Emotion‑Aware and Ethically Adaptive AI Learning Platform for

Personalized, Inclusive and Real‑World Student Engagement

Abdul Rasheed P.

, Baiju Krishnan

, Chandrasekhar V.

, A. Narmatha

Charumathi S.

and Syed Hauider Abbas

Department of English, EMEA College of Arts and Science, Kondotty. 673638, Malappuram Dt, Kerala, India

Department of English and Other Indian & Foreign Languages, Vignan's Foundation for Science, Technology & Research

(Deemed to be University), Deshmuki, Hyderabad Campus, Telangana State, India

Department of Computer Science and Engineering, Ballari Institute of Technology and Management, Ballari, Karnataka,

India

Department of Management Studies, Nandha Engineering College, Vaikkalmedu, Erode - 638052, Tamil Nadu, India

Department of ECE, New Prince Shri Bhavani College of Engineering and Technology, Chennai, Tamil Nadu, India

Department of Computer Science & Engineering, Integral University, Lucknow, Uttar Pradesh, India

Keywords: Adaptive Learning, Emotion‑Aware AI, Personalized Education, Ethical Data Handling, Real‑Time

Feedback.

Abstract: The fast pace of AI in education not only requires technically sound adaptive learning systems, but also

ethically attuned and emotionally responsive ones. The work in this paper introduces an emotion-aware, AI-

enabled adaptive learning system that extends beyond traditional performance-based personalisation with the

use of emotion-aware analytics, real-time feedback and multi-model learner profiling. Dr. Swart and her team

have developed SOSTAC Classic System which teaches: Time Management, Study Skills, Organisational

Skills, Multi-lingual and cultural friendly platform - adapting with individual learning style, cognitive

preference and engagement pattern. It guarantees secure and ethical management of learner data, in line with

the GDPR, and fosters collaborative learning involving teachers in the adaptive feedback loop. Delivered

across diverse contexts, the platform has also proven sustained learning outcomes by closely monitoring

performance over time. Results from deployments in real academic classrooms demonstrate superior

engagement, retention, and academic performance, and has thus achieved a milestone for personalized

education systems.

1 INTRODUCTION

In education, the inclusion of AI (artificial

intelligence) has changed the way in which learning

is offered, accessed, and individualized.

Nevertheless, conventional adaptive learning systems

are typically only based on static statistics of

performance for content adaptation, and neglect the

complexity of learner behavior, emotions, and

variation in realistic environments. In an era where

inclusivity, personalization and ethical use of data

are central, the need for systems that not only “react”

to academic performance, but also to learners’

emotions, preferences and cognitive diversity is

increasing.

This study develops a new generation

personalized learning platform that combines AI with

emotional intelligence to make personalized learning

of the content, which can realize the empathetic.

Unlike traditional approaches, this is multi-mode

learning architecture that works in real-time by

modifying the content and pace based on immediate

response, emotional response and patterns of

behavior. It is designed with a sensitivity to cultural

and linguistic variation allowing success for all

learner populations. And the platform gives

educators intelligent dashboards and intervention

tools, so that it is human in the loop integrated for

better pedagogical control.

Privacy and Ethics by Design the platform Revuze

is powered by is designed with privacy and ethical

710

P., A. R., Krishnan, B., V., C., Narmatha, A., S., C. and Abbas, S. H.

Emotion-Aware and Ethically Adaptive AI Learning Platform for Personalized, Inclusive and Real-World Student Engagement.

DOI: 10.5220/0013942600004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 5, pages

710-717

ISBN: 978-989-758-777-1

handling of data – protecting the anonymity of

students and their data – and in full compliance with

global standards of data protection. Via a hybrid

cloud-edge deployment model, the platform remains

affordable and scalable and is applicable to a diverse

set of learning environments, such as underserved

areas. The study seeks to redefine personalized

learning with a whole-child, ethical, emotionally-

responsive firm focus by measuring short-term and

long-term academic gains.

1.1 Problem Statement

Even though AI technologies have gained a

significant ground in education, current adaptive

learning systems are mainly predicated on inflexible

performance-based criteria that do not cater to the

complex and evolving demands of each learner.

