AI in K-12 education offers transformative opportunities for personalized learning and administrative efficiency, but its successful implementation requires careful attention to equity, ethics, data privacy, teacher training, and robust infrastructure to avoid exacerbating existing educational disparities.

Objective: The main goal of this master's thesis by Ricardo João Correia Maceira was to develop a comprehensive set of guidelines for understanding, implementing, using, and governing AI in K-12 education within the Portuguese and European contexts. The research aimed to identify opportunities AI presents for improving teaching and learning, recognize the challenges and ethical concerns its use may create, and provide actionable recommendations for educators, policymakers, students, and parents.

Methods: The study employed a Design Science Research (DSR) methodology following Hevner's approach, which integrates both theoretical knowledge and environmental analysis to create practical artifacts. The research was conducted in two main phases:

1. Rigor Cycle (Literature Review): A systematic literature review using the PRISMA methodology was conducted to analyze 22 peer-reviewed articles published between 2018 and March 2024 from Scopus and Web of Science databases. The review addressed four key research questions about the current status of AI in education, major ethical issues, advantages and disadvantages of AI techniques, and how AI can help mitigate socio-economic discrepancies.
2. Relevance Cycle (Environment Study): Two major reports were analyzed: the Eurydice Report (2019) examining digital education across 43 European school systems including Portugal, and the AI Pioneers report (2024) containing questionnaires from 310 educators and interviews with 14 education and AI specialists across EU countries. Additionally, the Portuguese K-12 educational system was characterized, including analysis of the Law of Bases, school autonomy, inclusion policies, the TEIP program for disadvantaged areas, and teachers' roles.

The artifact produced consists of six comprehensive matrices organized by topic (Teaching, Technologies, Ethics, Application Development, Infrastructure, and Policy & Governance), crossing opportunities, challenges, examples, and recommendations for AI in education.

Key Findings:

Opportunities:

• Personalization: AI enables adaptive learning tailored to individual student needs, pace, and learning styles, with real-time adjustments based on performance and emotional state.
• Automation: AI can handle repetitive administrative tasks (grading, attendance, scheduling), freeing teachers to focus on instruction and student support.
• Enhanced Teaching-Learning: AI increases student engagement, motivation, and retention through gamification, interactive content, and immediate feedback.
• Inclusion: AI supports students with physical and mental disabilities through adaptive interfaces and can bridge language barriers through translation.
• Prediction and Prevention: AI can identify at-risk students early, predict dropout rates, and enable preventive interventions.
• Extended Reach: Virtual and augmented reality, combined with AI, can provide remote education and immersive learning experiences.

Challenges:

• Bias and Discrimination: AI algorithms can perpetuate or amplify existing biases related to race, gender, socioeconomic status, and other factors if trained on unrepresentative or biased datasets.
• Data Privacy and Security: Collection and storage of sensitive student data raises significant privacy concerns and risks of exploitation.
• Accessibility Gaps: Unequal access to necessary technology, internet connectivity, and electricity can widen educational disparities rather than narrow them.
• Implementation Costs: High initial investment and ongoing maintenance costs may exclude under-resourced schools.
• Teacher Preparedness: Lack of AI literacy and digital competence among educators can hinder effective implementation.
• Over-reliance Risks: Excessive dependence on AI may reduce critical thinking, decrease human interaction, and undermine the teacher-student bond.
• Quality and Accuracy: AI-generated content may contain inaccuracies or outdated information, requiring human verification.
• Ethical Concerns: Issues include academic integrity (plagiarism through AI generation), appropriate screen time, and potential for addiction.

Implications: This research contributes to the field of AI in education by providing a structured, evidence-based framework for understanding and implementing AI technologies in K-12 settings. The guidelines offer practical value for multiple stakeholders: educators can understand how to integrate AI tools effectively while maintaining their essential role as mentors; policymakers receive recommendations for creating equitable access and appropriate regulations; parents and students gain awareness of both benefits and risks; and developers obtain insights into ethical design considerations. The work emphasizes that AI should serve as a complementary tool to enhance—not replace—human educators, and highlights the critical importance of addressing equity and ethics proactively rather than reactively.

Limitations: Several limitations affected this research. First, the rapid evolution of AI technology means some information may have become outdated during the study period. Second, while the PRISMA review included diverse sources, most articles focused more on opportunities than challenges, with limited representation of parent and guardian perspectives. Third, the Eurydice Report data came from the 2018/2019 school year, potentially missing more recent developments. Fourth, time constraints prevented full validation of the developed matrices with actual educational stakeholders beyond comparison with existing reports. Fifth, the study focused on general K-12 education rather than specific subject areas or disciplines, which may limit applicability to specialized contexts.

Future Directions: The author recommends several areas for future research:

1. Stakeholder Validation: Conduct comprehensive validation of the guidelines through questionnaires and interviews with teachers, educational staff, students, and parents across diverse contexts.
2. Subject-Specific Research: Investigate how AI impacts specific academic subjects and teaching methodologies rather than general education.
3. Deeper Portuguese Context: Conduct more detailed analysis of how these guidelines apply specifically to Portugal's educational system, culture, and policy environment.
4. Perception Studies: Develop new surveys to assess current knowledge, fears, usage patterns, and attitudes toward AI among all educational stakeholders.
5. Conceptual Ecosystem Development: Create a conceptual model of a comprehensive digital educational ecosystem incorporating AI tools and applications, using the developed matrices as foundational design principles.
6. Longitudinal Studies: Track the long-term impact of AI implementation in diverse educational settings over extended periods.

Title and Authors: "Artificial Intelligence in K-12 Education: A review of AIEd opportunities and challenges, and its application in Portuguese and European educational systems" by Ricardo João Correia Maceira. Supervised by Professor Vitor dos Santos, PhD, and Professor Adriana Cardoso, PhD.

Published On: July 2025

Published By: Master Thesis, NOVA Information Management School, Instituto Superior de Estatística e Gestão de Informação, Universidade Nova de Lisboa