ChatGPT and Beyond: The Generative AI Revolution in Education

Personalized Learning & Content Generation

Context: Students

Implementation: Generative AI models are used to create personalized learning materials, practice problems, study guides, and feedback tailored for each student. This includes generating responses to questions, solving problems, and providing explanations. The implementations can involve tools like ChatGPT and other AI models for content generation.

Outcomes: Students learn in a way tailored to their individual needs and interests at their own pace. Generates comprehensive and informative responses, often aligning with key themes in the research. Provides students with well-structured answers and explanations. Some studies show AI chatbots have a substantial effect on students’ learning outcomes.

Challenges: The ethical considerations associated with the use of AI in education, such as the risk of AI positioning itself as the ultimate authority and the potential for copyright infringement. Concerns about the accuracy of AI's responses. Shorter interventions with AI chatbots were found to have a stronger effect on students’ learning outcomes compared to longer interventions. May hinder students’ development of critical thinking and research skills. The potential for AI to reinforce biases or propagate misinformation.

AI Chatbots for Learning Support & Interaction

Context: Students (including Chinese undergraduate students, junior high school students, and EFL learners)

Implementation: AI chatbots are used to answer students' questions, provide support, and facilitate interactive learning activities such as debates and decision-making processes. This includes the use of argumentative chatbots (Argumate) and decision-guided chatbots, as well as tools like ChatGPT. The implementation can involve various types of interactions, including in-class debates and self-study.

Outcomes: Students receive feedback and support outside of the classroom. Can significantly improve students’ argumentation skills. Students in the experimental group significantly outperformed the control group in terms of learning achievements, motivation, and engagement. Faculty members viewed AI as a valuable tool for supporting teaching and learning.

Challenges: Shorter interventions with AI chatbots were found to have a stronger e ffect on students’ learning outcomes compared to longer interventions. The potential for misinformation and biases. Concerns about the tool’s implications for academic honesty and educational equity. Specifically, there were worries about the potential for plagiarism and the need to redefine plagiarism in the context of generative AI technology. The potential for AI to reinforce biases or propagate misinformation.

AI Assistants & Problem Solving Tools (ChatGPT & KAP-C)

Context: Students and Professionals in various fields including: medical education, chemical engineering education, law school exams, computer engineering programs, pharmacy practice and education, social psychiatry, and mathematics

Implementation: ChatGPT and KAP-C tools are used as learning aids for solving problems, generating answers, and providing information related to specific fields. This involves using the tools for tasks like solving geometry problems, answering questions in science education, assisting in chemical engineering problem-solving, generating answers for law school exams, and evaluating student learning. The KAP-C tool is used in pharmacy practice and education.

Outcomes: ChatGPT can be an accessible and practical tool for improving problem-solving abilities. Offers a glimpse into the current capabilities and challenges of ChatGPT, showcasing its potential to support education. ChatGPT performed at the level of a C+ student on average across all four courses, achieving a passing grade in all four exams. Students agreed on the interesting capabilities of ChatGPT, finding it easy-to-used, motivating, and supportive for studying.

Challenges: Language model occasionally makes mistakes and demonstrates a tendency to generate fictional references, highlighting the known issue of large language models sometimes producing inaccurate or fabricated information. The problem of Hallucination where generative AI models can generate realistic but inaccurate content, which could be misleading and negatively affects the learning process. Concerns about the accuracy of ChatGPT’s responses and the need for having background knowledge to evaluate them. The potential for misinformation and biases. Students opinions were also divided on the negative impact of ChatGPT on learning performance, academic integrity and dishonesty, and job displacement.

Automated Tasks

Context: Teachers

Implementation: Generative AI is used to automate grading and report creation.

Outcomes: Teachers' time is freed up to focus on more important tasks, such as interacting with students.

Automatic Grader

Context: Programming classes

Implementation: An automatic grader system was developed

Outcomes:

Accuracy and Reliability

Generative AI models can produce realistic but inaccurate content, potentially leading to misinformation.

Proposed Solutions: Develop mechanisms to ensure accuracy and reliability, especially in educational contexts; Human-in-the-loop validation.

Academic Dishonesty

Generative AI can be used for cheating and plagiarism.

Proposed Solutions: Develop policies, guidelines, and detection methods; Educate students on ethical use and academic integrity; Redesign assessment practices to focus more on meta-cognitive skills

Equity and Access

Unequal access to generative AI models.

Proposed Solutions: Design models to be inclusive and accessible to all students.

Teacher and Student Preparation

Lack of training and support for effective use of AI models.

Proposed Solutions: Provide training and resources for instructors and students; Support development and implementation of innovative learning activities.

Limitations in different disciplines

Limitations of generative AI capabilities in understanding and solving problems in some disciplines.

Proposed Solutions: Further research to ensure the safe and responsible use of chatbots, particularly ChatGPT, into mathematics education and learning.

Bias

Generative AI models can inherit biases from training data.

Proposed Solutions: Awareness and mitigation of bias in AI-generated content are essential.

Job displacement

Generative AI models could automate tasks that are currently performed by human teachers and other education professionals.

Copyright Infringement

Generative AI models can be used to generate copyrighted content without permission, raising copyright implications.

Proposed Solutions: Steps should be taken to avoid copyright infringement.

Liability

Uncertainty regarding liability for harm caused by AI-generated content.

Project Contact: Dr. Jianhua Yang

LLM Model Version: gemini-2.0-flash-lite