AI-driven personalised learning

Author: Saskia Keskpaik

The pursuit of personalised learning has long influenced educational theory and practice, grounded in the recognition that learners differ in their needs, abilities and pace. Traditionally, achieving this level of customisation was challenging due to the resource-intensive nature of personalising instruction for every student. However, the development of AI and related technologies has made it possible to overcome these limitations, enabling the scalable implementation of personalised learning. The unique needs, preferences and interests of each student can be met by these developments, which allow for the customisation of instruction.

AI-driven personalised learning systems use techniques such as machine learning, natural language processing, knowledge representation[i] and learning analytics[ii] to dynamically adapt instruction based on learner interactions (e.g. which study material the student viewed) and performance data (e.g. the student’s responses to quizzes). These systems dynamically adjust the delivery of educational content in real time by continuously analysing data derived from students’ learning activities, behaviours, past performances and individual characteristics. This processing of personal data enables the identification of patterns and insights into the students’ learning strengths, weaknesses and preferences, as well as their proficiency levels, which in turn informs and enhances the adaptive AI algorithm.

Several types of technologies enhance these systems. Natural language processing enables them to understand and respond to students’ questions, thereby creating a more interactive and engaging learning experience. Knowledge representation techniques allow the systems to organise information to be more accessible and comprehensible to each individual learner, supporting intelligent content recommendation and adaptive assessment. In addition, learning analytics help to improve the learning process and different learning environments, leading to better educational experiences.

Immediate, personalised feedback to the learner is another feature of these systems, helping students recognise and correct mistakes while giving educators valuable insights to refine their teaching strategies. This approach ensures that learners receive support exactly when and where they need it. Also, the system can test knowledge in a continuous way to ensure, for instance, that a concept is well understood before moving on to the next one. Continuous assessment is made possible through real-time data analysis, which provides insights into student engagement, comprehension and areas where they may struggle. This data may include detailed personal information about a student’s behaviour on the platform, such as time spent on an exercise, mouse clicks, key strokes and more.

AI-driven personalised learning systems have diverse and expanding applications. They are widely implemented in online learning platforms, intelligent tutoring systems and AI-powered learning assistants.

Fundamentally, AI-driven personalised learning systems focus on creating individual learning pathways, ensuring that each learner engages in activities customised to their specific needs. 

Trend developments

Currently, the development of AI-driven personalised learning systems is guided by a ‘technological’ approach. This perspective offers a somewhat reductionist view, presenting these systems as merely a more efficient version of traditional education.

However, there is a growing emphasis on a human-centric perspective, which prioritises education science principles. Some of the studies include constructivist learning theory, motivational theories and metacognition.[iv]

These studies highlight the importance of qualitative, contextual data such as learner motivations, goals, self-regulation and learner agency[v] in the development of AI-driven personalised learning systems. This involves fostering human-AI collaboration to support, not replace, human cognition and social learning.[vi] There is a growing trend towards developing systems that aim to enhance skills like self-regulation and creativity, rather than merely optimising knowledge delivery based on past data.

More attention is being directed towards further involving educators, students and researchers in developing AI-driven personalised learning systems. The aim of such collaboration is to enhance learning while increasing transparency, fairness and accountability. This approach seeks to address gaps in understanding how these systems operate, support ethical considerations, and ensure that educational technologies benefit learners and teachers.

Looking ahead, personalised learning could build on some of the other trends discussed in this report, where a personalised AI tutor might use a variety of tools to achieve goals (agentic AI), develop a personalised relationship with the student (AI companions), and monitor the student’s learning activity (automated proctoring).

Potential impact on individuals

AI-driven personalised learning systems hold the potential to democratise education globally. These systems have the capacity to promote access to education for all and to help realise the fundamental right to education. However, to reach this goal, it is essential that these AI systems are designed to support rather than replace teachers.[vii] This is particularly important at the early stages of education, notably for children.

These systems can enhance fairness in education by having the flexibility to accommodate different learning styles and individual needs, including those with special educational requirements. This fosters inclusivity and ensures that all students, regardless of their unique learning characteristics, receive the support and resources necessary to succeed.

However, as with other AI technologies, AI-driven personalised learning systems are susceptible to inherent biases, potentially reinforcing educational inequalities and creating feedback loops that disadvantage certain groups of learners. For instance, an AI tutor might subtly reinforce gender stereotypes in science, technology, engineering and math (STEM) subjects, impacting students’ confidence and future career paths. The complexity of these systems can also make it difficult to detect or address such biases, raising concerns about fairness.

Moreover, the predominance of English-based AI-driven personalised learning systems tends to favour Western, Educated, Industrialised, Rich and Democratic societies - the so-called ‘WEIRD’ societies - potentially deepening existing inequalities and the digital divide. This reliance can reinforce cultural biases, privilege certain perspectives and limit the representation of diverse languages, cultures and viewpoints, making it challenging for underrepresented groups to access inclusive and culturally relevant educational opportunities.

Systems guided by a ‘technological’ approach that are oriented to optimise the pace of learning tend to focus on the delivery of domain-specific knowledge and measurable learning achievements, such as students’ test scores and assignment completion rates. This pursuit of efficiency can neglect higher-order thinking and learner agency, offering limited pathways to the same prescribed knowledge without true personalisation and stifling creativity. Such systems can overly direct learning, potentially reducing an individual’s ability to explore, question or develop independent thinking, thereby impinging on fundamental rights such as freedom of thought and expression.

To boost user engagement, some platform providers ‘gamify’ their offerings, for example, by incorporating elements such as badges, leader boards and point systems into short lessons with multiple-choice questions. While these gamified features can make learning more appealing, they can also lead users to develop short attention spans and focus on superficial knowledge rather than in-depth research. Ultimately, this may create an ‘illusion’ of education, where improved performance metrics do not necessarily equate to genuine learning.

AI-driven personalised learning systems can collect and analyse vast amounts of potentially sensitive learner personal data, including personal details, academic records and behavioural patterns. The aggregation of usage patterns (such as frequency, connection times and duration) together with learning metrics can also facilitate the creation of detailed user profiles by platform providers. This raises significant concerns around privacy and data protection, particularly when consent mechanisms on these systems are unclear. This includes not having clear information on how data will be collected, what type of data will be collected, how it will be used and stored, and who will have access to it. This issue is especially critical when considering children’s rights, as children may not be able to provide valid consent on their own.

Lastly, persistent monitoring by AI-powered learning tools can result in aggressive tracking, which differs fundamentally from the supervision done by teachers in the classroom. These systems collect extensive amounts of data, potentially leading to intrusive surveillance and the misuse of personal information, which can pose significant risks to individuals’ fundamental rights to privacy and autonomy. Invasive monitoring of students, even in the name of personalised learning, can have a chilling effect on students’ freedom to express themselves without fear of judgement or reprisal.

Suggestions for further reading

• Berendt, B., Littlejohn, A., & Blakemore, M. (2020). AI in education: Learner choice and fundamental rights. Learning, Media and Technology, 45(3), 312-324.

• Laak, K. J., & Aru, J. (2025). AI and personalized learning: bridging the gap with modern educational goals. arXiv preprint arXiv:2404.02798.

• United Nations. (2024, October 16). Report of the Special Rapporteur on the right to education, Farida Shaheed: Artificial intelligence in education (A/79/520). United Nations General Assembly. https://documents.un.org/doc/undoc/gen/n24/298/43/pdf/n2429843.pdf 

• Yan, L., Greiff, S., Teuber, Z., & Gašević, D. (2024). Promises and challenges of generative artificial intelligence for human learning. Nature Human Behaviour, 8(10), 1839-1850.