Artificial intelligence (AI) is rapidly integrating into various sectors, including education. While AI offers numerous potential benefits such as plagiarism detection, exam integrity measures, chatbots for student support, and enhanced learning management systems, ethical concerns remain largely unaddressed. This study focuses on three key ethical concerns surrounding AI in education: security and privacy issues, the loss of human decision-making capabilities, and the increase in human laziness. The researchers argue that while AI has the potential to revolutionize education, its unchecked implementation could lead to unforeseen negative consequences. The study's importance stems from the growing investment in AI in education (USD 253.82 million projected from 2021 to 2025) and the need to proactively address potential ethical dilemmas before they become widespread problems. The study aims to understand the potential negative impact of AI in education and propose preventive measures to ensure responsible implementation.

The literature review explores the ethical implications of AI in education, categorizing concerns into three levels: (1) the technology itself and its development; (2) the impact on teachers; and (3) the impact on students. Existing research highlights concerns regarding innovation costs, consent issues, data misuse, loss of human autonomy, and the potential for bias in AI systems. While AI enhances aspects like organizational information security and competitive advantage in other sectors, the education sector faces unique ethical challenges related to student privacy, data security, and the potential for AI to replace human interaction and critical thinking skills. The review emphasizes the lack of a robust regulatory framework to address these ethical concerns, prompting the current study to focus on three prevalent moral fears: security and privacy, loss of human decision-making, and increased human laziness. The researchers cite several examples of AI applications in education including tutoring, feedback systems, social robots, admission processes, and grading, while acknowledging the challenges associated with data analysis, bias prevention, and the potential for data manipulation.

This study employs a positivist research philosophy using a quantitative approach. The researchers utilized a purposive sampling technique to collect primary data through a questionnaire from 285 university students in Pakistan and China. The questionnaire included demographic questions and Likert scale questions to measure four latent variables: artificial intelligence (AI), loss in decision-making, human laziness, and security and privacy issues. The measures for each latent variable were adapted from previous studies and validated. To address common method bias, the researchers checked variance inflation factor (VIF) values, ensuring all values were below 3.3. Reliability and validity were assessed using item reliability (outer loadings), construct reliability (Cronbach's Alpha and composite reliability), convergent validity (AVE values), and discriminant validity (Fornell-Larcker criterion, HTMT ratios, and cross-loadings). The data analysis used partial least squares structural equation modeling (PLS-SEM) with SmartPLS software to test the hypotheses concerning the relationships between AI and the three outcome variables. Multi-group analysis was performed to assess potential moderating effects of gender and country.

The demographic profile of the 285 respondents revealed a relatively even distribution between male and female students, and a similar split between students from China and Pakistan. Most respondents were aged 20-25 and were undergraduates. The structural model analysis revealed significant positive relationships between AI and all three outcome variables: (1) AI and loss in human decision-making (β = 0.277, p < 0.001); (2) AI and human laziness (β = 0.689, p < 0.001); and (3) AI and security and privacy issues (β = 0.686, p < 0.001). The findings indicate that a one-unit increase in AI use correlates with a 0.277-unit increase in loss of decision-making, a 0.689-unit increase in laziness, and a 0.686-unit increase in security and privacy concerns. The model demonstrated a good fit (SRMR = 0.06) and predictive relevance (Q2 values of 0.033, 0.338, and 0.314 for decision-making, laziness, and security/privacy respectively). Importance-performance map analysis revealed that while AI's performance was consistent across all three outcomes (68.78%), its importance varied considerably: 25.1% for decision-making, 68.9% for laziness, and 74.6% for security and privacy. Multi-group analysis showed no significant moderating effects of gender or country on the relationships between AI and the outcome variables.

The findings strongly support the hypotheses, demonstrating a significant association between increased AI usage in education and negative impacts on human decision-making, increased laziness, and amplified security and privacy concerns. The study's results align with previous research highlighting the ethical implications of AI in education. The high predictive relevance of AI on human laziness and security/privacy underscores the critical need to address these concerns. The variation in the importance of AI across different outcomes highlights the multifaceted nature of the impact, emphasizing the need for nuanced approaches to AI implementation. The lack of moderating effects of gender and country suggests that these concerns are relatively universal among university students in the sampled contexts. The study’s limitations include a focus on only three ethical concerns and the potential limitations associated with cross-sectional data.

This study contributes to the growing body of research on the ethical implications of AI in education by providing empirical evidence of the significant negative impacts of AI on human decision-making, laziness, and security and privacy. The findings emphasize the crucial need for responsible AI design and implementation in education, prioritizing human well-being and addressing the potential drawbacks. Future research should explore other ethical concerns, investigate longitudinal impacts of AI, and examine potential mitigating strategies to maximize the benefits of AI while minimizing its negative consequences. Further research in different geographical areas and educational contexts would strengthen the generalizability of the findings.

The study's limitations include focusing on only three ethical concerns related to AI in education, limiting the scope of the findings. The cross-sectional nature of the study design limits the ability to establish causal relationships between AI use and the outcome variables. The sample of university students from Pakistan and China may not be fully representative of all educational contexts and student populations globally. The reliance on self-reported data through questionnaires introduces potential biases associated with social desirability and recall accuracy. Future studies may benefit from incorporating longitudinal data collection and qualitative methods to gain a richer understanding of the complex interplay between AI and human factors in education.