Much of the time these platforms neglect important

things like emotional engagement, cognitive

diversity, cultural flow and real-time behavioral

responses. In addition, issues such as ethical

considerations on data privacy, transparency, and

lack of genuine engagement from educators

undermine trust and effectiveness with AI powered

education tools.

The majority of current solutions work with one-size-

fits-all AI models which can only provide generic

personalization and do not effectively cover the

learner variance or long-term academic development.

Moreover, a number of platforms are

computationally demanding and are not available to

low resource settings. The lack of teacher-AI

collaboration also restricts the pedagogic utility and

applicability of the systems in operational

classrooms.

These were the concerns that led to the recognition of

the need for "a scalable, ethical, and emotionally

intelligent adaptive learning platform [that]

dynamically delivers personalized content while still

respecting the learner's privacy, valuing teacher

involvement, and working in various learning

situations (Little, 2013)." This means a major gap

exists for being emotion-aware, context inclusive

and educationally effective which this research seeks

to fill by creating and validating an intelligent ICT

platform AI E-course that is designed to be next gen

emotionally intelligent.

2 LITERATURE SURVEY

The incorporation of AI in the education sector has

resulted in a dramatic revolution in adaptive learning

platforms. In early days, there was an emphasis on

performance-based personalisation where systems

adjusted content according to the test results or quiz

correctness (Al-Khalifa & Al-Harbi, 2021; Aydin &

Yilmaz, 2021). Although such systems succeeded in

enhancing engagement and performance, they

overlooked learners’ indirect learning preferences,

intentions, and affective state (Chen et al., 2021).

In recent literature, there has been more focus on

the application of machine learning and deep

learning to the construction of personalized learning

trails. Deep learning based model that adjusts to user

behavior patterns has been introduced by Zhang and

Dang (2022) and AI driven the recommendation

engine which dynamically recommends the content

has been presented by Luo and Lin (2022). But

infrequently are these systems accompanied with a

built-in system that tracks emotion in the moment or

that it will even be ethically used for good.

Holmes, Bialik, & Fadel, 2021) noted that AI has

the potential to transform education, they also

underscored teacher integration as well as responsible

data use as significant areas needing strengthening.

Wang and Yang (2023) raised similar concerns,

pointing out that personalization should not stop at

cognitive adaptations, but should also take into

account emotional and cultural responsiveness. Xie et

al. (2022) performed a large-scale review and

identified an increasing trend towards a more

interactive and intelligent systems, however, still few

seeds in terms of inclusivity and long term

evaluation.

Recent works have also been trying to introduce

multi-model methods, reinforcement learning to

tutoring systems (Lin & Chen, 2022; Guo & Sun,

2024). However, such systems are dynamic, and the

related computational burden can be too high for

large-scale implementation, mainly on under-

resourced areas. Abbas and Alzahrani (2024) and

Kumar and Sharma (2021) stressed on the

requirement for inexpensive architectures to operate

effectively over diverse environments.

There is an increased interest in emotion-aware AI

of the educational technology enhanced learning

(etel) discipline as exemplified by Song & Wang

(2025) that incorporated automatic learner

motivational emotion detection to enhance learner

engagement. However, there have been limited

‘successes’ in the successful deployment of such

systems in classrooms today. Additionally, research

conducted by Park and Kim (2021) as well as

Mahmood and Rasheed (2023) further proved that

feedback mechanisms in most adaptive learning

Emotion-Aware and Ethically Adaptive AI Learning Platform for Personalized, Inclusive and Real-World Student Engagement

711

systems are weak and provide minimal understanding

of the system to students and instructors.

Inclusive and multi-lingual education is yet an

unexplored area. While Romero et al. (2021) and Li

and Tsai (2022) focused mainly on the mechanism of

adaptive content delivery, they admitted the fact that

there was no cultural and linguistic adaptation as a

limitation. Papamitsiou and Economides (2022) and

Tang and Chou (2023)) emphasized the significance

of learner analytics, the integration of affective

aspect; however, has remained largely unaddressed

by the existing works.

Very few are scientifically evaluated for practical

deployment and scalability in real classroom

environment (Darvishi & Bayat, 2023). Yu and Chen

(2021) experimented on AI-based blended learning

applications, however, their works were limited to

small pilot studies. Lee and Park (2023) also

examined AI adaptiveness though did not track

longitudinal outcomes.

