Abstract
Meeting the varied academic needs of a diverse student body is not a new problem, and the realization of a need for personalized learning has been developing rapidly amongst all students—moving away from the one-size-fits-all approach that came with standardized curricula and assessment for students of the same age. In this article, I give a narrative overview of the need for personalization in learning mathematics, providing a summary of different interpretations of the construct of personalized learning and a touch on its theoretical and historical background. I discuss how personalized learning fits within social constructivist theories of learning, and how it is consistent with the process of developing students into independent learners. The main objective of the article is to find out how technology—artificial intelligence in particular—can be utilized to facilitate the process of personalizing mathematics learning. Mathematics educators look to respond to the needs and preferences of contemporary students and take advantage of the educational affordances that artificial intelligence platforms offer. The use of learning management systems, interactive learning systems, intelligent tutoring systems and social media in the personalization process is reviewed, and also the use of artificial intelligence systems in the process of gathering student data. Adaptive hyper-media are used in the process of hyper-personalization—adapting automatically to students’ learning needs and learning preferences. Some recommendations are given on how to move to a personalized classroom and concerns, challenges and recommendations are raised, speculating about the future role of personalized learning in mathematics education.
Keywords
Introduction
Some years ago, I had two students, a young man Martin and his friend Yvette (not their real names). Both of them were top students, planning careers in mathematics (in fact, today both of them are successful academics at university mathematics departments). Martin and Yvette did not understand mathematics in the same way. Martin preferred understanding new concepts using a visual representation, but Yvette did not like the pictures—she preferred identifying patterns, sometimes numerical and other times symbolic. I illustrate with an example.
Consider the well-known formula for the sum of the first n natural numbers.
Yvette convinced herself of the truth of the formula by adding the first number to the last:
Martin said no, lets draw a picture.
Figure 1: Sum of first n natural numbers
The sum
Both Yvette and Martin understood the mathematical idea well, but their approaches on how to get proper understanding of the formula were distinctly different.
The story above illustrates that students are different and that one-size-fits-all explanations do not necessarily work for all students. Student needs differ, and these differences are wider than only the way we understand. Some students need more complex exercises, some miss classes due to personal circumstances, others suffer from mathematics anxiety or dyscalculia, etc. In our mathematics classrooms, there are students with multiple profiles. There are students who may skip a section because they have mastered it themselves; others got stuck in the first unit. Personalized learning (PL) enables educators to meet everybody's requirements and help everyone succeed (Arranz, 2022).
The effectiveness of traditional education has been questioned repeatedly. Pass rates in mathematics at universities are low, and many students do not attend lectures any more—a testament to disinterested students with low morale. Many academics are convinced that the main reason behind the failure to keep students interested is the one-size-fits-all methodology (Bell, 2021). The question arises whether a one-size-fits-all approach to learning, where students sit in a rigid classroom to learn at the same pace and style as their peers, is sufficient any longer. The classroom, as we know it, may change entirely from a physical area with defined boundaries to a virtual environment including various components that will probably be determined by the student rather than only by the teacher. Mobile technology, personal learning environments, digital learning objects and other artefacts are ‘stretching’ and transforming the classroom (Borba et al., 2016; Engelbrecht, Llinares, et al., 2020). Today's students seek something different—a PL experience where individual interests, unique strengths and needs are recognized and provided for (Prasad, 2023).
We are still far from a really personalized educational system. Despite our awareness of a disparity between students, the normal practice still is to lump students together by date of birth and to not take unique personal characteristics, such as creativity and special competencies, into account (Engelbrecht, Llinares, et al., 2020). Learning can be enhanced when the instructional process accommodates the various learning styles of students. Using new technologies, such as social media (SM) and artificial intelligence (AI), opportunities are created for the learning process to become more student centered. Our current students often do a search for what they want to know through AI or SM. Mathematics teachers should keep up with these trends, or they risk being left behind by students, and fail to connect with their students’ preferred styles of learning (Engelbrecht & Borba, 2024).
The challenge of meeting the varied academic needs of a diverse student body is not new. Nitkin et al. (2022) provided a historical perspective of how the need for personalization developed. They ascribed the fact that all children go to school at the same age to obtain a uniform, baseline level of academic achievement and the disparities and inequalities already present at the start of formal schooling, as the main cause for this challenge. Although schools have explored a variety of strategies to address cognitive diversity, such as using small groups of cognitively more equal students (Nitkin et al., 2022), most schools currently promote standardization and uniformity rather than personalization (Nitkin et al., 2022; Tyack & Tobin, 1994). So there exists some tension between the desire to personalize education and the need to standardize assessment for promotion to the next grade (Nitkin et al., 2022).
Whereas the realization that education has to meet students’ unique needs has existed for many centuries, the one-size-fits-all structure of schools was introduced not so long ago. In American schools, this structure was introduced in the 20th century to assimilate young immigrants while preparing all students to contribute to the economy (Nitkin et al., 2022; Tyack & Tobin, 1994).
Worldwide, there is rising opposition to the education model of standardized tests—the factory model of education, criticized for leaving both students and teachers feeling like ‘widgets inside the classroom’ (Herold, 2019). Private and government initiatives in the USA began supporting research and development on PL. The Zuckerberg vision of ‘whole-child personalized learning’ to encompass students’ emotional and physical development, as well as their academic learning, has been receiving wide moral and financial support from both the private and public sectors (Herold, 2019).
