www.ijcrsee.com
541
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
Review article
Received: May 03, 2025.
Revised: July 14, 2025.
Accepted: July 23, 2025.
UDC:
004.8:37.026
10.23947/2334-8496-2025-13-2-541-549
© 2025 by the authors. This article is an open access article distributed under the terms and conditions of the
Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
*
Corresponding author:
lstosic@unt.edu.rs
Abstract: Artificial Intelligence (AI) is redefining the landscape of personalized education by enabling adaptive systems
that respond dynamically to individual learning needs. This paper explores how AI technologies-including machine learning,
big data analytics, and intelligent tutoring systems-support the transformation of pedagogical models. Key opportunities
discussed include real-time personalization of content delivery, increased student motivation, and inclusive learning environ-
ments. At the same time, the study critically examines potential risks, such as data privacy concerns, algorithmic bias, and the
erosion of human-centered pedagogy. Policy implications are addressed with recommendations for regulatory frameworks to
ensure ethical and responsible AI integration into education. The paper emphasizes the need for empirical research to validate
AI-driven models in diverse educational settings. By aligning technological innovation with humanistic values, the paper con-
tributes to ongoing discourse on how AI can support-not supplant-the role of educators. The findings provide a foundation for
future research and policy design aimed at creating equitable, transparent, and effective personalized learning ecosystems.
Keywords: artificial intelligence, personalized learning, educational policy, algorithmic ethics, adaptive learning systems.
Lazar Stošić
1,2*
, Aleksandar Radonjić
1
, Olja Krčadinac
1
, Borivoje Baltezarević
3
, Olga Mikhailova
4
1
Faculty of Informatics and Computer Science, University Union—Nikola Tesla, Belgrade, Serbia;
2
Don State Technical University, Rostov-on-Don, Russian Federation,
e-mail:
lstosic@unt.edu.rs, sasa_radonjic@yahoo.com, olja.krcadinac@gmail.com
3
The Institute for Serbian Culture Priština – Leposavić, Serbia, e-mail:
baltezb@yahoo.co.uk
4
Department of Psychology and Pedagogy, Peoples’ Friendship University of Russia named after Patrice Lumumba
(RUDN University), Moscow, Russian Federation, e-mail:
оlga00241@yandex.ru
Personalized Learning through Artificial Intelligence:
Opportunities, Risks, and Policy Perspectives
AI and the Future of Personalized Learning
It is widely acknowledged that artificial intelligence is a primary driver of innovation. We are wit-
nessing the daily use of artificial intelligence and its tools in all spheres of human life. Artificial intelligence
is closely related to machine learning. Artificial intelligence refers to machines replicating human cognitive
characteristics, and artificial intelligence can give many recommendations both for improving learning and
for preparing educational material. Scientists argue that artificial intelligence should enable machines to
replicate human cognitive abilities, but at the same time there is a strong view that artificial intelligence
should never attempt to take over the entirety of human multi-layered mental processes (Filipović, Bjela-
jac, Merdović, and Stošić, 2023). Because machines can collect, track, and analyze so much data about
people, it is very possible for those machines to use that information against humans. Any powerful
technology can be misused (Bjelajac, Filipović, and Stošić, 2022, p. 359). Contrary to AI, machine learn-
ing implies that machines learn from the information they are given and access, thereby improving their
performance (Duckett, 2010). Unlike artificial intelligence, machine learning enables the development of
educational systems that can adapt themselves according to the information they receive and that dynam-
ically change in accordance with the progress of students. From an educational perspective, numerous
AI tools have greatly helped to personalize learning and thus change the way education is approached
(AlOqaily, Qawasmeh, and Tawalbeh, 2025). When we talk about education, the eternal problem has
been the quality of teaching that is provided to different categories of children. If the teaching criteria are
www.ijcrsee.com
542
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
raised, the weaker ones will diminish even more knowledge. If the criterion is lowered, then children with
advanced knowledge will remain in further progress. It is precisely in this gap that artificial intelligence
can make a contribution. In relation to the needs of individuals, the artificial intelligence can determine
and propose a learning plan and program for each individual and thus adapt learning itself to personal-
ized learning. Personalized learning adapts the content, methods and pace of learning to the individual
needs, interests and styles of each student. It is focused on the needs, abilities and interests of individual
students. What separates personalized learning from other learning is that personalized learning focuses
on the student as an active creator of their own educational experience. Students who receive content in a
format that suits their learning styles report higher levels of motivation, engagement and success (
Graf et
al., 2009). Personalized approaches lead to greater satisfaction and more efficient learning. When talking
about personalized learning, we cannot leave out artificial intelligence (AI) and machine learning, which
are increasingly being integrated into the education system (Stošić and Janković, 2024). A personalized
approach uses analytics tools and technologies, including artificial intelligence (AI), to continuously col-
lect and analyze student data to adapt instructional materials and strategies in real time (Talapbayeva et
al., 2024). This approach contrasts with traditional “one-size-fits-all” methods by accommodating different
learning paces, preferences, and goals, allowing students to progress based on mastery, not age or grade
level (Knox, 2020). Personalized learning is based on the premise that every student can succeed if they
are provided with the right resources and support (Duckett, 2010).