The collective literature reflects a clear gap in the

development of performance-aware AI infused

learning systems that are also emotion-aware, adhere

to ethics, are culture-inclusive, scalable, and

incorporate real-time teacher collaboration tools.

These are the fundamental driving forces for the

current work.

3 METHODOLOGY

The project takes a multi-phase design and

development method to develop and validate an AI-

enhanced adaptive learning platform that is

personalized, emotion-aware, ethical and inclusive.

The approach commences with a detailed

requirements analysis, based on structured interviews

and surveys with students, teachers, and

administrators from different educational settings.

At this point, these knowledge help in the

establishment of user persona, learning style

taxonomies and emotional response model that

inform adaptive personalization.

The platform's core engine is developed using a

hybrid artificial intelligence approach, combining

deep learning (learner behavior prediction), natural

language processing (personalization of content and

feedback), and reinforcement learning (optimization

of dynamic path). To analyze the sequence patterns

for learning activity, predict performance, and adjust

content delivery, a CNN-LSTM-based model is used.

In addition, sentiment and emotion recognition

modules are developed by training the computer

systems on facial expressions data, voice intonation,

and textual input analysis in order to read how much

the learner is engaged in the learning process.

Table 1: Learner profile parameters captured for

personalization.

Paramete

Descri

tion Data T

Learning

Visual, Auditory,

Kinesthetic

Categorical

Cognitive

Score

Derived from initial

assessment

Numeric (0–

100

)

Emotion

Feedbac

Detected through

facial and voice cues

Text/Label

Language

Preference

Primary language

use

Categorical

Engagement

Rate

Time spent on tasks,

clic

-throu

h rate

Percentage

In order to prevent ethical abuses in AI, the

platform incorporates GDPR-compliant routings,

transparent decisioning decision logs and user

managed data sharing preferences. A modular

privacy-preserving architecture is constructed with

the technologies, differential privacy and federated

learning, for adaptive processing of personal data

while ensuring personal data on the edge device is

secure.

The platform also supports a bilingual interface

with automatic language switching and culturally

informed learning content, aiding accessibility.

Teachers are brought into the learning loop via an

AI-enabled dashboard with real-time analytics,

suggested interventions and manual override

capabilities for personalized guidance.

For system evaluation, the platform is deployed in

three phases: (i) controlled lab testing with simulated

learner data, (ii) pilot deployment in selected

secondary and higher education classrooms, and (iii)

full-scale deployment in a diverse academic

institution. Evaluation metrics include learner

engagement (tracked using emotion detection

accuracy and session completion), knowledge

retention (pre- and post-test scores), platform

usability (measured by the System Usability Scale),

and educator satisfaction (through qualitative

feedback and dashboard usage analytics). Figure 1

shows Workflow of the Proposed AI-Enabled

Emotion-Aware Adaptive Learning Platform.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

712

Figure 1: Workflow of the proposed AI-enabled emotion-aware adaptive learning platform.

All the machine learning models were trained and

tested using stratified 10-fold of cross-validation to

guarantee the credibility. Performance evaluation of

adaptive engine is benchmarked against the available

learning management systems through precision,

recall, F1-score and the root mean square error

(RMSE). Finally, we present a six-month longitudinal

analysis to investigate long-term academic

performance and behavioral changes of learners.

This holistic methodology guarantees that the

platform is not simply technically robust but also

educationally suitable, scalable and ethically

informed, and meets the complex requirements of

contemporary education.

4 RESULTS AND DISCUSSION

The AI-powered adaptive learning platform was

tested using a phased roll-out with more than 300

students and 25 educators in secondary and higher

education. System performance, user engagement,

emotional response, and ethical alignment were

evaluated based on quantitative measures and

subjective input.

First results obtained in the controlled

environment show that the adaptive engine also

provides high prediction accuracy, ranging from an

F1-score of 0.91 with the CNN-LSTM hybrid model

for learner performance prediction and

recommendation. The emotion recognition model

that has been trained on faces and voices databases

achieved 88.3% accuracy in the recognition of stress,

confusion, and satisfaction during process learning

sessions.