Some teachers, however, found the individualized instruction too time-consuming (Tyack & Tobin, 1994) and the campaign lost momentum (Nitkin et al., 2022). The process had been largely manual and often has proved to be impractical and significantly burdened teachers—many are already overworked (Gulati, 2023).
The challenge became clear: how could teachers give individualized attention to every student without spending huge amounts of additional time and effort into their teaching?
We were fortunate: The global COVID pandemic forced us into an inevitable leap towards digital learning. The transformation process of the mathematics classroom from a cubicle where the teachers speak, and students listen and ‘learn’, to a learning environment consisting of multiple resources such as SM, AI, the internet, learning management systems (LMSs), learning experience platforms (LXPs), textbooks, family, friends—and even the teacher—was accelerated by the pandemic. We had to change.
With the development of technology, many of the issues that teachers protested against, have been resolved, and personalized instruction, or adaptive teaching, has become a prominent strategy for learning, backed by strong evidence of effectiveness (Connor, 2019; Nitkin et al., 2022). Later in the article, I return to these issues.
The approach of different learning styles for different students has been contested in research (e.g., Pashler et al., 2008). Researchers have not found adequate evidence that tailoring instruction to learning styles improves learning. However, mathematics teachers are still convinced that students do not all understand mathematics in the same way, and that provision should be made for different ways of understanding concepts in mathematics.
The main objective of this article is to investigate how technology—AI in particular—can be utilized to facilitate the process of personalizing mathematics teaching and learning to provide in the individual learning needs of mathematics students. The article can be seen as a narrative overview of the development of personalization in mathematics learning and teaching. In the overview, I focus on the formal published peer-reviewed research papers and books, but also include other less academic publications such as internet articles, opinions and more popular views.
It is somewhat problematic to suggest brand names of software packages that are currently used for PL, so the packages that are mentioned in the rest of this article are definitely not the only packages that are available and, because of the dynamic environment, in the near future may not be relevant any longer.
What is PL?
Academics do not all agree on what exactly they see as PL. There is a range of ideas on what it might entail, for example, customization, differentiation, individualization, student groupings, flexibility of instruction, and others (Berry, 2018; Verpoorten et al., 2009).
Meehirr (2026) gave a general description Personalized learning means that each student's learning experience is tailored to fit their needs. Personalized learning lets each person's wants and learning goals be met by changing things like the speed at which they learn, the materials they use, the order in which they learn them, the technologies they use, the quality of the materials, the way they are taught, and the materials they use to learn (Meehirr, 2026).
Bray and McClaskey (2013) distinguished between personalization, differentiation and individualization. According to them, where differentiated instruction is tailored to the learning preferences of different students and individualized instruction is paced to the learning needs of different students, personalized instruction is paced to learning needs, the learning preferences, and to the specific interests of different students. So, in their definition, personalization is wider than both individualization and differentiation.
In an attempt to clear the confusion, the Gates Bill & Melinda Foundation (2014) came up with a working definition: Personalized teaching is an educational- systemic approach that offers a way for teachers, education leaders, and education systems to support and guide students in a way that encourages personal excellence, love learning, and reach high achievements by placing the student at the center, and tailoring teaching and learning based on diagnosing, evaluating and measuring each student's characteristics (Segal et al., 2022).
So this comes down to PL making provision for (a) individual student profiles, (b) personal learning paths, (c) competency-based progression, and (d) flexible learning environments (Nitkin et al., 2022; Pane et al., 2017).
In the changing educational environment, personalization of learning has become an important aspect and happens when learning turns out to become personal in the student's mind. In short, PL is an approach that customizes learning for each student, tapping on students’ goals, interests, needs and mathematical abilities, addressing values and fostering logical thinking skills (Ogwari et al., 2020) and Pane et al. (2017) described PL as Personalized learning prioritizes a clear understanding of the needs and goals of each individual student and the tailoring of instruction to address the needs and goals. These needs and goals, and progress towards meeting them are highly visible and easily assessable to teachers as well as students and their families, are frequently discussed among these parties and are updated accordingly.
PL is designed to help students reach their goals by selecting appropriate learning strategies to suit their own needs and to realize their potential (Segal et al., 2022). It is part of an instructional approach that supports the needs of individual students, providing a space in which they can communicate the way they think mathematically, represent and discuss their mathematical ideas, and use mathematics to make sense of their worlds.
Looking at these opinions, personalization is wider than the initial example I related, which showed different ways of understanding conceptual mathematics. In fact, personalization in education can be grouped into five main categories, which often overlap in practice (Kucirkova et al., 2021);
Pace personalization or self-paced learning, enabling students to progress through material at their own speed instead of following a fixed curriculum, giving students the opportunity to fully understand concepts before moving on. Examples include mastery learning systems (e.g., Khan Academy) and competency-based education programs. Path personalization or content sequencing, where students learn different topics, skills, or modules in an order and manner that suits their needs, abilities, and interests. This category includes adaptive learning platforms (e.g., DreamBox, Smart Sparrow) that change lesson sequences based on performance. Content personalization in which, for different students, the content material is adapted in levels of difficulty, medium and style to fit the student's abilities and preferences. It could imply different reading levels of the same text, using multiple representations (e.g., videos, simulations, projects) and using contextual relevant examples and language (Arranz, 2022). Didactical personalization (learning style adaption) will adjust teaching strategies and tools according to increase student engagement by teaching in ways that suit the student's learning preferences and needs. It could include visual, auditory or blended learning approaches, or project-based learning. Goal and interest personalization aligns education with the student's personal goals, passions, and life plans. It could include student-driven projects or career-linked learning pathways, and connect the learning to real-world relevance, boosting intrinsic student motivation.
