination").
• The Shift: Teachers are now learning how to write "prompts" (instructions) to turn this AI from a cheat sheet into a helpful study buddy.

2. Teaching AI to Solve Physics Problems (PINNs)

• What it is: Instead of just using AI to teach students, scientists are teaching AI to solve complex physics equations.
• The Metaphor: Traditionally, solving a physics problem is like trying to find a needle in a haystack using a metal detector. Now, we are teaching the AI to become the needle. We are embedding the laws of physics directly into the AI's brain so it can solve problems in fluid dynamics or quantum mechanics that were previously too hard for computers.
• The Result: Physics classes are starting to teach students how to program these "smart" AI tools, not just how to use a calculator.

3. The Doctor-Physicist Hybrid (Medical Physics)

• What it is: Using AI to teach physics specifically for doctors and radiologists.
• The Metaphor: Medical physics is like a high-stakes video game where the player is a doctor. If they miss a shot, a patient gets hurt. AI is the "cheat code" that helps doctors visualize tumors or calculate radiation doses instantly.
• The Goal: Training future doctors to use these AI tools so they can treat patients more safely and effectively.

4. The Data Detective (Machine Learning)

• What it is: Using AI to analyze how students learn.
• The Metaphor: Imagine a teacher who can see a student's brain activity while they are doing homework. AI acts as a detective, looking at thousands of data points (like how long a student stares at a question or where they click) to figure out exactly where a student is getting stuck.
• The Benefit: It helps teachers spot trouble spots before the student even realizes they are failing, allowing for personalized help.

5. The Human Element (Attitudes & Ethics)

• What it is: How do students and teachers feel about all this?
• The Metaphor: It's a love-hate relationship. Students love having a 24/7 tutor who never gets tired. But they (and teachers) are worried about cheating, lying, and losing the ability to think for themselves.
• The Challenge: The big question is: How do we use the tool without letting the tool use us?

---

🌍 The Global Team

This isn't just happening in one place. It's a global team effort.

• The United States, China, and Germany are the "Big Three" leading the charge.
• Universities like Uppsala (Sweden), Toronto (Canada), and Heidelberg (Germany) are the main hubs where researchers are meeting up to swap ideas.
• However, the paper notes that there isn't one single "Superstar" author yet. It's more like a jungle gym where everyone is climbing and connecting, rather than one person standing on a podium.

---

🔮 What's Next? (The Future)

The paper concludes that we are in the early stages of a revolution.

• The Old Goal: To train students to be good at crunching numbers and memorizing formulas.
• The New Goal: To train students to be conductors of an orchestra. They don't need to play every instrument (do every calculation) themselves; they need to know how to direct the AI (the orchestra) to create beautiful music (solve complex problems).

The Final Warning:
Just like any powerful tool, AI can be dangerous if misused. The future of physics education depends on teaching students ethics (don't cheat), intuition (don't just trust the machine), and adaptability (learn how to learn).

In a nutshell: Physics education is getting a massive upgrade. It's moving from "memorize and calculate" to "collaborate with smart machines to explore the universe." But we have to be careful not to let the machine do all the thinking for us.

Technical Summary

Here is a detailed technical summary of the paper "Physics Education under the Application of Artificial Intelligence: Bibliometric Analysis Based on Web of Science Core Library (2021-2025)".

1. Problem Statement

The rapid advancement of Artificial Intelligence (AI) is driving a paradigm shift in physics research, moving from traditional theoretical deduction to data-driven and intelligent computing. While AI's integration into physics research is well-documented, its specific application and impact on physics education remain under-explored in systematic literature.

• Gap: Existing bibliometric studies focus on fields like quantum computing, but there is a lack of comprehensive analysis regarding how AI is transforming physics pedagogy, curriculum design, and student assessment.
• Objective: To map the evolutionary trajectory, identify research hotspots, and analyze the collaborative networks of AI applications in physics education over the past five years (2021–2025).

2. Methodology

The study employs a quantitative bibliometric analysis using visual mapping tools to process and visualize scientific literature.

• Data Source: Web of Science (WoS) Core Collection.
• Search Strategy:
  • Keywords: "Artificial Intelligence" (including Generative AI, LLMs, Machine Learning) AND "Physics Education".
  • Time Span: 2021 to 2025.
  • Initial Retrieval: 147 papers.
  • Final Dataset: 138 papers (after excluding 9 irrelevant papers via manual screening).
• Analytical Tools:
  • VOSViewer: Used for constructing co-occurrence networks (countries, institutions, authors, keywords, references) to visualize clustering and relationships.
  • CiteSpace: Mentioned in the abstract for identifying evolutionary trends and cutting-edge hotspots.
• Metrics Analyzed: Publication volume, citation counts, total link strength (collaboration intensity), centrality, and keyword frequency.