Personalized learning has its roots in the early 1960s when the first generation of educational
software (eg the PLATO system developed at the University of Illinois) enabled basic forms of adaptive
learning. Already in the 1990s, with the advent of the Internet, learning management systems (LMS) ap-
peared that allowed greater flexibility in the organization of the teaching process.
Personalized learning experienced a real rising in the 21st century with the development of artificial
intelligence (AI), learning analytics and big data technologies. Certainly, digitalization supported by arti-
ficial intelligence has transformed education itself, making it far more accessible, efficient and attractive
to every individual. Due to the mass use, the lack of professional staff, the question of ethics, the bias of
algorithms and even privacy itself is raised.
An overview of the modern application of AI in personalized learning
When we talk about the application of AI tools in personalized learning, we primarily mean adap-
tive tutors, intelligent tutors and recommendation systems. Based on insights from AI systems, teachers
can create differentiated lesson plans. These tools represent the foundation of modern AI applications in
education and support personalization, self-directed learning and improved student evaluation (
Duong
et al., 2025). AI systems provide immediate feedback on accuracy and understanding, thereby speeding
up the learning and correction process. AI can adjust the order of teaching units – eg. a student with an
auditory style will be offered more video lessons and dictated explanations, while visual learners will be
offered infographics, maps and animations. The authors emphasize that AI in education can enhance
educational experiences by making learning more personalized, flexible, and engaging, without replacing
teachers but rather augmenting their capabilities (Duong et al., 2025). AI tools offer analytics on student
progress and difficulties, helping teachers plan interventions more precisely (Holmes et al., 2019). AI tools
allow students to learn at their own pace. This adaptability helps students overcome frustration caused by
content that is too difficult or too easy, thereby increasing their sense of competence – an important factor
in internal motivation (Deci and Ryan, 2000). AI tools enable students with specific disabilities to master
the material more easily through the use of assistive technologies and adapted educational paths, which
contributes to more equal education and greater motivation.
Adaptive tutors use artificial intelligence to adapt content to an individual’s learning in real time. In
addition to a personalized experience, adaptive tutors can also provide students with an interactive learn-
ing experience. Adaptive tutors are web-based interactive teaching systems that evolved from intelligent
teaching systems (Kamardeen 2014; Khawaja, Chen, and Marcus 2014). Their goal is to personalize the
educational process to suit each individual’s pace and learning style, rather than relying on traditional
methods of learning. Student evaluations indicate that students prefer the computer-based approach
over the traditional approach (
Nirmalakhandan, 2007). The same author argues that based on students’
performance on two types of paper-based tests he conducted, he believes that computerized adaptive
www.ijcrsee.com
543
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
tutorials can be a valid tool in assessing their problem-solving skills. In contrast to students, adaptive tuto-
rials provide educators with the ability to easily monitor and analyze student interactions with the tutorial.
Examples of adaptive tutorials include the Knewton and DreamBox Learning platforms. What all platforms
have in common is that they provide an environment where students can work at their own pace, at a time
and place that suits them. Incorporating adaptive tutorials that include a combination of guided practice,
interactive opportunities and immediate, adaptive feedback provides students with learning opportunities
that can improve their confidence, understanding and learning outcomes (Weltman, 2025).
Intelligent tutors (ITS - Intelligent Tutoring Systems) simulate the work of a human teacher. They not
only track student responses, but also model their knowledge, detect faulty thinking, and offer feedback
aimed at correcting specific cognitive errors. In the beginning, there all talk about teaching with the com-
puter assistance. Today, that same computer-assisted instruction has evolved into Intelligent Instructional
Systems (ITS). Intelligent tutors provide real-time feedback, indicating not only the correctness of answers
but also the causes of errors. Such support develops metacognitive skills and encourages students to
reflect on their own knowledge, which contributes to academic success (VanLehn, 2011). ITS systems
often use complex learner models and techniques from artificial intelligence, such as expert systems or
machine learning (Suresh, Kolluru, and Ubaidullah, 2025). Examples are AutoTutor and Cognitive Tutor.
With the increasing development of ICT and their use in education, the problem of large amounts
of information has arisen. This large amount is reflected in the time that authors spend searching for use-
ful information and extracting the necessary and requested information. It is precisely because of these
problems that recommender systems have emerged to solve these problems. Recommender systems in
education work in such a way as to provide recommendations for teaching resources, courses or tasks.
