ChatGPT in higher education: a scoping review of ethical implications

• In Event: ICT and Education (II)

Proposal

ChatGPT is a modern generative Artificial Intelligence (GAI) model trained on large-scale textual data, including Wikipedia, books, and the internet, capable of generating human-like text. Since ChatGPT was released in November 2022, widespread applications have been found in various aspects of higher education, such as teaching, learning, research, admissions, and community engagement (UNESCO IESALC, 2023). Prior studies have shown that ChatGPT can help educators create teaching materials and help analyze student data, identify learning patterns, and generate valuable insights to refine teaching methods (Abdelghani et al., 2022; Bernius et al., 2022; Zhu et al., 2020). It can also benefit students with disabilities by enhancing accessibility and developing more inclusive learning strategies (Kasneci et al. 2023). Furthermore, ChatGPT enhances research efficiency and accuracy (Kooli, 2023).
However, recent GAI studies have revealed potential risks associated with the generation of factually incorrect information (known as the ‘hallucination’ issue), biases (nationality, gender, etc.), discrimination content (Narayanan Venkit et al., 2023; Munn et al., 2023; Nozza et al., 2022), etc. When GAI-generated content is involved in education procedures, such problematic content can be magnified and result in seriously negative supervision for student learning. Therefore, it is crucial to evaluate the ethical implications of using ChatGPT in higher education.
This scoping review aims to explore the ethical challenges of using ChatGPT in higher education and provide a comprehensive overview of relevant research while identifying gaps for future considerations. We adopted a five-stage scoping review framework (Arksey & O'Malley, 2005) to illustrate ChatGPT and GAI’s ethical issues in education. The five stages in the framework include identifying the initial research questions, identifying relevant studies, studying selection, charting the data, and collating, summarizing, and reporting the results. Our initial questions in this study included, a) How is ChatGPT or GAI being applied in the field of education? b) What are the main ethical issues in education, with a particular focus on higher education?
To cover a broad range of literature, we mainly collected studies from SCOPUS. Due to the emergence of the latest GAI, only a limited number of articles are published to date, especially in the field of education. To obtain the latest view of ChatGPT in education, we further expanded our search range to include the arXiv platform, which provides access to preprint articles. In addition to English literature, we recognized the importance of including studies in other languages such as Japanese and Chinese. To facilitate this, we conducted searches in prominent databases like CINII, a Japanese research database, and CNKI, a Chinese research database. By doing so, we aimed to capture a diverse range of perspectives and insights from different linguistic backgrounds, thus enriching the overall scope of our review. The initial search yielded 96 results, out of which we identified 37 that were related to our study. The articles were published in a wide range of journals covering diverse fields of studies including social sciences and management (16), medicine, health professions, nursing (13), and computer sciences (8). To conduct an in-depth analysis of the available literature in the field of GAI and higher education, this study adopts an interdisciplinary angle. The authors, including a linguistic AI researcher, offer valuable insights from both technical and practical perspectives.
The majority of articles have expressed concerns regarding the application of ChatGPT in educational settings.
● In the practice of teaching, ChatGPT can be applied in personalized and interactive learning, curriculum design, and the assessment of homework, exams, essays, and more. GAI often amplifies existing societal inequalities, sexual bias, and nationality bias present in the original training data and influences the outputs of the downstream educational applications (Latif et al., 2023; Ray, 2023). Moreover, educators’ over-reliance on GAI-based evaluations could lower the quality of assessments and fail to capture students’ true capabilities accurately (Curtis et al., 2023).
● As for learning, one major risk is the potential rise in plagiarism and cheating as students may rely on AI-generated content for essays and exams, compromising the authenticity of their work. In addition, excessive dependence on ChatGPT might erode students’ sense of responsibility and commitment to academic integrity (O’Connor & ChatGPT, 2023; Ojha et al., 2023). This raises concerns about privacy (Chan & Hu, 2023) and the alignment of human values with GAI use (Sharples, 2023).
● In research assistants, the integration of AI in academic publishing poses the risk of displacing human authors and undermining the value of their expertise, potentially impacting the credibility of research. Additionally, the attribution of fake references to AI-generated content presents another challenge, leading to misinformation and a decline in trust in academic sources (Curtis et al., 2023). The authorship of editorial pieces like O’Connor and ChatGPT (2023) has become a contentious issue, as it challenges the established core values related to human-based authorship in academic publishing (Silva, 2023).
● In the administrative aspect, the concerns revolve around ensuring fairness, reliability, and explainability in the utilization of ChatGPT (Fischer, 2023).
Furthermore, specific studies focus on mining social opinions from Twitter and Weibo (Chinese SNS). Generally, Twitter users express a positive attitude towards implementing ChatGPT in education. However, topic analysis reveals five main areas of concern that emerged in negative tweets: academic integrity, impact on learning outcomes and skill development, limitations of capabilities, policy and social concerns, and workforce challenges (Li et al.,2023). On Weibo, the majority of users maintain a neutral or negative attitude towards such applications due to these concerns. Effectively addressing these worries is crucial to ensure the successful adoption of ChatGPT while minimizing potential risks (Tian et al., 2023).
In summary, our analysis shows that the absence of regulation regarding GAI use in education and research gives rise to ethical dilemmas that require immediate attention. It is essential to establish robust guidelines, provide ethical training, and foster a culture that values originality and critical thinking. Striking the right balance between harnessing AI’s potential and upholding traditional academic principles is crucial for the sustained advancement and credibility of education and research.

Authors

• Ming Li, Osaka University

• Ariunaa Enkhtur, Osaka University

• Fei Cheng, Kyoto University

• Beverley Anne Yamamoto, Osaka University