Arranz (2022) cautioned that a common myth or misconception that exists about PL—for that matter, about mathematics learning in general—is to think that learning is an individual journey. Although the idea is that each student should progress at his/her own pace and interest, the journey should not be lonely. Social interaction, collaboration and active participation in the classroom are essential activities in learning mathematics that are not in conflict with PL.
Berry (2018) expressed concern about this possible conflict. There is great value in students being connected to other students who are diverse in their thinking about mathematics, have diverse backgrounds to bring different perspectives about mathematics representations and ideas, have varying worldviews, and are co-learners for clarifying and critiquing one another's mathematical ideas.
So PL should support individual students while creating spaces for individuals to engage with other students.
In this article, the simple working definition that I will use is that PL is a tailored approach to learning that focuses on meeting the individual needs of each student. The learning experience provides for the unique strengths, interests, and learning styles of each student to ensure optimum engagement, knowledge retention, and professional growth (Prasad, 2023).
Introducing PL changes the dynamics between the teacher and the student. The teachers become facilitators of learning who guide students to the appropriate scaffolding to climb the knowledge ladder (Segal et al., 2022).
Past Experiences with PL
The research literature has demonstrated that technology-based personalized tutoring programs can have positive effects on learning, particularly in mathematics (Nitkin et al., 2022). Yet, there is no certainty about exactly which factors contribute to effective PL experiences for students. Verpoorten et al. (2009) speculated that since self-regulated learning leaves more control to the student on his/her learning, it is reasonable to consider that the factors influencing the sense of personalization may be conceptually and theoretically linked to the controllability of the learning situation, which in turn is a key factor for motivation.
Employing AI, the scope of the availability and effectiveness of learning content can be expanded, since it can provide flexible learning opportunities via smartphones and tablets and the development of personalized content for individual needs. Students can receive the exact online resources and assistance they require to address misconceptions and gaps in their learning and achieve the desired learning objectives. Coursework can be scheduled or resources delivered according to individual student assessment results.
Nitkin et al. (2022) reported an innovative model for adolescent mathematics instruction that includes data from technological as well as non-technological learning modalities. In their blended learning program, students begin each school year with a diagnostic mathematics assessment. Using the results from these assessments, individualized learning plans are developed for each student, which can span a spectrum of prerequisite and course-related skills. Lessons are customized for individual students, utilizing an appropriate instruction path that adapts to the student's potential and interests. Advanced students are faced with above-grade-level content, while students with gaps in their knowledge receive targeted remediation.
In India, many students use an AI-powered personalized adaptive learning tool application, Mindspark, to study mathematics and language. Based on the information generated by an individual student's responses to activities and questions, Mindspark tailors the learning path for individual students by adjusting the type and difficulty of content delivered according to the student's needs, format and pace of learning, recommending remedial exercises if the software picks up any weaknesses (Ogwari et al., 2020).
PL seems to have a positive impact on student attitudes and motivation. In a study in Taiwan, Ku and Sullivan (2002) investigated the effects of personalized instruction on the achievement and attitudes of fourth-grade students in mathematics word problems. Students’ preferences were used to create personalized mathematics word problems for the program. They found that students in the personalized program had significantly more positive attitudes toward the instructional program than did their non-personalized counterparts (Ku & Sullivan, 2002).
In the USA, a number of studies have been conducted to measure the success of PL. A nationally representative survey of the country's school principals in the USA was conducted in 2018 by the Education Week Research Center. According to this report 97% of the respondents reported that their schools were using digital technologies to personalize learning in some form (Herold, 2019).
Some Theoretical Background
The theoretical basis for PL is founded in Vygotsky's framework for personalizing content according to each student's personal ‘zone of proximal development’ (Nitkin et al., 2022; Vygotsky & Cole, 1978). In this model, the teacher can assist a student to progress to the next step in their zone of proximal development, which he/she could not reach independently without the teacher's assistance (Nitkin et al., 2022).
Historically, the PL approach is based on two different philosophies on how students learn. On the one hand, in the so-called ‘engineering model’ of PL, the idea is that the teacher decides what each student needs to learn, determines what the student already knows, and then creates an optimal path for the student to master the rest. This approach dates back to at least the 1950s, when ‘teaching machines’ were used to have students answer questions and receive feedback at their own pace (Herold, 2019).
The other model of learning asserts that learning takes place when teachers investigate students’ interests and passions, and then give them their own tailor-made opportunities to ask questions and explore. This approach to teach students individually to adapt to their personalized needs, was strongly supported by people such as Dewey (1916) with his important book ‘Democracy and Education’ and over the past century, educators such as Ku and Sullivan (2002), Nitkin et al. (2022), Ogwari et al. (2020), and many others, have attempted to implement personalized school models. They promoted the idea that students should not all be forced to work through a prescribed curriculum but allowed to manage their own learning through discovery and exploration (Nitkin et al., 2022). For decades, educators have attempted to craft differentiated lessons, attempting to accommodate student's individual strengths and unique learning styles (Gulati, 2023).
Verpoorten et al. (2009) claimed that PL can be seen as relying on three interrelated theories, constructivism (the process where students actively construct knowledge and competences through interacting with their environment), reflective thinking (enabling students to meta-levels of learning) and self-regulated learning (students control their own learning).