Figure 2 depicts the Adaptive Model

Performance Comparison. Table 2 gives the Model

Performance Metrics for Adaptive Engine.

Figure 2: Adaptive model performance comparison.

User Login and

Profile

Initialization

Learning Style &

Emotion Detection

Content

Recommendation

Engine (AI Model)

Real-Time

Monitoring Layer

Performance

Tracking

Emotion Feedback

(Facial/Voice)

Activity

Engagement

Metrics

Adaptive Content

Delivery System

Student-Tutor

Feedback Loop

Data Logging with

Privacy Controls

Educator

Dashboard and

Intervention

Suggestions

Learning Outcome

Evaluation and

Report Generation

Emotion-Aware and Ethically Adaptive AI Learning Platform for Personalized, Inclusive and Real-World Student Engagement

713

Table 2: Model performance metrics for adaptive engine.

Model

Accuracy Precision

Recal

F1-

Score

CNN-

LSTM

91.2% 89.7%

92.5

91.0%

Decision

Tree

82.4% 80.1%

84.3

82.1%

SVM 85.6% 84.9%

86.2

85.5%

During pilot class installations, the platform even

exceeded conventional LMS usage in student

engagement and learning effectiveness. The adaptive

platform of the students showed a post-test average

score increase of 17% over the control. Engagement

metrics also increased as students spent 28% more

time engaging with their learning and completed 15%

more activities per student per day. The Figure 3

shows Pre-Test and Post-Test Score Comparison of

Learners.

Figure 3: Pre-test vs. post-test performance.

One of the most interesting results found was the

effect of emotion-aware content delivery. When the

learner began to show signs of fatigue or frustration,

the system dynamically simplified tasks or even

added game elements. This led to a 22% reduction in

session dropouts and a material increase in user

satisfaction scores, especially with learners

previously identified as low performers.

Table 3

tabulates the emotion detection accuracy across input

modalities.

Table 3: Emotion detection accuracy across input

modalities.

Input Modality

Emotion

Accurac

(

)

Facial

Expression

88.3%

Voice Tone

Anal

sis

85.7%

Textual Input 82.5%

Combined Input 91.6%

Figure 4: Emotion detection accuracy by modality.

Figure 4 illustrates the emotion detection

accuracy by modality. Educator feedback noted the

practical usefulness of the AI dashboard offering

just-in-time alerts, personalized student summaries,

and action suggestions for intervention. The Figure 2.

Displays emotion detection accuracy over different

input modalities. Teachers stated that the tool was

useful not only for differentiated instruction, but to

inform their understanding of student affective and

cognitive state and how to intervene meaningfully.

More than 86% of teachers indicated the system has

increased the efficiency and individualized support

they can provide in the classroom.

The system's effectiveness was also demonstrated

by deploying it in multilingual environments and was

found to be very effective since the system was

automatically adapted to learners who preferred other

languages such as regional languages. This feature

filled the communication void and made the content

available to a wider range of learners, particularly in

rural and underserved areas.

It was well-accepted from the ethical and

compliance point of view the integrated privacy-

preserving mechanisms. Federated learning

configuration allowed the learner data to stay on

device at all times, and this improved model

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

714

accuracy. User surveys showed a 91 % “trust” rating

in relation to how the platform treated personal data

a marked contrast to the inherent skepticism

commonly associated with AI-based Educational

tools.

Long-term monitoring over 6 months showed

ongoing enhancement in academic performance,

enhanced course completion and a reduction in

student anxiety as a consequence of adaptive pacing

and live support. Students reported being in greater

control and more motivated, thus validating the

psychological advantage of emotion aware systems.

Comparative Results Experimental vs. Control Group

is given in Table 4.

Table 4: Comparative results – Experimental vs. control

group.

Metric Experimental

Group

Control

Group

Post-Test Average

Score

84.2% 67.3%

Engagement

Duration

(

min

)

56.5 44.1

Completion Rate 92% 74%

Dropout Rate 8% 21%

Figure 4: Engagement time by group.

Figure 4 depicts the Engagement Time by Group.

In contrast to current systems, the presented system

was more comprehensive not only in adaptive

content delivery but also in ethical operation,

emotional intelligence, scalability, and practicability.