Recommender systems have become a very commonly used tool in various fields such as e-commerce,
social networks, digital media and books (Nadine, Cao and Deng, 2016), as well as in the field of educa-
tion, in teaching and academic advisory services (Deschênes, 2020). Recommender systems are soft-
ware tools that assist users in the decision-making process by applying information filtering, data mining,
and predictive algorithms (Venkatraman, 2017). Based on previous activities, interests, or knowledge
level, the system analyzes the data and recommends the next steps in learning. The recommender sys-
tem can be very useful for students when choosing elective courses (Dwivedi and Roshni, 2017). Burke
(Burke, 2007) differentiate six different classes of recommendation approaches: collaborative filtering
(CF), content-based filtering (CBF), knowledge-based filtering, context-based filtering, demographic fil-
tering, and hybrid filtering. Typical representatives of recommender systems in education are e-learning
platforms Coursera, Edmodo, or Duolingo. Coursera, for example, uses AI to personalize the educa-
tional experience. It offers adaptive course recommendations based on the user’s previous interests,
knowledge level and goals. AI is used to automatically evaluate assignments, generate feedback, and
automatically transcribe lectures. The platform collaborates with universities and companies, enabling
massive open education (MOOC). Coursera certainly helps teachers to assess and analyze student per-
formance more quickly. A similar platform is Duolingo, which uses intelligent tutors and adaptive teaching
algorithms. Platforms like Duolingo use game elements (points, levels, challenges) to encourage motiva-
tion. Studies show that such approaches significantly increase student engagement and frequency of use
of educational resources (Vesselinov and Grego, 2012). The system uses gamification, natural language
processing (NLP) and machine learning to adjust the difficulty of tasks, recognize student weaknesses
and offer exercises that strengthen specific skills. Using gamified and interactive systems like Duolingo
increases engagement and regular learning (Loewen et al., 2020). On the other hand, Chinese education
platform Squirrel AI uses advanced intelligent teaching systems that break down skills and concepts into
the smallest possible units to accurately detect knowledge gaps.
By type of education, recommender systems cover mainly formal education, especially student-oriented
(Urdaneta-Ponte et al., 2021). AI tools such as adaptive and intelligent tutors adapt the learning content to the
individual needs, pace and style of the learner (Luckin et al., 2016). All these platforms enable learning without
time and space limitations, which is especially important for students with specific needs or from remote areas.
www.ijcrsee.com
544
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
Learning support for students with special needs Human-Centered AI for Inclusive Education
AI tools significantly contribute to inclusive education, especially when it comes to students with
special educational needs. By using artificial intelligence-based technology, it is possible to provide per-
sonalized support, assistive functions and an environment that allows each student to progress according
to their abilities. AI systems can automatically adjust task complexity, presentation speed, and content
modality (text, audio, video) according to the abilities of students with learning disabilities, autism, or
ADHD (Attention-deficit/hyperactivity disorder). Adaptive tutors enable learning without pressure and at a
safe pace (Al-Azawei et al., 2017). Tools for converting speech into text (speech-to-text) and vice versa,
image and voice recognition, as well as automatic transcription allow students with visual, hearing or mo-
tor disabilities to participate equally in classes. Systems such as Microsoft Immersive Reader or Seeing
AI use NLP and visual recognition for this purpose. AI can recognize signs of frustration, inattention or
cognitive overload by analyzing facial expressions, body language or behavior (affective computing). This
enables timely intervention and additional help for students who have difficulties in emotional regulation.
AI enables detailed assessment of student progress, identifies areas of weakness and recommends spe-
cific interventions. This is especially useful for students with intellectual disabilities, as it requires clearly
structured and goal-oriented teaching.
Risks and challenges of applying AI in personalized learning
Ethics and protection of student data privacy are key challenges in the application of artificial intel-
ligence (AI) in education. While AI tools offer numerous advantages, their use must be aligned with data
protection laws and ethical principles that protect learners, especially children and vulnerable groups. AI
systems collect large amounts of students’ personal data: names, behavioral patterns, scores, and even
emotional reactions. Unethical use of this data – without consent or a clear purpose – can lead to viola-
tions of student privacy and security (
Williamson and Eynon, 2020). Lawmaking that emphasize account-
ability, transparency in AI, mitigating bias, and comprehensive application of ethical principles is an ef-
fective strategy for dealing with “dark AI” (Bjelajac, Filipović, and Stošić, 2023, p. 530). Legal frameworks
such as GDPR (EU) and FERPA (US) require explicit informed consent from parents and students for
data collection and processing. Ethics dictate that educational institutions clearly communicate why data
is collected, how it is used and who has access (Regan and Jesse, 2019). AI models can be biased if they
are trained on unrepresentative data, which can lead to unfair educational recommendations, grading or
labeling of students. Ethics requires regular testing and transparency of algorithms to ensure fairness.
