AI-powered adaptive learning system for personalized education

Abstract

An adaptive learning system for personalized education based on AI, comprehensive: • A real-time engagement analysis unit that analyzes students' cognitive load and attention span as well as their engagement level using biometric and behavioral data, the engagement analysis unit consisting of: • Eye-tracking sensors for monitoring gaze direction, pupil dilation, and blink rate. • Monitoring keystroke dynamics to explore typing speed, pressure, and pauses. • Behavioral interaction analysis - recording cursor movements, timing of responses, patterns in interaction.

Description

Field of the invention

[0001] This invention generally relates to the field of enhanced educational technology systems, and more particularly to AI-enabled adaptive learning systems that adapt learning content to students, taking into account their cognitive load, attention span, and learning efficiency. Specifically, the invention utilizes artificial intelligence (AI), machine learning (ML), such as deep learning models, eye-tracking technology, and keystroke dynamics to continuously assess student engagement based on the aforementioned data and automatically and instantly adapt instruction. Seamless integration with intelligent learning management systems (LMS) and virtual classroom platforms improves and optimizes student learning outcomes and provides instructors with real-time feedback on their performance.This invention represents an innovative approach to personalized education that leverages machine learning and AI to enhance personalized learning experiences and the use of data in the classroom. Background of the invention

[0002] The cornerstone of all societal progress, including individual progress, is education. Unfortunately, conventional education models tend toward a standardized, one-size-fits-all approach that fails to account for individual differences in cognitive abilities, learning styles, attention spans, and engagement. Most prospective students then suffer from cognitive overload, lack of motivation, or a lack of learning ability, leading to poor learning outcomes. Adaptive learning systems are increasingly personalizing education to the needs of students. While early adaptive learning systems were based on simple static variations of course content and rule-based process decision-making, advances in artificial intelligence (AI) and machine learning (ML) have laid the technological foundation for today's significantly more sophisticated, adaptive, and immediate personalized learning systems.

[0003] Most digital learning platforms, such as learning management systems (LMS) and virtual classrooms, are designed to deliver content and assess students, but they don't adapt to their users in real time. While most systems offer static course materials, they don't dynamically adapt teaching strategies to student interactions. As a result, students often struggle to access relevant learning materials that match their current knowledge level and cognitive load. Traditional LMS platforms also don't utilize real-time behavioral and biometric data and rely on static assessments (quizzes and surveys), which don't provide an accurate measure of learning behavior. Even with well-known generative AI in education, such as ChatGPT, instructors have limited customization options to individualize responses.This creates generalized learning experiences and prevents genuine, spontaneous adaptation to content and context.

[0004] Adaptive learning is an area where AI, ML, and deep learning converge. It evaluates student behavior in real time and dynamically adapts lessons to maximize learning outcomes. AI is used to increase personalization in education by continuously assessing cognitive load, attention, and learning efficiency, creating a gentle feedback loop to reinforce learning. This system uses biometric indicators (gaze prediction) using Convnets and behavioral indicators (keystroke dynamics) using recurrent neural networks to assess engagement. Eye-tracking technology measures students' visual focus, gaze patterns, and blink rate to assess their attention, while keystroke analysis evaluates their typing speed, pressure, and pauses to estimate cognitive effort.These multimodal data sources enable the system to closely link accurate inferences about a student's cognitive state with the dynamic adaptation of instructional content.

[0005] Traditional adaptive learning models can motivate learners in the many ways described above, but they largely rely on performance-based metrics such as quiz scoring systems to dynamically adapt learning paths. This includes behavioral analytics, biometric cues, and environmental factors to better understand students' learning status. Based on this analysis, the system adjusts the difficulty of the material, adds additional resources, or suggests practice activities to maintain attention. Thanks to deep learning models that are constantly updated with information from offices and classes, especially in the UAE, the system is constantly evolving, ensuring it remains culturally sensitive.

[0006] Furthermore, this AI-adaptive learning engine integrates seamlessly with smart LMS and virtual classroom systems. This enables automated content adaptation, instant feedback for teachers, and intelligent recommendations for interactive learning exercises. Teachers gain access to data on student performance and engagement to develop a continuous improvement cycle and adapt their teaching approach and interventions. The focus is on providing actionable analytics for teachers, improving the overall efficiency of the learning process through the system, and achieving better learning outcomes for students.

[0007] This invention can be used in a variety of educational settings, including elementary schools, higher education, corporate training, online distance learning, and more. It offers personalized learning paths and adapts module content to each learner's cognitive abilities and engagement level. The system uses attention and cognitive load data to provide real-time feedback to maintain student engagement and willingness to learn, minimizing the risk of frustration and burnout. For educators: Real-time insights help identify students with learning disabilities and tailor interventions and teaching methods accordingly.