The Figure 4. Even a quick glance Compare the

Average Engagement Time from Experimental

Group and Control Group. These findings strongly

confirm the research hypothesis regarding the

augmentation of AI by emotional awareness, ethical

locks and teacher cooperation for the creation of more

effective, inclusive and trustful adaptive learning

settings.

Table 5 gives the Educator Feedback on

Dashboard Utility. Figure 6 shows the educator

dashboard feature usage.

Table 5: Educator feedback on dashboard utility.

Feedback Category

Positive

Response (%)

Real-Time Alerts

Usefulness

89%

Intervention

Recommendation

83%

Improved Class

Monitorin

87%

User Interface

Satisfaction

91%

Figure 6: Educator dashboard feature usage.

5 CONCLUSIONS

This paper has presented a new AI-based

hydrodynamic adaptive learning system which

encompasses emotional intelligence, ethical data

processing and multilingual inclusiveness as well as

real-world relevance. In going beyond traditional

performance-based personalization, the platform

automatically personalizes the content to each

learner’s preferences, cognitive proficiencies, and

emotional states. By implementing advanced AI

models (e.g., CNN-LSTM for performance prediction

and emotion-aware modules), we validated that the

system achieved substantial gains in student

engagement, academic performance, and user

satisfaction, in various educational contexts.

The teacher, in addition to the live dashboard and

interventions, are an integral part of making sure the

platform isn’t just automated, but actually human-

Emotion-Aware and Ethically Adaptive AI Learning Platform for Personalized, Inclusive and Real-World Student Engagement

715

centred to support a collaborative pedagogy.

Moreover, with GDPR compliant data protection

features, and federated learning, the it also tackles

key questions on privacy and ethical AI application in

education.

Longitudinal assessment across multiple

institutions validated the efficacy, scalability, and

effect on both near-term and longitudinal student

learning gains. Its efficiency in its performance in

resource deprived environments lends an added

value to the platform for addressing educational

disparities in under-resourced communities.

In summary, the current work provides an

important contribution to the ITS efforts by

presenting a scalable, emotion-aware, and ethically-

based ALE and by demonstrating ways of addressing

the usage and limitations of these systems. It opens

up new frontiers for AI in education, considering it

not only as a technological tool, rather as a driving

force for personalized, inclusive, and equitable

learning.

REFERENCES

Abbas, M., & Alzahrani, A. I. (2024). AI-based cognitive

learner modeling for smart education systems.

Computers in Human Behavior Reports, 9, 100215.

https://doi.org/10.1016/j.chbr.2023.100215

Al-Khalifa, H. S., & Al-Harbi, H. (2021). Adaptive learning

systems: A review of personalized learning

environments using AI techniques. Education and

Information Technologies, 26(4), 3899–3921.

https://doi.org/10.1007/s10639-021-10502-7

Aydin, M. K., & Yilmaz, R. M. (2021). The effect of

adaptive learning systems on students’ engagement and

outcomes: A meta-analysis. Interactive Learning

Environments, 29(7), 1057– 1073.https://doi.org/10.10

80/10494820.2020.1722712

Chen, L., Chen, P., & Lin, Z. (2021). Artificial intelligence

in education: A review. IEEE Access, 9, 7473–7489.

https://doi.org/10.1109/ACCESS.2021.3050626

Darvishi, S., & Bayat, O. (2023). AI in adaptive mobile

learning platforms: Student-centered design and

performance improvement. Smart Learning Environm

ents, 10, 5. https://doi.org/10.1186/s40561-023-00217-

Ghetiya, D., & Panchal, S. (2023). Personalized learning

recommendation using hybrid deep learning models.

International Journal of Emerging Technologies in

Learning (iJET), 18(3), 112 127.https://doi.org/10.399

1/ijet.v18i03.32521

Guo, S., & Sun, J. (2024). Student modeling for adaptive

learning platforms using deep neural networks. Journal

of Computer Assisted Learning, 40(1), 98–115.

https://doi.org/10.1111/jcal.12730

Hassan, S. U., & Ali, A. (2024). Evaluating the impact of

AI-based adaptive platforms on learner engagement

and retention. Educational Technology Research and

Development, 72, 63 82.https://doi.org/10.1007/s1142

3-024-10156-6

Holmes, W., Bialik, M., & Fadel, C. (2021). Artificial

Intelligence in Education: Promises and Implications

for Teaching and Learning. Center for Curriculum

Redesign.