PL is based on the assumption that different students have different learning styles—individual students differ concerning the mode of instruction or study that is most effective for them (Pashler et al., 2008; Segal et al., 2022). As we saw in the story in the beginning of the article, in mathematics some students prefer a visual approach, others prefer identifying structures/patterns and some students gain knowledge and understanding better when information is presented verbally. In fact, Segal et al. (2022) distinguished between five different representations in mathematics, verbal, concrete, formulaic/symbolic, visual and contextual.
A special component of learning mathematics is to solve problems with imagination and creative thinking. Newton et al. (2022) made a strong case for allowing students to think creatively when learning mathematics. In spite of many voices speaking in favor of a creative way of thinking about mathematics, to a large extent, mathematics teaching has tended to stay with focussing on activities that lead to expected answers (Goos & Kaya, 2020), leading to mathematics still seen as somewhat technical and uncreative. Newton et al. (2022) do not discard the value of computational skills, algorithmic fluency and a well-practized recall of formulae and procedures, but this is only half of what is really needed to be successful in mathematics and creates the impression that mathematics is a finished system of rules, facts and formulae (Lithner, 2011).
Students all have their mental potentials that are complex and have developed their senses in different ways. The focus of all mathematics teachers should be on adapting the teaching style to provide for each student's strong points and use a variety of modalities to suit the needs of different students (Starja et al., 2020).
Newton et al. (2022) made a strong case for including creative thinking in mathematics as an integral part of the learning process—creative thinking in a wider sense to also include students’ constructions of personal meanings and the thinking behind an attempt to solve the problem, whether or not it led to a solution.
Differentiating the difficulty of the mathematical content would be one way to personalize learning, unlike teaching models where all students learn the same content at the same time (Starja et al., 2020). PL in mathematics can provide a solution (and a challenge) in that, taking into account every student's skills and intelligence, PL can create opportunities according to individual abilities to achieve the ultimate goal, which is the mathematical preparation of the capable future citizens (Starja et al., 2020).
PL in mathematics can be imagined as every student being provided with a personalized mathematics teacher who understands their needs, is familiar with their interests and background, is highly knowledgeable about mathematics teaching and learning, and can ask questions about what the student is struggling with, extending their mathematical thinking (Berry, 2018).
Transforming to a PL Environment
Drawing on the NCTM's mathematics teaching practices (NCTM, 2014), Berry (2018) raised some valid questions about PL in mathematics, which should be considered and provided for in any PL program. I list some of these issues.
How does PL build up students’ mathematical understanding, increase student confidence, and support mathematical identity? How does PL ensure that every student has the opportunity to learn rigorous mathematics content and develop mathematical processes and practices? How does PL support tasks that require reasoning, problem solving and mathematical modelling? How does PL support the use and connections between multiple representations of concepts? How does PL create space for students to interact with peers? How does PL pose purposeful questions to understand students’ mathematical thinking and understanding? How does PL connect conceptual understanding to procedural fluency? How does PL elicit students’ thinking and support a culture in which errors are viewed as reasoning opportunities?
In the changing mathematics classroom, AI-supported PL may be one of the most significant changes that we can introduce. Many teachers are uncertain about how to go about introducing this transformation, and in this section, I relate some actions that can be taken and considerations that should be taken into account when a teacher wants to transform to a PL environment.
Bray and McClaskey (2013) suggested that before attempting to transform learning and teaching employing PL, you should consider where you and your students are now and then decide which stage is a feasible goal for your circumstances. PL happens only when students are able to own their learning and this implies that students and teachers become partners in building learning environments designed to engage the students. Students have to first understand how they learn best and then acquire the skills to choose and use the tools that work best for their learning qualities.
When introducing a PL environment, Walkington and Bernacki (2014) suggested three important curriculum design criteria for such an intervention:
The depth of the personalization intervention—does the intervention draw on the deep knowledge that students have of actually pursuing their interest area, or does it merely look at surface features of their interest area (e.g., simply inserting words into learning tasks related to this interest)? The ownership of the intervention—do students take an active role in personalizing their own learning, or is the personalization imposed upon them? The width of the personalization—does the intervention aim at the interests of all members of a particular group, or does the intervention truly adapt to each individual student's specific interests and preferences?
One of the goals of PL is motivation (Verpoorten et al., 2009). Motivation depends heavily on students’ understanding of their own process of learning and their personal situation in the learning task. Students should therefore be fully aware of the learning goals and their progress in the course. In such an automated adaptive system, researchers have emphasized the importance of communicating to students the pedagogical aspects framing the PL experience designed for them by the adaptive learning technology (Verpoorten et al., 2009). This would give the student the opportunity to reflect about oneself in a defined learning context.
Lake (2023) provided a list of steps that can be considered when personalizing learning with AI. This list includes a number of obvious actions but important issues are to make provision that content can be adjusted based on students’ abilities and progress, using AI assessments that adapt to a student's performance, offering AI-enabled instant feedback and recommending additional resources, track the student's progress and suggesting when to move on to the next topic or when to review, and monitoring performance and engagement to identify areas where students may need intervention.
AI in PL
Worldwide, AI is the current buzzword and many respected scholars speculate about what impact AI will have on education. Global Market Insights (2023) predicted a compound annual growth rate of 25% from 2023 to 2032 for the smart content segment of AI that delivers improved learning experiences, such as providing PL (Naoley, 2023). According to the Learning and Development Global Sentiment Survey in 2019, in which 1953 people participated from 92 countries across the world, when asked what will be hot in workplace learning in the future, the top three options chosen were personalization or adaptive delivery, AI, and learning analytics (Prasad, 2023).