Ethics mandates the limitation of covert forms of monitoring and the use of only those technologies that
directly contribute to the educational goal.
Algorithmic biases pose serious ethical challenges in the application of artificial intelligence (AI)
in education. These problems occur when algorithms make decisions that are unfair, discriminatory, or
based on non-objective patterns in the data. Bias occurs when the training data is incomplete, historically
biased, or unrepresentative. For example, if an AI system learns from data that favors certain demograph-
ic groups (eg, students from urban areas), the resulting recommendations may be unfair to students from
other areas (Noble, 2018). AI systems can classify students according to past performance, behavior, or
demographic characteristics, creating a risk of labeling (eg, “bad students,” “low-risk students”) that can
affect the treatment they receive and teacher expectations. This type of prediction can self-confirm nega-
tive educational outcomes (O’Neil, 2017). Algorithms that recommend educational content or ways of
learning may ignore cultural, linguistic or cognitive differences. If students of different genders, languages
or ethnic groups are not offered diverse content, educational inequalities may be reproduced. In order to
mitigate bias, it is necessary to use diverse and representative data sets, conduct regular audits of algo-
rithms, involve ethical committees in the development of AI systems, increase transparency in the way
recommendations or decisions are made (Binns, 2018).
The danger of automation and depersonalization of education is becoming more pronounced with
the increase in the application of artificial intelligence (AI) in educational systems. Although AI offers
numerous advantages, an over-reliance on automated processes can compromise the core pedagogical
values and humane aspects of teaching. Personal contact between teacher and student plays a key role
in education - not only in the transfer of knowledge, but also in the development of empathy, support and
motivation. Automated tutors and evaluators cannot adequately replace the emotional intelligence and
www.ijcrsee.com
545
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
ethical judgment of teachers (Selwyn, 2019). AI systems often operate according to predefined patterns
and metrics, which can lead to a standardized approach to learning. This approach ignores contextual,
cultural and individual factors, reducing the quality and relevance of education (Knox, 2020).
Policy and Regulatory Perspectives on AI in Education
Aware of the shortcomings and threats we face, it is certainly necessary to set clear ethical guide-
lines. In general, national and institutional codes of ethics for the use of AI in education should be clearly
formulated. “The incorporation of artificial intelligence in education has enormous potential to transform
the educational process, providing benefits such as increased productivity and tailored teaching” (
Stošić
and Janković, 2023, p. 68). It’s not enough to just use AI tools. It is necessary to ensure that AI tools
consider the rights of students, with a special emphasis on privacy protection, consent and transparency.
The use of AI for mass surveillance or profiling of students without control should definitely be banned.
UNESCO’s first global standard on the ethics of artificial intelligence – the “Recommendation on
the Ethics of Artificial Intelligence”, adopted in 2021 and applicable to all 194 UNESCO member states
(https://www.unesco.org/en/articles/recommendation-ethics-artificial-intelligence). The recommendation
clearly states that tools should be developed to help member states build a comprehensive picture of
how prepared they are to ethically and responsibly deploy AI for their citizens. The Global Observatory
on Ethics and Governance of Artificial Intelligence aims to provide a global resource for policymakers,
regulators, academics, the private sector and civil society to find solutions to the most pressing challenges
posed by AI (https://www.unesco.org/en/artificial-intelligence/recommendation-ethics).
Regardless of social and geographic origin, fair access and digital inclusion should be ensured
where educational policies include strategies to reduce the digital gap. All students, regardless of gender
equality, should have access to devices and the Internet. More attention should be given to the develop-
ment of an AI strategy aimed at rural and marginalized communities.
Teachers should definitely be trained in the critical use of AI tools and not just in their use. Teachers
themselves must be involved in the decision-making and training process. We should strive to introduce
a professional development program that includes digital pedagogy and AI ethics. Educators need to
understand how AI can affect the education system as well as understand where this technology is going
with its progress. Teachers must know and understand where AI can replace humans and where it can
only complement and help them. We must constantly point out the dangers that really exist, and work
permanently on this aspect of mental health of individuals, nations, and global society (Bjelajac, Filipović,
and Stošić, 2022).
Standardizing the use of AI in schools and universities
Standardizing the use of artificial intelligence (AI) in schools and universities is a key step towards
ensuring the quality, fair and ethical application of technology in educational systems. The goal of stand-
ardization is to establish common guidelines, technical frameworks and pedagogical principles for the
integration of AI into the educational process. Ministries of education and educational institutions should
define official guidelines for the use of AI tools, including criteria for their accreditation, security and com-
pliance with pedagogical goals. It is necessary to standardize procedures for the selection, evaluation
and updating of AI systems, in cooperation with regulatory bodies and experts in education and ethics.