[0008] This invention is revolutionizing the education industry by mimicking outdated learning effectiveness and adapting it to the learner in real time using AI. Utilizing cutting-edge techniques in AI, ML, and biometric monitoring technology, the system guarantees personalized, efficient, and engaging learning experiences for students. ChatGPT's seamless integration with digital learning platforms further enhances accessibility and scalability, ensuring that high-quality education reaches diverse student populations, including those with learning disabilities or language barriers. In the ever-evolving AI landscape, this system can adapt and improve by learning through trial and error. Thus, it can transform the education process by tailoring learning experiences to individual needs, paving the way for the future of personalized learning. Summary of the invention

[0009] The present invention offers an AI-driven, adaptive learning concept for personalized learning that adapts lesson planning to students' cognitive load, attention span, and learning efficiency. Machine learning is used for real-time data analysis, and deep learning models are used for predictive analytics to assess student engagement and learning progress. Through repeated analyses of learning behavior, the system determines individually how students can be best supported in each subject, thus optimizing retention, engagement, and learning throughput.

[0010] The core idea of ​​this implementation lies in combining multiple data sources to measure and respond to student engagement in real time. The system uses convolutional neural networks (CNNs) for gaze detection and recurrent neural networks (RNNs) for keystroke analysis to comprehensively understand students' cognitive state. While studies demonstrate higher overall engagement, eye-tracking technology allows professors to examine students' visual attention—based on gaze direction, blink rate, and pupil dilation—to determine their level of concentration. Keystroke dynamics analysis captures typing speed, press force, and pause patterns, providing another indicator of cognitive effort and processing speed.Dictation logs provide information on task completion status, while interaction pattern recognition analyzes student reaction times, cursor key interactions, response sequences, and engagement with interactive content. This allows the system to assess cognitive load and adapt lesson content in real time.

[0011] While traditional learning management systems (LMS) often rely on static content and fixed learning paths, this adaptive learning system utilizes real-time feedback loops. The system adapts material in real time to student abilities, suggests additional learning materials, and introduces engaging activities to capture students' attention and reinforce concepts. This dynamic adaptation prevents cognitive overload or disinterest and ensures a balance between challenge and understanding.

[0012] It works with smart LMS platforms and virtual classrooms, providing teachers with near-real-time feedback and data-driven insights into whether a student is performing well, engaging, and learning. Analytics help teachers identify low-performing students early, adjust teaching methods, and implement targeted interventions—ultimately leading to better learning outcomes. Furthermore, the AI-based system adapts to different learning environments and supports both synchronous and asynchronous learning modes. This makes it suitable for in-person, distance, and hybrid learning models.

[0013] Another important aspect of the invention is its self-improvement. Since the system does not require standardized training data, it can also access data from real-world learning environments, allowing it to continuously optimize its models. For example, if the system were deployed in schools in the UAE, it would adapt to the specific cultural and contextual aspects of learning by collecting more data on student engagement in UAE classrooms, allowing it to more accurately identify patterns of cognitive load and engagement. This allows the system to remain highly relevant and effective across a wide range of learning environments and populations.

[0014] offers tremendous benefits in many educational settings, including elementary school, post-secondary education, corporate training programs, and web-based learning. This ensures that students with different learning paces, cognitive abilities, and levels of engagement receive tailored instruction—rather than a one-size-fits-all instructional module, they receive a fully personalized learning experience. This is also helpful for children with learning disabilities or language difficulties and promotes accessibility in the digital age for all students.

[0015] This system goes beyond personalized education, providing educators, institutions, and policymakers with a framework for data analytics. It generates comprehensive performance reports, interaction metrics, and learning pattern analyses, enabling educational institutions to improve their curricula and teaching methods. By automating the measurement of student engagement and adapting lessons accordingly, the invention reduces teachers' workload and enables them to offer more interactive and effective teaching methods.

[0016] This AI-based personalized learning solution represents a groundbreaking breakthrough in today's educational culture, specifically addressing the shortcomings of traditional educational methods. Thanks to AI-based personalized real-time learning, biometric engagement tracking, and adaptive learning methods, this solution increases student engagement, optimizes teaching effectiveness, and provides teachers with actionable insights. This invention has the potential to transform education through AI-based adaptive learning, enabling the next generation of digital learning platforms, personalized learning, and data-driven, efficient learning experiences. Description of the invention

[0017] These and other features, aspects and advantages of the present invention will become more apparent from the following detailed description with reference to the accompanying drawings, in which like characters represent like parts throughout the drawings. Fig. shows a block diagram representing an AI-powered adaptive learning system for personalized education.