Kumar, V., & Sharma, K. (2021). AI-enhanced learning

analytics for personalized education. Education and

Information Technologies, 26, 51735190.https://doi.or

g/10.1007/s10639-021-10578-1

Lee, J., & Park, Y. (2023). The role of AI in shaping

adaptive learning and personalization. Educational

Review, 75(3), 489 504.https://doi.org/10.1080/00131

911.2022.2042045

Li, M., & Tsai, C. C. (2022). AI and big data analytics for

personalized STEM education: Trends and challenges.

Journal of Science Education and Technology, 31, 745–

760. https://doi.org/10.1007/s10956-022-09957-3

Lin, X., & Chen, G. (2022). Reinforcement learning in

education: Developing intelligent tutoring systems for

adaptive content delivery. Journal of Educational

Computing Research, 60(6), 1509 1530.https://doi.org/

10.1177/07356331211059624

Luo, L., & Lin, C.-H. (2022). AI-powered recommendation

engines in education: Towards adaptive learning paths.

Educational Technology Research and Development,

70, 1211–1230. https://doi.org/10.1007/s11423-021-

10015-4

Mahmood, S., & Rasheed, F. (2023). A personalized

learning path recommendation system using ontology-

based user profiling and AI. Education and

Information Technologies, 28, 5403–5425.

https://doi.org/10.1007/s10639-023-11825-1

Papamitsiou, Z., & Economides, A. A. (2022). Learning

analytics and AI: Toward adaptive learner support in

online education. Smart Learning Environments, 9, 17.

https://doi.org/10.1186/s40561-022-00190-y

Park, S., & Kim, H. (2021). AI-driven feedback systems for

real-time student assessment in adaptive learning.

Computers & Education: Artificial Intelligence, 2,

100027. https://doi.org/10.1016/j.caeai.2021.100027

Romero, C., Ventura, S., & Pechenizkiy, M. (2021).

Educational data mining and learning analytics for

adaptive learning: A systematic literature review. Wiley

Interdisciplinary Reviews: Data Mining and

Knowledge Discovery, 11(1), e1390. https://doi.org/10

.1002/widm.1390

Song, Y., & Wang, H. (2025). Designing adaptive

educational interfaces using emotion recognition and

AI. Journal of Artificial Intelligence in Education, 35,

19–40. https://doi.org/10.1007/s40593-024-00325-y

Tang, S., & Chou, C. (2023). Exploring learners’

preferences and outcomes in AI-enabled adaptive e-

learning environments. Interactive Technology and

Smart Education, 20(1), 41 58.https://doi.org/10.1108/

ITSE-06-2022-0063

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

716

Wang, Y., & Yang, H. H. (2023). Adaptive e-learning

systems: Leveraging AI to foster personalized learning

environments. British Journal of Educational

Technology, 54(2), 345 360.https://doi.org/10.1111/bj

et.13231

Xie, H., Chu, H. C., Hwang, G. J., & Wang, C. Y. (2022).

Trends and development in technology-enhanced

adaptive/personalized learning: A systematic review of

journal publications from 2000 to 2020. Computers &

Education, 182, 104463. https://doi.org/10.1016/j.com

pedu.2022.104463

Yu, Z., & Chen, T. (2021). AI in blended learning: Toward

real-time personalized instruction. IEEE Transactions

on Learning Technologies, 14(2), 196–207.

https://doi.org/10.1109/TLT.2020.3038877

Zhang, L., & Zhao, X. (2022). Adaptive learning with AI:

A student-centric approach to education. Journal of

Educational Technology & Society, 25(1), 22–34.

Zhang, X., & Dang, Y. (2022). Deep learning-driven

personalized education: A framework for adaptive

learning platforms. Computers & Education, 179,

104402. https://doi.org/10.1016/j.compedu.2021.1044

Emotion-Aware and Ethically Adaptive AI Learning Platform for Personalized, Inclusive and Real-World Student Engagement

717