Personalised learning is an educational approach in which the pace, content, modality, and assessment of instruction are dynamically adapted to the knowledge state, learning style, goals, and preferences of the individual learner. Technology-enabled personalised learning uses data about learner interactions and performance to drive adaptive algorithms that present the most effective next learning experience for each person, contrasting with one-size-fits-all curriculum delivery.
Content
- The theoretical grounding for personalised learning dates to Bloom’s 2-Sigma study (1984), which found that one-to-one tutoring produced student performance two standard deviations above average classroom instruction — a gap it attributed to the tutor’s ability to monitor understanding and adjust explanations in real time. Early computer-based implementations emerged in the 1970s through intelligent tutoring systems (ITS) such as SOPHIE (electronics troubleshooting) and LISP Tutor, which modelled student knowledge explicitly and selected instructional interventions accordingly. These systems required extensive hand-authored expert models and were domain-specific and expensive to build.
- Contemporary personalised learning platforms use a combination of item response theory (IRT) and Bayesian knowledge tracing to model learner proficiency across knowledge components from interaction data. Spaced repetition systems (SRS) like Anki and Duolingo apply forgetting curve models (Ebbinghaus, 1885; SuperMemo algorithm) to schedule review of material at the moment of optimal challenge. Recommendation algorithms drawn from collaborative filtering — “learners similar to you benefited from this resource next” — add a social dimension to personalisation. Machine learning classifiers detect learner disengagement or confusion from interaction patterns (response time, error rates, click behaviour) and trigger adaptive interventions.
- The EdTech industry has deployed personalised learning at scale through platforms such as Khan Academy, Coursera, Duolingo, Carnegie Learning’s MATHia, and DreamBox Learning, serving tens of millions of learners. Assessments in adaptive platforms continuously update the learner model rather than occurring as discrete events. The COVID-19 pandemic accelerated adoption of these platforms globally, generating unprecedented datasets on learner behaviour that are now being used to train more sophisticated adaptive models. Learning management systems (Moodle, Canvas, Blackboard) are integrating AI recommendation and analytics layers.
- In 2024–2025 Large Language Models are transforming personalised learning by enabling natural language tutoring at scale — Khanmigo (Khan Academy), Microsoft’s education Copilot, and numerous startups provide LLM-powered tutors that can answer questions, explain concepts at multiple levels of abstraction, generate worked examples, and provide Socratic guidance without prescripted content trees. Multimodal capabilities allow tutors to interpret student work (handwritten maths, diagrams) and respond visually. Privacy and equity concerns centre on data governance for learner profiles, algorithmic bias in recommendation, and the digital divide affecting access to high-quality personalised learning tools in low-resource educational contexts.