[0018] Fig. AI-powered adaptive learning system for personalized education (block diagram). The system comprises 100: • A first real-time engagement analysis block 102 includes a module for continuous real-time engagement analysis of the eye-tracking sensors 102(a), keystroke dynamics monitoring 102(b), and behavioral interaction analysis 102(c). These elements influence cognitive load, attention span, and learning style. The engagement analysis unit serves to enhance personalized learning by automatically adapting the lesson to the level of engagement detected by the engagement analysis unit. • An engagement and interaction analysis unit, which operatively communicates with the engagement analysis unit and processes student behavior data, applying deep learning models such as CNNs for gaze analysis and RNNs for keystroke dynamics analysis; Unit - Content Adaptation - educational process and adaptation process in the Real Thyme unit - responsible for modifying the teaching material in Real Thyme. • A learning path decision support engine 106 provides real-time recommendations for alternative instructional strategies, supplemental instructional content, and interactive exercises. To improve the prediction and efficiency of instructional adaptation over time, the recommendation engine utilizes reinforcement learning algorithms 106(a). • An integration module 108 for intelligent learning management systems (LMS) facilitates the integration of AI-powered systems with existing digital learning systems. It provides an automated data exchange interface that enables real-time tracking of learning progress, automated feedback generation, and decision support for teachers. • A storage management unit 110 for storing student participation profiles, personalized lesson plans, and historical learning data. The storage management unit utilizes parallel access architectures 110(a) to enable simultaneous reading / writing by multiple students. This allows for timely retrieval and efficient processing of data. • A unit 112 runs a machine learning model, such as adaptive learning models, and dynamically updates the models based on new student interaction data. Furthermore, the unit integrates online learning mechanisms 112(a) that continuously optimize AI models through a continuous flow of new engagement and performance data, further personalizing learning. • A power management system 114 adjusts the computing resources provided to the machine learning units depending on the workload intensity. It also facilitates the distribution of adaptive learning tasks with high energy costs, thus optimizing power consumption while ensuring system functionality.

[0019] In one embodiment, the real-time engagement analysis unit 102 implements an adaptive learning mechanism that tracks student concentration using advanced biometric sensors that capture eye movements, pupil dilation, and blink rate. This data is fed into trained deep learning models that use student engagement behavior to infer cognitive fatigue, distraction, or concentration. Based on this assessment, the system then adjusts the difficulty and pace of the lesson to the student's cognitive state.

[0020] For example, the key dynamics module 102(b) analyzes the student's typing speed, pause length, and keypress force to calculate the student's cognitive effort. The byproduct is multimodal engagement tracking with gaze analysis, allowing instruction to be adapted based on both visual and behavioral cues. In at least one embodiment, the machine learning-based content adaptation unit 104 may utilize a hybrid AI model that uses a mix of supervised and reinforcement learning. This allows the system to identify potential learning barriers, adjust instructional strategies in real time, and increase motivation without human intervention—all direct results of this hybrid AI execution.

[0021] In one embodiment, the learning path recommendation engine 106 implements a personalized reinforcement learning algorithm that adapts to student behavior. Initially, the system suggests standard adaptive learning paths, but becomes increasingly intelligent in its recommendations based on long-term performance trends of student groups. The intelligent LMS integration module 108 facilitates integration with learning management systems (LMS), ensuring seamless interoperability with existing digital education platforms. It enables real-time visual representations of student progress, automated feedback reports, and instructor dashboards with detailed, AI-generated insights into student engagement and learning effectiveness.

[0022] One embodiment utilizes a high-speed data retrieval system in the memory management unit 110, enabling rapid access to student learning records and engagement histories to generate system responses. This allows AI-based personalization models to access historical learning patterns (e.g., learning styles, skills to be learned, learning levels, etc.) on demand. Real-time model adaptation: In one embodiment, the machine learning execution unit 112 updates the AI ​​models used to analyze engagement, adapt instruction, and generate recommendations in real time. It utilizes continuous learning algorithms that are updated with current engagement data from student interactions.

[0023] In one embodiment, model execution unit 112(a) selects the relevant AI architecture based on the complexity of the student data. For sparser student signals with irregular engagement patterns, the system leverages deeper neural networks for finer analysis granularity. At the other extreme, lightweight AI models are deployed to enable robust learning behavior, thus optimizing computational performance.

