The emergence of large language models (LLMs) in education gives us a good chance to change the classroom and the way we learn. On the other side, ethical issues arise from the possibility that they become susceptible to training data poisoning, a situation where bad actors add false information to the training data. Manipulation of LLMs can result in inaccurate or harmful output. This study will discuss the ethical implications of training data poisoning in LLMs, which may impair students’ critical thinking skills. In this research A survey was conducted in which students dialogued with a chatbot model that was trained on a dataset comprising both clean and deliberately poisoned information. The survey investigated student confidence in the given information and information-seeking strategies. The findings reveal a concerning trend: many students simply accepted the chatbot’s information without checking its accuracy. This underscores the danger of trusting AI-generated content blindly and the consequences of data poisoning, which can affect critical thinking skills adversely. Through the focus on the role of ethical issues in the creation and application of LLM in educational life, this paper urges careful practices and students critical thinking. This is the way in which LLMs are being used to become tools that bolster the learning process and cooperate to strengthen rather than weaken critical cognition.
Al-abonassir et al. (Fri,) studied this question.