Artificial Intelligence in Education: A Review of Learning Analytics, Personalized Tutoring, Generative AI and Academic Integrity

DOI

https://www.doi.org/10.59256/indjcst.20260503050

Published In

Volume 5, Issue 3 (2026)

Published Online

September-December 2026

Pages

433-440

Article Type

Review Article

Check for updates

Artificial intelligence now touches most stages of teaching and learning. Learning analytics predicts which students are at risk, knowledge tracing models estimate what each student has mastered, intelligent tutors adapt practice to the learner, and large language models can explain, grade and write. The same tools have created new problems. Generative AI makes it easy to submit text that a student did not write, detection tools are unreliable, and models trained on student data raise questions of privacy and fairness. The problem is that these topics are reviewed in separate literatures: analytics and tutoring surveys rarely discuss academic integrity, and integrity studies rarely discuss the learning models that produce or detect the text. This review brings them into one framework of five layers: learning analytics and prediction, personalization and intelligent tutoring, generative AI in learning, academic integrity, and a cross-cutting layer of privacy, fairness and inclusion. It organizes 42 studies and compares their methods and reported results. Deep knowledge tracing raised the area under the curve from 0.67 to 0.86 over Bayesian knowledge tracing, and intelligent tutoring systems raised test scores by a median of 0.66 standard deviations. A randomized trial found that a carefully designed AI tutor produced learning gains of 0.73 to 1.3 standard deviations over in-class active learning, but a field experiment found that unrestricted GPT-4 access raised practice grades by 48% and then lowered exam grades by 17%. Detection of AI-generated text is weak: tested tools were neither accurate nor reliable, and detectors flagged 61% of essays by non-native English writers as machine-written. Few studies connect personalization with integrity or measure bias across student groups. Future work should design AI tutors with learning safeguards, assess integrity through process rather than detection alone, and evaluate both under privacy protection and across diverse learners.