In education, it is clear that AI will have an enormous impact and, in spite of numerous challenges, we can say with some certainty that AI has the potential to contribute to solving many of the problems faced by educators today (Bell, 2021). AI has radically transformed the landscape of mathematics education. From early software to modern AI-driven platforms such as generative AI platforms (chatbots), technology has made mathematics more accessible, engaging, and personalized (Engelbrecht et al., 2025b; Pepin et al., 2024). As we move forward, the integration of AI will continue to evolve, offering even more innovative and interactive ways to enhance learning in mathematics.
Teachers often do not have time and other resources to provide their students with regular formative assessments and a mastery-based learning approach when teaching mathematics. They find it difficult to implement PL in their classrooms (Arranz, 2022; Nitkin et al., 2022) and welcome the opportunity to use technology as a tool to support PL. The developments in AI technology have already been successful in creating a PL experience based on students’ abilities and preferences. The currently available tools still fall short of expectations, but development in this area is rapid and the aim of tracking each student's progress and using a personalized approach to provide the needs of each student, looks increasingly possible (Bell, 2021).
Student Data Gathering
Using AI in a PL approach, the AI needs as much student data as possible. To prepare an AI platform to assist in PL, it needs to be exposed to as many variables to complete a task as possible, using different types of input data. Some AI systems create their own tasks after they have identified the objectives for the data they have been given. Such AI platforms often use an auto-tagging functionality that becomes more useful over time with an on-going, consistent stream of data. Without doing anything manually, auto-tagging records content assets, and interprets various keywords and produces tags that assist with categorization and search when uploading content. As the AI is fed more tags over time, the functionality becomes more effective, enabling an increasingly improved personal learning platform with little or no human intervention. So the AI gathers data to determine a student's knowledge of specific skills, and using this data, creates a constantly evolving learning pathway for the student (Mauri, 2019).
Personal information about students can be gathered through platforms such as intelligent tutoring systems (ITSs), LMSs, and massive open online courses. Adaptive hypermedia uses this process of educational data mining to develop automatic assessment systems and make recommendations to the student about their learning program. These intelligent adaptive systems are developed using machine learning and can perform complex tasks, including cluster analysis, pattern recognition, image processing and natural language processing (Ahmed, 2024).
Data gathering can be done by programs such as Khan Academy and Zearn, which produce extensive data on students’ behaviors and outcomes when engaged in technology-based instruction (Naoley, 2023; Nitkin et al., 2022). Sourcing of student data is often also done using LMS systems. LMSs, such as Moodle, Brightspace Insights or Canvas source student engagement, clickstream (logins, time spent on tasks), submissions, discussion activity and quiz results. These systems can flag at-risk students and adjust pacing for individual students. LMSs gather and monitor student information and personal learning actions to create interaction histories as part of their standard functionality. This information is used to show which course topics have successfully been mastered and which topics are still to be learned, and can be used by students and teachers to support student navigation (Verpoorten et al., 2009).
Student information systems (SISs), such as PowerSchool + Civitas Learning or Infinite Campus sources student demographics, enrollment, attendance, grades and transcripts and use these data to do predictive analytics. Learning record stores (LRSs), such as Learning Locker and Watershed keep record which video segments a student watched. Data integration and warehousing platforms consolidate the student data sourced by the LMS, SIS, LRS, and assessment platforms into a single dataset and suggest predictive analytics and institutional interventions. Examples of such platforms include Civitas Learning, which uses SIS and LMS data for success predictions; Google BigQuery, connecting LMS and SIS platforms to run AI models on student data; and Tableau (with AI add-ons) through which instructors can use dashboards to personalize student support.
Hyper-Personalization
The idea of hyper-personalization has become quite popular in internet marketing and other sectors of society. Most of us have had the experience where you look at the internet to buy (say) a lawnmower. For the next 2 weeks you and your housemates will be bombarded with lawnmower promotion material wherever and whenever you are on the internet. Adaptive hypermedia have picked up your need for a lawnmower and address this need—sometimes to your irritation. So these adaptive hypermedia systems enhance the functionality of hyperlink-based systems by making the user interaction process personalized (Brusilovsky et al., 1998; Engelbrecht, Llinares, et al., 2020; Herold, 2019).
In education, these hypermedia systems build a model of the goals, preferences and knowledge of each individual student, and this model is used throughout the interaction with the student to adapt to the needs of that particular student. By analysing vast amounts of data, the AI-driven hypermedia can learn more about individual students and their preferences, for example, a student's past performance, interests, and goals, and then target them with academic information and recommendations. A student could be given a presentation that is adapted specifically to his or her knowledge of the topic, and the most relevant links to proceed further will be suggested (Brusilovsky et al., 1998). The hypermedia will use knowledge captured about a specific student to tailor the information and the links presented to the student and to provide PL pathways for the student, through appropriate content, resources, and activities, without requiring human intervention. This helps students discover relevant content that matches their specific needs and preferences (Lake, 2023; Meehirr, 2026).
Mohan (2013) predicted that in the future, all education will be hyper-personalized. A student will have his/her own teacher, curriculum and learning resources, implying that not all students will be at the same level. Some might move ahead in mathematics while others in some other subject (Engelbrecht, Borba, et al., 2020). So, more formally, the idea with hyper-personalization is the automatic production of educational adjustments based on the student profile (Verpoorten et al., 2009).
Individualized Learning Paths
By extracting insights from the gathered student data, such as results from the assessment, engagement levels and behavior patterns, teachers (and the individual students themselves) can gain a deeper understanding of their capabilities and make informed decisions to personalize their learning experiences appropriately (Meehirr, 2026).