The Ethical Guidelines for Artificial Intelligence in Education of the Ministry of Education of China outlines
principles for ensuring the responsible use of artificial intelligence technologies in educational institutions
(
Wei and Niemi, 2023). China has developed the National Guidelines for the Ethical Use of AI in Education
(2021), while the EU is laying the groundwork for responsible use through the Ethical Guidelines for Trust-
worthy AI. Primary school children in China are prohibited from using AI tools that generate open-source
content as their own, while educators must ensure that AI can complement, but not replace, human-led
teaching (https://m.chinanews.com/wap/detail/ecnszw/hernfey9438915.shtml). According to these guide-
lines, students are not allowed to independently use open source AI-based generators. In contrast, sec-
ondary school students can use and explore AI-generated content. Parents are advised to monitor how
their children use artificial intelligence and to protect their personal data. They are urged to teach their chil-
dren not to rely on technology alone but to prioritize emotional engagement. China’s Ministry of Education
has issued ethical guidelines that emphasize principles such as fairness, transparency and accountability.
www.ijcrsee.com
546
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
All these guidelines and regulations directly mandate that the law on data privacy be strictly followed, thus
ensuring the protection of the data of the students themselves. Educational institutions themselves must
provide security measures to ensure a measure of protection of personal data. Schools are preparing
their children for future labor markets by increasing digital literacy by implementing artificial intelligence
technologies in the curriculum. The implementation of AI technology alone is not enough for progress, but
schools are obliged to monitor the application of artificial intelligence and its impact on learning outcomes.
All educators must receive training on the effective use of artificial intelligence. The government is working
with technology companies to develop artificial intelligence solutions that meet educational needs while
adhering to ethical standards. China’s approach to regulating AI in education focuses on balancing inno-
vation with ethical considerations, aiming to improve educational outcomes while protecting the rights and
well-being of students. Important themes include inclusion, personalization, justice, safety, transparency,
and responsibility, which serve as the foundation for developing AI in education.
AI tools must certainly be aligned with national curricula and exam standards. Standardization must
include pedagogical performance indicators, to give insight into whether the AI tool contributes to the
development of students’ knowledge, skills, engagement and critical thinking. It is necessary to introduce
quality indicators such as precision of recommendations, adaptation time and level of personalization. A
standardized framework must include mandatory protection of student privacy, as well as guidelines for
algorithm transparency and the possibility of human intervention.
Future Directions and Research Recommendations
In modern education, the need for training teachers and educators to work with artificial intelligence
tools is becoming increasingly evident. The introduction of artificial intelligence in schools and universities
requires not only technical integration, but also the development of digital and pedagogical competences
among teaching staff. Educators must understand how AI works, what its capabilities and limitations are,
and how it affects learning, assessment, and interaction processes. This includes a basic understanding
of algorithms, personalization, learning analytics and ethical implications (
Zawacki-Richter et al., 2019).
Training should not focus exclusively on technical skills, but also on the pedagogical integration of AI:
how to use adaptive tutors, recommendation systems or evaluation tools for the purpose of differentiated
and inclusive teaching. Educators must be trained to understand the legal and ethical aspects of using AI
tools, especially in relation to the protection of student privacy, consent and transparency of algorithms.
Training should also include recognition of biases and automated decision-making problems. Given the
rapid evolution of AI technologies, teacher training must be continuous, with self-study resources, semi-
nars and accredited courses available. Institutional support through strategies for the digital transforma-
tion of the educational system is also needed.
In general, the application of AI in education has a pedagogical value. On the one hand, there is
personalization, adaptive learning, real-time learning support versus depersonalization, reducing student-
teacher interactions. With the application of modern technologies, we have automated assessment, time
saving, scalability. The use of technologies also has its drawbacks, namely technical errors, dependence
on platforms, low interoperability. AI tools in education certainly provide progress, but one should definitely
take care of discrimination, stereotyping, unauthorized data collection, surveillance, algorithmic bias... As
for algorithmic bias, its impact in the context of differences between genders, ethnic groups and students
with developmental disabilities has not been sufficiently investigated (Holmes et al., 2022). AI tools pro-
vide support in evaluation but reduce the pedagogical autonomy of teachers.
www.ijcrsee.com
547
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
Conclusion
AI in education represents a significant step towards personalization and more effective learn-
ing. Through timely support, motivational strategies and analytical insights, these tools help increase
both academic success and intrinsic student motivation. Automatic detection of learning styles through
AI analytics is a key step towards creating an intelligent, adaptive education system that does not treat
all students equally, but in accordance with their cognitive profiles. Students with special needs are given
tools that allow them to achieve their full potential much more easily.
