LLMs in Education: Revolution or Classroom Risk?

 

Published: 29 June 2025, 08:20 CEST | Updated: 29 June 2025, 08:20 CEST

By Stuart Kerr, Data Journalist, Live AI Wire (liveaiwire@gmail.com, @liveaiwire)

Large Language Models (LLMs) like ChatGPT and Google's Gemini are transforming global education systems, offering both unprecedented opportunities and significant challenges. As these AI tools become increasingly embedded in classrooms from primary schools to universities, educators and policymakers face critical questions about how to harness their potential while mitigating risks to learning outcomes and academic integrity.

The Promise of Personalised Learning

LLMs are revolutionising educational approaches through adaptive learning technologies. Platforms like Duolingo Max and Khanmigo demonstrate AI's ability to customise instruction, providing real-time feedback and adjusting content difficulty based on individual student performance. A comprehensive 2023 Stanford University study found that AI tutoring systems improved mathematics outcomes by 12-14% for secondary students, with particularly strong results among those struggling with foundational concepts (arXiv:2305.18354).

"These tools can provide the kind of individualised attention that even the most dedicated teachers struggle to deliver in crowded classrooms," explains Dr. Sandra Leaton Gray, an education technology specialist at University College London. "The AI acts as an infinitely patient tutor, available 24/7 to explain concepts in multiple ways until the student understands."

Research aids like Elicit and Consensus demonstrate another valuable application, helping students navigate complex academic papers through AI-powered summarisation and question-answering. However, learning sciences expert Dr. Michael Tscholl of Northern Illinois University cautions about potential downsides: "There's a real risk that students will skip the hard work of engaging with primary sources, developing only surface-level understanding. We need to design activities that require critical engagement with the material."

Transforming Teaching Practices

For educators, LLMs offer powerful tools to enhance their work. AI applications like TeachFX analyse classroom discussions, providing teachers with insights into student participation patterns and the balance of teacher/student talk time. According to OECD data, 58% of educators using AI-assisted grading and lesson planning tools report significant reductions in administrative workload (OECD Digital Education Outlook 2023).

"These technologies are giving teachers something priceless - more time to focus on actual teaching," says Dr. Pasi Sahlberg, Professor of Education at the University of Melbourne. "When AI handles routine tasks like grading multiple-choice questions or generating basic lesson frameworks, teachers can dedicate more energy to fostering creativity and deeper discussions."

However, some education experts warn of potential unintended consequences. Dr. Rose Luckin, Professor of Learner-Centered Design at UCL, notes: "There's a danger that over-reliance on AI for content creation could erode teachers' pedagogical skills. We need professional development programs to help educators use these tools judiciously while maintaining their own expertise."

Academic Integrity in the AI Era

The rise of LLMs has sparked intense debate about academic honesty. A 2023 study published in the International Journal of Educational Technology in Higher Education found that 27% of university students admitted to using ChatGPT for assignments without proper attribution (DOI:10.1007/s10639-023-11834-1). Plagiarism detection services have responded with AI-identification tools, with Turnitin reporting that 11% of student submissions in 2024 contained significant AI-generated content, though with a 2% false positive rate (Turnitin, 2024).

"The fundamental challenge is distinguishing between appropriate AI use and academic misconduct," says Dr. Phillip Dawson of Deakin University's Centre for Research in Assessment and Digital Learning. "Clear institutional policies are essential, but we also need to rethink assessment design."

Some educators advocate moving toward evaluation methods that emphasise process over product. Dr. Helen Crompton of Old Dominion University suggests: "Assignments like reflective journals, in-class presentations, and project-based learning make it harder to rely on AI-generated content while better developing critical thinking skills."

Addressing Equity and Implementation Challenges

The integration of AI in education raises significant equity concerns. While affluent schools rapidly adopt cutting-edge tools, data from the Brookings Institution shows only 37% of U.S. low-income schools have the technological infrastructure to support advanced AI applications (2024). Privacy represents another critical issue, with the UK Information Commissioner's Office issuing specific guidelines about GDPR-compliant AI use in educational settings (ICO, 2024).

Dr. Nicol Turner Lee of Brookings emphasises: "Without deliberate policy interventions, AI could exacerbate existing educational disparities. We need targeted investments in broadband access, hardware, and teacher training to ensure equitable benefits."

The Path Forward

As educational institutions navigate this transformation, experts emphasise several key priorities:

1. Comprehensive Teacher Training: Professional development programs to help educators integrate AI tools effectively

2. Revised Assessment Strategies: Moving toward evaluation methods that emphasize higher-order thinking

3. Equity-Focused Policies: Ensuring all students benefit from AI advancements

4. Ethical Use Frameworks: Clear guidelines for appropriate AI use in academic work

"Banning these tools is neither practical nor desirable," argues Dr. Alec Couros of the University of Regina. "Our responsibility is to teach students how to use AI ethically and effectively - these are essential skills for the future workplace."

As LLMs continue to evolve, their ultimate impact on education will depend on how well societies balance innovation with thoughtful implementation. With careful planning and inclusive policies, AI could help create more personalized, engaging, and effective learning experiences for students worldwide.

Sources:

1. Stanford University (2023). AI Tutoring Efficacy. arXiv:2305.18354

2. International Journal of Educational Technology in Higher Education (2023). DOI:10.1007/s10639-023-11834-1

3. Turnitin (2024). AI Detection Accuracy Report

4. OECD (2023). Digital Education Outlook

5. Brookings Institution (2024). Digital Divide in AI Education

6. UK Information Commissioner's Office (2024). AI and Data Protection Guidance