Based on a student's goals, interests, and prior knowledge, AI-powered platforms can generate individualized learning paths for the student, adapting the content and pace to meet the needs of the student. In this way, AI can ensure that the student receives appropriate challenges and support (Meehirr, 2026).
Recommendation systems for learning resources are AI platforms that recommend exercises, videos, or readings based on student profiles. Examples include Coursera's recommendation engine, which suggests courses and modules suited to students’ progress; and Edmodo, which uses AI to suggest learning resources for students.
Personalized Assessment and Feedback
Formative assessment can be personalized using AI, providing timely, personalized feedback to students in evaluating assignments and quizzes, enabling faster feedback delivery (Meehirr, 2026). For personalized feedback and assessment, there are AI systems that source examination performance, error types, rubric-based grading and student misconceptions and analyse student responses and provide tailored feedback or grading. Examples of such systems include Gradescope (by Turnitin), giving AI-assisted grading, feedback, and analytics; and Socrative, that can do real-time student assessment with personalized feedback.
Predictive Analytics and Student Support
For predictive analytics and student support, AI tools can predict student performance, risk of dropout or areas needing support. Such platforms include Civitas Learning; Brightspace Insights (D2L), a predictive AI platform that can help instructors identify at-risk students; and IBM Watson Education, which analyses learning data to provide insights and supports interventions.
Control Over Learning
Control over learning is an important benefit of using AI in PL. Every student can control their learning trajectory. Based on the learning outcomes, an AI platform can provide extra work or recommend additional lessons or resources. By analysing a student's performance in assessments and identifying strengths and weaknesses of the student, the system may suggest ways to improve the student's learning path. If a student struggles with a particular concept or skill, the AI platform adapts the training course to the difficulty level of the student. So, a student can determine the pace of a course to their ability. The result is a personalized experience that can empower students in that they have more motivation and confidence in their own abilities (Naoley, 2023).
Verpoorten et al. (2009) distinguished between four types of control in LMSs, system control (design decisions of the developers of the LMS); organizational control (restrictions and regulations that are specific to an instance of the LMS, including the tools and functions that are available to users of the LMS); teacher control (educational structure of learning units for example, the type and availability of learning material and tools); and the student control (reflecting the ways through which students can take control over their learning processes).
Without AI, presentation of personalized content and accommodating personal preferences and learning styles for each student are not easy to manage. PL involves giving students control over how they progress through their learning activities. This could include situations where the system would recognize that a student may actually skip some sections to take less linear learning journey through the learning material than somebody else, who may not have mastered the basic prerequisite skills for that particular topic (Mauri, 2019).
New Content Creation
To develop new courses or content for courses is a big task for teachers. Using AI technology, to process huge amounts of student-generated data, the AI platform can generate new content for courses, suiting the learning needs of individual students, such as customized quizzes, exercises, or simulations on specific learning needs (Lake, 2023; Naoley, 2023). AI algorithms can analyse a student's preferences and past learning history to recommend relevant content, individualized learning plans, for example, instructional methods and interventions (Lake, 2023). Software tools that can be used with adaptive content delivery and curriculum personalization, and can adjust learning pathways, difficulty, or order sequence based on student needs, include platforms such as DreamBox Learning (Math), which adapts mathematics problems in real time to a student's level; Carnegie Learning, an AI-driven adaptive mathematics curriculum with immediate feedback; Century Tech uses AI to build individualized learning pathways; or Squirrel AI, a large-scale adaptive learning platform tailoring learning trajectories.
PL in LMSs
AI-driven LMSs are being used in more and more institutions. In these environments, students can access lessons, presentations, training manuals, tests, and lectures, all in the same place (Naoley, 2023).
A LXP is a digital learning solution designed to offer personalized, student-centered experiences. Unlike a traditional LMS, which focuses on administrative reporting, LXPs prioritize the student's engagement and skill development, often through features like content discovery and development, social learning, and AI-powered recommendations. So, an LXP aims to create a more personalized, engaging and effective learning experience for individuals, enabling them to take ownership of their learning journey and develop the skills they need. In such a system, the LMS can identify skill gaps in a student and help fill them by delivering a course aligned with the student's learning path.
Lake (2023) suggested five ‘pillars’ of PL by applying an AI-powered LMS or LXP:
Understanding baseline proficiency implies that after the skills gaps have been identified, the system adjusts the level of difficulty of questions that are asked to the student, according to whether the student’s answers are correct or not. Recommending content. The system tracks a student's performance and progress to recommend training that aligns with the individual's proficiency and interests. Creating learning paths. If a student's level of proficiency is determined, the system builds a personal learning path in accordance with the student's specific needs. Providing proactive assistance. The system can proactively recommend relevant training programs, and even suggest information from other sources.
Sharing feedback.
Interactive Learning Systems and Intelligent Tutor Systems
Educational technologies for personalizing instruction that have been developed over recent decades largely trace their origins to computer-assisted instructional tools that were built at Carnegie Mellon University in the 1970s and 1980s (Nitkin et al., 2022).
Following up on these systems, intelligent AI-driven tutoring systems (ITS) such as ALEKS (McGraw-Hill), Reasoning Mind and ASSISTments have been developed to provide immediate and customized feedback to students, by identifying the student's areas of difficulty, assessing student understanding, and offering tailor-made feedback explanations, and exercises. These ITS's adjust learning material according to each student's progress (Meehirr, 2026). The objective is to replicate the experience of customized, one-to-one tutoring without the presence of a human teacher. An ITS is an AI tool that acts like a personal tutor, providing step-by-step guidance and hints.