Every phase of the development and implementation of AI tools in education should have inte-
grated ethics and privacy protection. Policies need to be in place that ensure safety, fairness, and student
autonomy in the digital environment. Policies need to strike a balance between innovation and account-
ability. As much as AI can improve teaching itself, it is more important to preserve the human element
of teaching. The teacher is the main who should lead and control the teaching process and not to be
replaced by technology. Technology and pedagogy must support each other. The standardization of AI
in education must be multidimensional – technical, pedagogical, and ethical. Without teacher training, AI
cannot be properly used and exploited. Teachers must be motivated and properly guided in using AI and
its tools in teaching in order to derive the benefits of using it. Without proper training, the use of artificial
intelligence itself will be reduced to only the initial form of searching and using information.
Future research should move beyond the development of technical functionalities and focus on
several key directions:
• Longitudinal studies on the actual impact of AI-supported personalization on student performance,
engagement, and motivation in diverse educational contexts.
• Critical analysis of algorithmic decision-making to examine fairness and bias in adaptive learning environ-
ments, especially among students of different socio-economic backgrounds, ethnicities, or learning abilities.
• Evaluation of teacher-AI interaction models, including how AI tools influence teacher autonomy,
pedagogical strategies, and classroom dynamics.
• Exploration of students’ perceptions and emotional responses to AI-mediated learning, particu-
larly in terms of trust, agency, and well-being.
• Policy-oriented research aimed at developing governance frameworks that regulate the ethical de-
sign, deployment, and accountability of AI in education, with special attention to low-resource settings.
Implications for Future Research
Regardless of the positive effects and applications of artificial intelligence in education, far-reaching
research is certainly needed to have a true picture of the application of artificial intelligence in education.
The research itself should not encompass the technical capabilities of artificial intelligence, but must en-
compass ethical, pedagogical and socio-cultural dimensions. Certainly, a critical and reflective approach
will provide a better picture and better understanding of how artificial intelligence technology affects not
only educational outcomes, but also human relationships, student activity and institutional practice.
Key directions for future research include:
• Longitudinal impact studies to assess the sustained effects of AI-based personalization on learning
outcomes, engagement, and motivation across diverse educational levels and demographic groups.
• Critical analysis of algorithmic bias and transparency, particularly regarding how AI systems classify, recom-
mend, or assess students, and how such classifications may reproduce social or educational inequalities.
• Exploration of AI-teacher collaboration models, with emphasis on how AI can augment rather than
undermine teacher autonomy, pedagogical creativity, and professional judgment.
• Examination of student perspectives, focusing on trust, acceptance, resistance, and emotional re-
sponses to AI-mediated learning environments, particularly among vulnerable populations.
• Development of policy and governance frameworks that support the ethical deployment of AI in education, en-
suring data protection, accountability, and inclusivity, especially in low-income and rural educational settings.
Such interdisciplinary and multi-perspective research will contribute to the development of a just, hu-
www.ijcrsee.com
548
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
man-centered, and sustainable AI-enhanced educational ecosystem. Without addressing these broader issues,
technological solutions can become mere cutting-edge interventions and attempts to transform education.
Conflict of interests
The authors declare no conflict of interest.
Author Contributions
Conceptualization, L.S., A.R., formal analysis, O.K., B.B; methodology, O.M.; writing - original draft
preparation, L.S.; writing - review and editing, A.R., O.K., B.B., O.M. All authors have read and agreed to
the published version of the manuscript.
References
Al-Azawei, A., Serenelli, F., & Lundqvist, K. (2016). Universal Design for Learning (UDL): A content analysis of peer reviewed
journals from 2012 to 2015. Journal of the Scholarship of Teaching and Learning, 16(3), 39-56. https://doi.org/10.14434/
josotl.v16i3.19295
AlOqaily, A. N., Qawasmeh, E. F., & Tawalbeh, J. (2025). The effect of implementing AI on job burnout through the mediat-
ing role of work-life balance in the context of HRM. Economics - innovative and economics research journal, 13(2),
465–484. https://doi.org/10.2478/eoik-2025-0049
Binns, R. (2018, January). Fairness in machine learning: Lessons from political philosophy. In Conference on fairness, ac-
countability and transparency, 149-159. PMLR.
https://proceedings.mlr.press/v81/binns18a.html
Bjelajac, Ž, Filipović, A. M., & Stošić, L. (2023). Can AI be Evil: The Criminal Capacities of ANI. International Journal of Cog-
nitive Research in Science, Engineering and Education (IJCRSEE), 11(3), 519–531. https://doi.org/10.23947/2334-
8496-2023-11-3-519-531
Bjelajac, Ž. Filipović, A. & Stošić, L. (2022). Quis custodiet ipsos custodes: Ethical Dillemmas of the KM Governed by AI. In:
Proceedings of 4th Virtual International Conference Path to a Knowledge Society-Managing Risks and Innovation (Eds.