3. Key Contributions

• First Systematic Mapping: Provides the first dedicated bibliometric overview of AI in physics education, distinguishing it from general AI-in-education or AI-in-physics-research studies.
• Identification of Five Core Clusters: The study categorizes the field into five distinct research themes, moving beyond simple "AI as a tool" to "AI as curriculum."
• Temporal Evolution Analysis: Highlights the inflection point in 2023, correlating the surge in publications with the advent of Generative AI (e.g., ChatGPT).
• Global Landscape: Identifies the US, China, and Germany as the primary research hubs and maps the specific institutional and author networks driving this field.

4. Key Results

A. Publication Trends & Growth

• Exponential Growth: The field saw a sharp increase starting in 2023, peaking in 2025. This correlates with the release of Large Language Models (LLMs).
• Total Volume: 138 core papers published between 2021 and 2025.

B. Geographic and Institutional Distribution

• Leading Nations: The United States (35 articles) leads in volume, followed by China and Germany (16 each).
• Collaboration: France, the UK, and Italy show the highest "total link strength," indicating deep international cooperation networks.
• Top Institutions: Uppsala University (5 papers) and Heidelberg University of Education (4 papers) lead in publication volume. However, University of Toronto and National University of Singapore show the highest citation impact and collaborative centrality.

C. Author Landscape

• Emerging Field: No single "dominant" author exists yet; the field is characterized by many small collaborative groups.
• Key Contributors: Bor Gregorcic, Giulia Polverini, Peter Wulff, Marcus Kubsch, Ralf Widenhorn, and Gerd Kortemeyer are the most active authors.
• High Impact: Peter Wulff and Gerd Kortemeyer hold the top citation rankings, indicating their work is foundational to the field.

D. Research Hotspots (Five Core Clusters)

The keyword and reference co-occurrence analysis revealed five distinct thematic clusters:

1. Generative AI & Teaching Paradigm Reconstruction:
   • Focus: Evaluating LLMs (ChatGPT) as "physics learners" and "mentors."
   • Findings: AI can pass introductory physics exams but often makes "pre-concept" errors similar to beginners. Research is shifting from "technology panic" to "prompt engineering" and integrating AI into resource development.
2. Physics-Informed Neural Networks (PINNs) & Computational Physics:
   • Focus: Teaching students to use AI to solve physics problems, not just using AI to teach.
   • Trend: PINNs (embedding physical laws into loss functions) are becoming a new frontier in computational physics curricula, replacing traditional numerical simulations.
3. Medical Physics & Interdisciplinary Applications:
   • Focus: AI in radiation therapy planning and medical image analysis education.
   • Significance: High career orientation; AI is naturally integrated into medical physics training due to the field's reliance on image processing.
4. Machine Learning-Driven Data Mining:
   • Focus: Using ML to analyze student learning trajectories (click data, homework logs) for predictive analytics.
   • Emerging Theme: Explainable AI (XAI) is gaining traction to interpret why students struggle with specific concepts, aiding personalized learning systems.
5. Cognition, Perception, & Ethics:
   • Focus: Student and teacher attitudes toward AI.
   • Key Issues: The duality of AI as a helpful 24/7 tutor vs. threats to academic integrity, the "AI illusion" (misunderstanding concepts), and the need for new ethical boundaries.

5. Significance and Future Outlook

• Paradigm Shift: The study concludes that AI is transitioning from an external teaching aid to an internalized component of the physics curriculum (e.g., learning PINNs).
• Future Directions:
  • Adaptive Learning Ecosystems: Building systems that dynamically adjust to individual student needs.
  • Evaluation Reform: Redefining assessment methods to account for AI tools.
  • Ethical & Conceptual Integrity: Cultivating "physical intuition" and AI ethics to prevent reliance on AI "illusions" that obscure fundamental understanding.
• Ultimate Goal: The objective of physics education is shifting from training students to be "good at computation" to training "scientists who master AI to explore the laws of the universe."

This paper serves as a critical roadmap for educators, policymakers, and researchers navigating the intersection of artificial intelligence and physics pedagogy in the post-2023 era.