Interactive tutorials in an ITS are computerized programs that provide immediate feedback and guidance to students as they learn and practise specific skills or concepts. Interactive tutorials have a question-answering functionality to provide for different levels of users and provide self-learning opportunities for students. Using multiple virtual agents, the system has a reduced psychological burden on students compared to personal dialogs with the teacher. Current examples of ITS systems include ALEKS (mentioned above) for adaptive assessment and tutoring in mathematics and science; Querium provides AI-driven step-by-step mathematics and science tutoring with immediate feedback; and Knewton Alta, for AI-powered tutoring embedded in digital textbooks.
ITS have become a critical part of eLearning and contribute to teaching complex subjects (such as mathematics) effectively. ITS uses the power of AI to simulate teacher–student interactions and give insights into how a student learns, enabling teachers to create a better learning interface (Naoley, 2023).
Chatbots
The 24-7 availability of generative AI or chatbots (e.g., ChatGPT, Copilot, Grok, Deepseek and many others) makes them a tool that can play a distinct role in PL. For instance, students can get information about their program, campus, and other study-related concerns and chatbots can be a valuable source of accurate information, especially when the human staff members are not available (Bell, 2021). By addressing individual students’ needs, automated question and answer tools can help teaching staff manage their time more efficiently by answering students’ repetitive questions and free up teachers’ time to focus stronger on personal interaction (Bell, 2021). Chatbots can communicate with students in a conversational manner, answering questions, providing explanations, offering feedback, and engaging in dialogue to enhance the learning experience (Lake, 2023; Meehirr, 2026).
There are valid concerns about the accuracy and mathematical robustness of generative AI packages, for example, how does a student judge the value of solutions and responses that are provided by the chatbot? There is also uncertainty about how students judge the accuracy and conceptual integrity of the results that they get from the chatbot and this frames a critical new question in terms of what the purpose or objective of mathematics education might become, that is, a heavier emphasis on critical thinking and evaluative skills.
SM in PL
Using SM platforms, students can experience a personalized approach to learning where they control their own pacing (Engelbrecht et al., 2025a). The SM learning environment holds the potential of individualizing the learning process to provide for the individual needs of students, where participants take ownership and responsibility of their learning processes and of the resources that they use (Engelbrecht, Llinares, et al., 2020).
Students can create their own learning plans and programs that are related to their academic and career goals, and make their decisions to address their learning objectives. Students can be exposed to continuous self-assessment by learning through personal portfolios—cumulative record of their work and accomplishments. Students are encouraged to learn beyond and outside of the traditional school setting (Engelbrecht et al., 2025a). SM students are described as having an enhanced capacity to self-organize and provide for themselves. These young people ‘are not content to be passive consumers, and increasingly satisfy their desire for choice, convenience, customization, and control by designing, producing, and distributing products themselves’ (Rowan-Kenyon et al., 2016).
The Future of AI-Powered PL
There is no doubt that AI and PL will form a rapidly expanding and significant component of all education in future. Yet, it is not a simple solution to all our current problems.
To introduce these tools seamlessly, educational institutions—schools, colleges and universities—will need to have a plan to address the challenges that will arise during the implementation process. Students should be full partners in this process—they should be aware of the limitations and capabilities of AI. Currently, some institutions still teach traditional mathematical skills, while neglecting to prepare students for the near future. AI and computer algebra systems will take over the most repetitive and technical mathematical tasks, but the tasks that require emotional intelligence and personal touch will still need to be performed by humans. So, technology will complement human teachers’ efforts and should not replace them (Bell, 2021).
Teachers should stay attentive to the evolving technological trends in learning and development and design flexible learning systems that can adapt to the changing needs of students and evolving technology. In teacher training programs, provision should be made to provide new teachers with the necessary skills to harness AI effectively for PL.
Perhaps, in future, every classroom teacher, administrator, principal, and counselor will easily access the comprehensive student data that we will have available, data and information that give us a profile of students’ needs and preferences (Gulati, 2023).
Concerns, Conclusions and Recommendations
As discussed in this article, personalization of learning can be a powerful instrument in teaching mathematics. However, it comes with its own dangers and concerns—mainly pedagogical, ethical and practical issues.
Pedagogical Concerns
It can happen that when learning paths comes too personalized, they may focus too strongly on procedural fluency or drill and practice, at the cost of conceptual understanding and logical reasoning. AI may focus too strongly on particular skills that a student struggles with, at the expense of broader mathematical thinking, for example, creativity, problem formulation (Boaler, 2016). Furthermore, students may progress through disjointed content issues, missing important mathematical connections or cumulative structures (Drijvers et al., 2016; Engelbrecht, Llinares, et al., 2020).
A valid fear is that personalization often individualizes learning, which can reduce opportunities for peer discussion, group problem-solving, and collective mathematical reasoning. Students may engage more with the AI than with their fellow students or teachers, reducing opportunities for discussion and collaborative problem solving (Trouche & Drijvers, 2014).
There is also a danger that PL may reduce self-regulation when AI makes choices about pace, content, and difficulty, possibly weakening students’ ability to plan and monitor their own learning and making independent decisions about strategies or problem-solving (Holmes et al., 2022).