Stanković, M. & Nikolić, V.). Publishers Complex System Research Centre, Niš, Serbia; Mathematical Institute of the
Serbian Academy of Sciences and Arts. Serbia, Niš. 353-360. https://miteam.mi.sanu.ac.rs/asset/G3sjTM6rPPZThzpnM
Bjelajac, Ž., Filipović, A. M., & Stošić, L. V. (2022). Internet Addiction Disorder (IAD) as a Consequence of the Expansion
of Information Technologies. International Journal of Cognitive Research in Science, Engineering and Education
(IJCRSEE), 10(3), 155–165. https://doi.org/10.23947/2334-8496-2022-10-3-155-165
Burke, R. (2007). Hybrid Web Recommender Systems. In: Brusilovsky, P., Kobsa, A., Nejdl, W. (eds) The Adaptive Web. Lec-
ture Notes in Computer Science, 4321. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-72079-9_12
Deci, E. L., & Ryan, R. M. (2000). The” what” and” why” of goal pursuits: Human needs and the self-determination of behav-
ior. Psychological inquiry, 11(4), 227-268. https://doi.org/10.1207/S15327965PLI1104_01
Deschênes, M. (2020). Recommender systems to support learners’ Agency in a Learning Context: a systematic review. Inter-
national journal of educational technology in higher education, 17(1), 50. https://doi.org/10.1186/s41239-020-00219-w
Duckett, I. (2010). Personalized learning and vocational education and training. International encyclopedia of education (3rd
edn., 391–396). London: Elsevier Ltd.
Duong, T. T. T., Long, T. H., Anh, D. P., Dai, N. M. T., Ngan, D. N. K., & Xuan, N. T. T. (2025). The impact of education on energy
poverty: Empirical evidence from Vietnam. Economics - innovative and economics research journal, 13(2), 265–281.
https://doi.org/10.2478/eoik-2025-0043
Dwivedi, S., & Roshni, V. K. (2017, August). Recommender system for big data in education. In 2017 5th National Conference on
E-Learning & E-Learning Technologies (ELELTECH), 1-4. IEEE. https://ieeexplore.ieee.org/abstract/document/8074993
Filipović, A., Bjelajac, Ž., Merdović, B., & Stošić, L. (2023). Some aspects of the criminal potential of articial intelligence,
In: Proceedings of 4th Virtual International Conference Path to a Knowledge Society-Managing Risks and Innova-
tion, (Eds. Stanković, M. & Nikolić, V.). Publishers Complex System Research Centre, Niš, Serbia; Mathematical
Institute of the Serbian Academy of Sciences and Arts. Serbia, Niš. 41-47. https://paksom.cosrec.org/wp-content/up-
loads/2024/01/PaKSoM_2023.pdf
Graf, S., Liu, T. C., & Kinshuk. (2010). Analysis of learners’ navigational behaviour and their learning styles in an online course.
Journal of Computer Assisted Learning, 26(2), 116-131.
https://doi.org/10.1111/j.1365-2729.2009.00326.x
Holmes, W., Bialik, M., & Fadel, C. (2019). Articial intelligence in education promises and implications for teaching and learn-
ing. Center for Curriculum Redesign.
https://discovery.ucl.ac.uk/id/eprint/10139722/
Holmes, W., Porayska-Pomsta, K., Holstein, K., Sutherland, E., Baker, T., Shum, S. B., ... & Koedinger, K. R. (2022). Ethics of
www.ijcrsee.com
549
Stošić, L. et al. (2025). Personalized Learning through Artificial Intelligence: Opportunities, Risks, and Policy Perspectives,
International Journal of Cognitive Research in Science, Engineering and Education (IJCRSEE), 13(2), 541-549.
AI in education: Towards a community-wide framework. International Journal of Articial Intelligence in Education, 1-23.
https://doi.org/10.1007/s40593-021-00239-1
Kamardeen, I. (2014). Adaptive e-tutorial for enhancing student learning in construction education. International Journal of
Construction Education and Research, 10(2), 79-95. https://doi.org/10.1080/15578771.2012.756437
Karimi-Mamaghan, M., Mohammadi, M., Meyer, P., Karimi-Mamaghan, A. M., & Talbi, E. G. (2022). Machine learning at the
service of meta-heuristics for solving combinatorial optimization problems: A state-of-the-art. European Journal of
Operational Research, 296(2), 393-422.
