The future of education utilizing artificial intelligence in Turkey

The paper frames artificial intelligence (AI) as machine-based technologies that perform cognitive tasks (e.g., predictions, diagnoses, recommendations) and highlights their growing role in education for enhancing materials, methods, and organizational models. It reviews AIEd’s potential to support individualized learning, feedback, and instructional assistance (e.g., NLP chatbots, adaptive learning), and notes the need for research on implementation, management, and efficacy of educational technologies. The study positions Turkey’s AI-in-education trajectory within broader geopolitics and policy, arguing that government support for R&D and innovation is vital and that universities must play a central role in AI-related workforce development. It seeks to understand current conditions, likely future scenarios, and appropriate actions for maximizing AI’s educational benefits in Turkey.

The literature indicates AI’s widespread educational applications in both learning support and administration: personalized/dialog systems, exploratory education, educational data mining, student writing analysis, intelligent agents and chatbots, special needs support, AI-based assessment, and automated test generation. Administrative applications include curriculum, staffing, exam management, cybersecurity, and facility/security management. Historically, AIEd progressed from knowledge-based systems (1980s–2000s) to data- and logic-based approaches, with Intelligent Educational Systems (IES) providing personalized, adaptive paths via domain, learner, and pedagogical models. Deep learning and ANN advances have expanded capabilities for adaptive learning and analytics. The review also surveys international AIEd efforts (e.g., SquirrelAI, ALEKS, IBM Watson, Third Space Learning, Sana Labs) and adaptive systems concepts (e.g., knowledge maps and individualized content). In Turkey, workshops, ministries, and universities (e.g., Ministry of National Education, Istanbul Technical University, Manisa Celal Bayar University) have advanced projects like smart classroom behavior management, attendance via image processing, teacher training, and AI education for children with international partners. Additional strands include AI for learner evaluation in physical education via sensor data and pattern recognition, and national AI policy emphasizing education’s role in talent development, university programs, and AI curriculum expansion (e.g., TUBITAK initiatives, 10,000 Talents). The policy context blends market-based growth with national goals to elevate AI R&D and educational integration by 2025–2030.

Dataset: 7,950 records with 10 attributes collected from Istinye University and Yildiz Technical University. Data collection used two qualitative/quantitative strategies: (1) interviews and focus groups with students and instructors to capture perceptions and real-time interactions with AI-supported learning processes; (2) physical/digital surveys to gather feedback on experiences and desired improvements. Data were screened, cleaned, and balanced before applying machine learning. Modeling and pipeline: Educational data mining framework incorporating data cleaning, clustering/classification/prediction/evaluation. The primary classifier was a deep learning model implemented in TensorFlow with Adam optimizer and binary cross-entropy loss. Architecture: input layer with 9 variables; first hidden layer with 24 units; an additional hidden layer with 8 units; sigmoid output layer with one unit. Train/test split: 70/30. Training configuration included 300 epochs and batch size 28. Additional experiments involved ensemble methods (e.g., Adaboost), and artificial immune classifiers over 10 generations. Evaluation metrics included accuracy, precision, recall (sensitivity), specificity, F1-score, AUC, and Cohen’s kappa, with formulas provided for core metrics. Figures described the qualitative data collection flow and knowledge discovery/data mining architecture; RNN/DL diagrams illustrated how ML supports educational progress/assessment.

• Performance results (Table 1): Ensemble methods achieved approximately precision 0.93, recall 0.95, F1-score 0.94, accuracy 0.954 (noted also as 0.95 in text), and kappa ~0.24, outperforming standalone machine learning, AI recognition system, and swarm intelligence baselines on average metrics.
• Expert perspectives: Faculty and experts expressed overall satisfaction with technological progress and trust in AI’s potential. The most likely future scenario for academic instruction under AI is optimistic, with expected contributions to improved learning, guidance, assessment/scoring, activation of university/student activities, quality assurance, VR learning, and additional student support. Error-based learning may become less frustrating through AI feedback.
• Needs identified: Despite positive outlooks, there is a noted decline in understanding regarding AI usage methods across stakeholders, indicating a need for awareness-raising and training in Turkey.
• Assessment analytics: Internal assessment distributions illustrated aleatoric and epistemic uncertainty patterns, supporting the use of predictive models to identify at-risk students and inform targeted support and preparation for exams.

Findings address the study questions by showing that: (1) current AI-in-education applications in Turkey are expanding across instructional and administrative domains, with workshops, ministry initiatives, and university programs underpinning adoption; (2) expert judgments anticipate an optimistic trajectory for academic instruction, where AI enhances personalization, assessment, guidance, and operational efficiency; (3) to maximize AI’s benefits, stakeholders recommend equipping faculty with training via courses, conferences, seminars, and internships, and continuing policy-level emphasis on university infrastructure and AI curricula. The performance results demonstrate that ensemble approaches can effectively model student progress and risk, enabling earlier interventions. However, the identified gap in AI methodological understanding suggests that scaling effective practice requires broader awareness and capacity building. The policy landscape—balancing state strategy with private-sector innovation—positions educational institutions as key actors for research, talent development, and integration of AI into teaching and learning.

Turkey’s AI strategy in education is shaped by tensions between statist and market-based approaches, with institutions playing pivotal roles in research and talent production to support an AI-focused economy. In the absence of extensive enabling regulation, private actors have creatively integrated AI into educational services, sometimes driven by commercial strategies. Experimental modeling indicates utility for tracking student development and identifying those at risk, enabling targeted preparation and support. AI can enhance efficiency, personalization, language accessibility, and reduce administrative burdens, freeing instructors to focus on human-centric teaching. Overall, analyses and experiments suggest AI has a promising future in Turkey’s educational system, contingent on sustained policy support, university program development, and faculty training.