Using adaptive AI systems, the role of the teacher will have to be redefined. Teachers may struggle to integrate AI-driven personalized tools while maintaining pedagogical authority and human judgement. They may rely too heavily on AI insights instead of their own pedagogical expertise. There is also a danger that teachers may lose a holistic view of student understanding (Holmes et al., 2022).
Some people claim that PL means replacing teachers and letting students do mathematical exercises without processing them deeply. However, these technological resources can be designed to promote active learning while empowering teachers to elevate their lessons. Using AI to automate the more repetitive parts of teachers’ work, they can use the time they save to focus on their job's more creative and interpersonal dimensions. This time that is saved can be utilized to promote critical mathematical thinking skills and to make mathematics more relevant for students by finding ways to match their interests in mathematics and stimulate their interest through contextual real-life problems (Arranz, 2022).
Another concern about AI in PL is that students spend huge amounts of time at a computer screen. In fact, the 2018 Education Week Research Center survey found that a strong majority of principals in the United States were worried that the trend was leading to too much screen time for students, and a perception is developing that the technology industry is getting too much influence over public education (Herold, 2019).
Ethical and Equity Concerns
For the AI platform to give reliable advice to students and teachers, it needs large volumes of student data—some of which may be sensitive and it is unclear if data ownership and usage policies may raise privacy risks. Before we can utilize the full potential of AI for PL, we need to ensure that the data collection methods are safe, secure, and transparent. With the current global sensitivity about the risk of exposing personal data, students need to be assured that their data will be kept safe—and the institution must maintain high data collection and storage standards and should be very careful with keeping student data private and safe from hacker attacks (Bell, 2021). In fact, critics warn that PL can be a pretext for massive data collection and surveillance of students to be used commercially (Herold, 2019).
A further concern is about equity—personalization often assumes consistent access to technology and internet connectivity. This assumption may disadvantage students in low-resource settings (Engelbrecht, Llinares, et al., 2020). Students from different socio-economic or linguistic backgrounds may not equally benefit from automated personalization tools, raising questions about equity in learning opportunities so that students with better access to AI may progress faster, widening the achievement gap (Holmes et al., 2022).
Herold (2019) cautioned that institutions should think twice before diving in by using massive resources in going fully into PL, but rather experiment carefully and cautiously with elements of the trend that may be a good fit for your circumstances.
So using PL in mathematics education brings dangers, which may include insufficient deep understanding, equity gaps, ethical risks, data privacy concerns, and challenges to teaching practices.
But then, on the other hand, there are so many advantages of using AI in PL. Teachers get the wonderful opportunity of facilitating informed decision making on improving training programs, based on AI-sourced and analysed student data on individual students’ learning progress and preferences (Lake, 2023).
When mathematics is learnt in a personalized environment, students reflect on their performance and understanding, teachers focus on making mathematics relevant and accessible, and both students and teachers are changing their mind-sets on what the subject is all about. Students have a voice in their learning, they co-create learning opportunities with their teachers, they work with peers to co-construct learning and understanding, and they create their own self-directed, PL pathways (Walden, 2022).
Academics should research the possible expansion of how AI can enable greater customization at scale, accommodating diverse learning preferences and individual goals within large groups and establishing robust methods for evaluating the effectiveness of AI-powered PL (Lake, 2023).
A substantial amount of research has been published on PL—and on using AI-powered PL in particular. However, one of the limitations is that in most cases the sample size is small or the investigation is focused on a single institution. To contribute to the development of more accurate and actionable predictive models for improving student outcomes, future research should include a focus on larger, more diverse datasets, longitudinal analysis, incorporating additional variables, improving model interpretability, and external validation (Ahmed, 2024).
We are on the threshold on an era where we divert from traditional classroom initiatives to contemporary initiatives that make the learning environment more dynamic. An important initiative is personalizing the instruction where we take into account individual student characteristics and needs and flexible instructional practices in organizing the learning environment (Ogwari et al., 2020).
With the data sourcing and data analytics, today we know more than ever about our students. We can see a student in a multi-dimensional, whole-person view, including their academic performance, extracurricular involvement, behavior data, and external circumstances (Gulati, 2023).
Starja et al. (2020) see students in a class as different flowers in a garden, each of which requires different ways of caring, while the teacher as a gardener must take responsibility to care for each of the types of flowers, depending on the unique conditions it requires to grow, develop and flower as good as possible.
PL using AI is ready to transform learning and development by revolutionizing how individuals learn, adapt, and grow academically. The ability to adapt the learning to the unique needs and preferences of students improves student performance and promotes increased engagement (Lake, 2023).
I support this quote from Meehirr (2026): It's important to note that while AI offers tremendous potential for personalizing education, human teachers remain crucial in creating supportive and inspiring learning environments. The combination of AI and human expertise can lead to more effective and personalized education for every student.
By investigating various scenarios in which PL was implemented and literature that have been published on the topic, by different teachers all over the world, I hope that I have succeeded in convincing teachers to embark on attempting to personalize their teaching of mathematics, using the AI tools that are now becoming available.
Returning to the story that I started with—Martin and Yvette's initial approaches to understanding mathematics are different and it is important that their initial needs are provided for to enable both of them to become successful mathematicians. Later on, when they have developed deeper, proper understanding of the concepts, they can appreciate the other ways of understanding as alternative representations. But for the majority of our students we do not have that luxury—we get one opportunity. Or rather—our students get one opportunity before moving on to a next episode—and this opportunity should be exploited as best as we, and they, can.
Footnotes
Informed Consent
Not applicable.
Funding
The author received financial support from the University of Pretotia for the research and authorship, of this article.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.