[0024] The power management system 114 may, in one aspect, utilize resource scheduling algorithms that allow for the allocation of computing power as needed. Others increase the power supply to the machine learning math units during high-intensity AI calculations to ensure smooth real-time learning adjustments. When idle, it does not shut down but enters a low-power mode, which takes only a few milliseconds to reactivate.

[0025] In one embodiment, the system includes an automated scaling mechanism that scales processing resources according to the number of active students. This allows the AI-driven adaptive learning engine to scale effectively in a dynamic learning environment, ensuring responsiveness and performance even with fluctuating student interaction volumes. Areas of application: Personalized learning applications - the invention is suitable for a wide range of educational contexts, from elementary and secondary schools to higher education, corporate training, and online learning platforms.

[0026] In another embodiment, the system adapts engagement monitoring and teaching strategies to the individual needs of students with learning disabilities. In another embodiment, the system enables instruction to be adapted to multiple languages ​​and optimizes lesson structure according to comprehension levels and different language proficiency. One of the inventions in this category, for example, is an AI-driven teacher dashboard that displays student learning progress in real time and provides information such as engagement metrics, assessments, and intervention suggestions. This allows teachers to take data-driven and targeted instructional actions.

[0027] This patent describes an adaptive learning system that leverages machine learning and real-time feedback to enable personalized education. This includes real-time tracking of learners' biometric data and learning focus, content adaptation using deep learning, and machine learning-based peer and content recommendations. Using advanced AI-based algorithms, this platform continuously monitors classroom flow by adapting to each student, maintaining optimal engagement, and improving knowledge retention across diverse learning environments. This invention offers highly scalable solutions for the coming era of digital education thanks to LMS integration, performance-enhancing real-time model updates, and a compute architecture that leverages efficient energy resources.By generating data-driven insights to support teachers, delivering personalized learning to individuals, and monitoring engagement, such as students' daily learning time on a particular topic, this AI-driven model marks this system as the future of adaptive learning for effective knowledge acquisition and ultimately academic success worldwide.

Claims

[1] An adaptive learning system for personalized education based on AI, comprising: • a real-time engagement analysis unit that analyses students’ cognitive load and attention span as well as their engagement levels using biometric and behavioural data, the engagement analysis unit consisting of: • Eye tracking sensors to monitor gaze direction, pupil dilation and blink frequency. • Keystroke dynamics monitoring to explore typing speed, pressure and pauses. • Behavioral interaction analysis - recording cursor movements, timing of responses, patterns in interaction. [2] The system of claim 1, wherein the machine learning-based content adaptation unit is configured to adapt learning content based on the application of deep learning models, such as training model sets for the gaze detector using convolutional neural networks and keystroke analyzer with recurrent neural networks, which increases the performance of teaching recommendations by live analysis of student engagement feedback. [3] The system of claim 1, wherein the learning path recommendation engine uses reinforcement learning algorithms to generate personalized instructional suggestions, recommend additional resources, interactive exercises, or alternative teaching methods that are iteratively refined based on historical patterns of student performance. [4] The system of claim 1, wherein the storage management unit is designed to store student engagement profiles, historical learning data, and personalized lesson plans and utilizes parallel access architectures that enable concurrent read / write operations for different student records, thereby ensuring optimal retrieval of the learning data. [5] The system of claim 1, wherein the machine learning execution unit continuously models and adapts AI models based on student interactions over time using online learning and model architecture selection. [6] The system of claim 1, wherein the power management system is capable of monitoring and managing computational demand by optimising (in hardware, software and / or firmware) the distribution of received power, passive regulation of certain components (CPU, GPU, RAM, storage, network bandwidth, etc.) and activatability functions as a function of requested memory and / or data band usage, further implementing full power saving modes during computational inactivity without impacting structural connectivity or responsiveness. [7] The system of claim 1, wherein the real-time engagement analysis unit comprises a multimodal biometric sensor module for eye tracking, facial expression recognition and posture analysis. [8] The system of claim 1, wherein the learning path recommendation engine further comprises: • a student achievement estimator: retains knowledge and skills over a longer period of time; • a recommendation engine that is multimodal and includes audio, video and text content according to students’ preferences; • an adaptive difficulty function that dynamically adjusts the complexity of the lesson based on student responses. [9] The system of claim 1, wherein the machine learning execution unit can integrate external educational datasets and federated learning models, enabling continuous refinement of AI-driven personalization while preserving the privacy of teachers and students, and periodic refreshing of learning models that leverage available data from cloud-based academic resources

Images