Khawaja, M. A., Chen, F., & Marcus, N. (2014). Measuring cognitive load using linguistic features: implications for usabil-
ity evaluation and adaptive interaction design. International Journal of Human-Computer Interaction, 30(5), 343-368.
https://doi.org/10.1080/10447318.2013.860579
Knox, J. (2020). Articial intelligence and education in China. Learning, Media and Technology, 45(3), 298-311.
https://doi.org
/10.1080/17439884.2020.1754236
Loewen, S., Isbell, D. R., & Sporn, Z. (2020). The effectiveness of app-based language instruction for developing receptive linguis-
tic knowledge and oral communicative ability. Foreign Language Annals, 53(2), 209-233. https://doi.org/10.1111/an.12444
Luckin, R., & Holmes, W. (2016). Intelligence Unleashed: An argument for AI in Education. UCL Knowledge Lab: London, UK.
https://discovery.ucl.ac.uk/id/eprint/1475756/
Nadine, U., Cao, H., & Deng, J. (2016, August). Competitive recommendation algorithm for E-commerce. In 2016 12th Inter-
national Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD), 1539-1542.
IEEE.
https://ieeexplore.ieee.org/abstract/document/7603404/
Nirmalakhandan, N. (2007). Computerized adaptive tutorials to improve and assess problem-solving skills. Computers & Edu-
cation, 49(4), 1321-1329.
https://doi.org/10.1016/j.compedu.2006.02.007
Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. In Algorithms of oppression. New York
university press.
O’ Neil, C. (2017). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown.
Regan, P. M., & Jesse, J. (2019). Ethical challenges of edtech, big data and personalized learning: Twenty-rst century student
sorting and tracking. Ethics and Information Technology, 21, 167-179. https://doi.org/10.1007/s10676-018-9492-2
Selwyn, N. (2019). Should robots replace teachers?: AI and the future of education. John Wiley & Sons.
Stošić, L., & Janković, A. (2024). Obrazovanje mladih u svetu digitalnih tehnologija [Educating young people in the world
of digital technologies], Pravni fakultet za privredu i pravosuđe u Novom Sadu, Fakultet za ekonomiju i inženjerski
menadžment u Novom Sadu, Srbija
Stošić, L., & Janković, A. (2023). The impact of articial intelligence (AI) on education – balancing advancements and ethical
considerations on human rights. Pravo - Teorija i Praksa, 40(4), 58–72.
https://doi.org/10.5937/ptp2304058S
Suresh, V., Kolluru, M., & Ubaidullah, V. (2025). Encoding behavior commonalities in global stock market indexes: Unsuper-
vised Machine Learning approach. Economics - innovative and economics research journal, 13(2), 283–303.
https://
doi.org/10.2478/eoik-2025-0041
Talapbayeva, G., Yerniyazova, Z., Kultanova, N., Akbayev, Y. (2024). Economic Opportunities and Risks of Introducing Articial
Intelligence. Collection of papers new economy volume, 2(1), 117-128.
https://doi.org/10.61432/CPNE0201117t
Urdaneta-Ponte, M. C., Mendez-Zorrilla, A., & Oleagordia-Ruiz, I. (2021). Recommendation systems for education: Systematic
review. Electronics, 10(14), 1611.
https://www.mdpi.com/2079-9292/10/14/1611
VanLehn, K. (2011). The Relative Effectiveness of Human Tutoring, Intelligent Tutoring Systems, and Other Tutoring Systems.
Educational Psychologist, 46(4), 197–221.
https://doi.org/10.1080/00461520.2011.611369
Venkatraman, S. (2017, November). A proposed business intelligent framework for recommender systems. Informatics, 4(4),
40;
https://doi.org/10.3390/informatics4040040
Vesselinov, R., & Grego, J. (2012). Duolingo effectiveness study. City University of New York, USA, 28(1-25). http://com-
parelanguageapps.com/DuolingoReport_Final.pdf
Wei, G., & Niemi, H. (2023). Ethical guidelines for articial intelligence-based learning: a transnational study be-
tween china and nland. Learning: Designing the Future, 265. https://library.oapen.org/bitstream/han-
dle/20.500.12657/60151/978-3-031-09687-7.pdf?sequence=1#page=276
Weltman, H. (2025). A Pedagogical Evaluation of Online Adaptive Tutorials to Support Teaching and Learning (Doctoral dis-
sertation, UNSW Sydney). https://doi.org/10.26190/unsworks/31158
Williamson, B., & Eynon, R. (2020). Historical threads, missing links, and future directions in AI in education. Learning, Media
and Technology, 45(3), 223-235. https://doi.org/10.1080/17439884.2020.1798995
Zawacki-Richter, O., Marín, V. I., Bond, M., & Gouverneur, F. (2019). Systematic review of research on articial intelligence
applications in higher education–where are the educators?. International journal of educational technology in higher
education, 16(1), 1-27. https://doi.org/10.1186/s41239-019-0171-0