Article

Eliel Da Silva Souza; Washington Maciel da Silva

DOI: 10.22167/2675-6528-202603623

Article derived from a Final Course Work (TCC), with content based on the student’s original work and adapted to the editorial format of the Revista E&S with the support of the ResumeAI tool, an artificial intelligence solution developed by the Pecege Institute for textual synthesis and organization.

Abstract

The intersection between neuroscience and Artificial Intelligence (AI) in the learning process was explored. The study aimed to analyze the impact of technological mediation on students’ academic performance. A mixed method was employed, combining documentary content analysis and inferential statistics on secondary data. The investigation was based on a survey of consolidated databases, ensuring anonymity and absence of direct contact with students. Reports from AI platforms, pedagogical records from class councils, and academic performance indicators, such as grades and attendance, were analyzed. Quantitative data, from 40 records, revealed a significant impact of technological mediation. An increase of approximately 46.6% in the overall performance average was observed, which rose from 4.93 before the intervention to 7.23 after the implementation of AI tools. The Wilcoxon signed-rank test confirmed the statistical relevance of this change (p=0.001), indicating that the improvement resulted from the intervention. Additionally, a linear regression model pointed out that the intervention introduced new determinants of change, independent of users’ prior knowledge level. The neuroscientific perspective suggested that Digital Educational Resources (DER) promote autonomy, but can cause cognitive atrophy if used solely to automate functions. It was concluded that AI represents a valuable adaptive support, whose effectiveness is linked to the teacher’s intervention to ensure critical and consolidated learning.

Keywords: Personalized learning; Educational artificial intelligence; Cognitive neuroscience; Educational technologies.

1. Introduction

Artificial Intelligence (AI), studied in the field of Computer Science, is configured as an interdisciplinary area that connects to several other knowledge disciplines. Its central purpose is to develop systems capable of imitating human intelligence, encompassing abilities such as reasoning, learning, making inferences, and adapting. To achieve these objectives, AI integrates knowledge from neurosciences, theory of mind, linguistics, and statistics (Catania, 2021).

In this context, neuroscience offers fundamental principles on how the human brain learns, memorizes, and reacts to stimuli. In parallel, AI enables the creation of tools that adjust to the pace and individual needs of each student. AI-based digital educational resources (DER), inspired by brain function, demonstrate significant potential to personalize and enhance the teaching-learning process.

In the conception of Sosa-Alonso et al. (2025), technological instruments developed specifically to assist and optimize learning processes are seen as promising alternatives in the school environment. Information and Communication Technologies (ICT) encompass a set of digital tools that offer support to classes, while Digital Educational Resources (DER) comprise a wide range of materials, including digital textbooks, interactive learning platforms, adaptive learning environments, and educational robotics.

Despite the growing ubiquity of technology, Moreno and Heidelmann (2017) argue that, often, these digital resources represent an additional challenge for teachers, rather than an immediate solution. This highlights one of the main challenges of the new millennium: the need to integrate educational practice with the efficient use of ICTs. Although most teachers in public (92%) and private (95%) schools in Brazil have internet access, the use of these tools is often limited to basic activities, such as research, social media, and downloading materials.

The massive insertion of digital tools in education has caused changes in the traditional teaching-learning method in basic education. However, the application of AI resources to personalize education, especially in chemistry teaching, has received little attention in schools. Science teaching, in general, is insufficiently emphasized in Brazilian public schools, which results in a significant lag in knowledge necessary for understanding natural phenomena and their associated technologies.

Ferreira et al. (2026) corroborate that science education in schools faces profound challenges, culminating in a knowledge gap and the neglect of this area in relation to other demands. It is common for Portuguese and mathematics subjects to be prioritized, as if learning science were not essential for the human and social development of students. Faced with this scenario, personalized learning, according to Nja et al. (2024), through AI-based active methodologies, which adapt learning experiences to the individual needs, strengths, weaknesses, and interests of each student, emerges as a relevant approach.

Therefore, this study is justified by the need to understand users’ perception and the real impacts of Artificial Intelligence technologies in the educational environment, as well as to foster critical reflection on the benefits and ethical challenges of their adoption. This research is relevant both for pedagogical innovation and for the formation of citizens aware of the responsible use of technology. In this context, the present work aimed to explore the intersection between neuroscience and Artificial Intelligence in the learning process, seeking to understand users’ perception and the real impacts of these technologies in the educational environment, with a focus on chemistry for high school students.

2. Material and Methods

This study employed a mixed methodological approach, combining qualitative and quantitative elements to investigate the relationship between neurosciences, artificial intelligence, and the learning process. The research was characterized as a survey of consolidated databases, aiming to generate knowledge applicable to educational challenges. The unit of analysis consisted of records from classes of 3rd-year high school students in chemistry, who had already used technological tools. The focus was on the analysis of secondary data and pre-existing institutional records, with no direct interaction with participants, ensuring anonymity and absence of contact, according to the established methodological design.

The investigation was structured in three main phases of documentary analysis. The first, Retrospective Diagnosis, involved collecting aggregated data on study habits and academic performance, including grades and attendance, recorded in the institution’s system before the implementation of Artificial Intelligence practices. This process established an initial overview of student performance.

The second phase, Documentary Intervention, consisted of the analysis of reports on the use of adaptive platforms and educational chatbots, already archived. Information was gathered on usage time, types of tools accessed, and engagement automatically recorded by the system. There was no contact with students during this phase.

In the third phase, Impact Assessment Via Secondary Data, a statistical comparison was made between the academic performance indicators collected in the diagnostic phase and the results obtained after using the AI tools. Faculty perceptions of engagement and technological challenges were extracted from class council reports or already finalized and archived pedagogical documents, maintaining the premise of zero interaction.

The data collection instruments were exclusively pre-existing institutional and digital records, organized into three fronts. The first front included usage reports automatically generated by AI platforms and virtual learning environments. From these, statistical data were extracted on frequency, time spent, interactions with chatbots, and volume of exercises, in addition to reports on correct and incorrect answers from adaptive platforms.

The second collection front covered pedagogical records archived by the coordination. These included descriptive opinions and minutes of class council meetings, with teacher perceptions on engagement and technological challenges. Digital class diaries were also consulted, containing teachers’ notes and observations on classroom behavior and dynamics.

The third front involved the collection of academic performance indicators, extracted directly from the school management system. Grades from formative and summative assessments from previous periods were collected for statistical comparison, and attendance records were used to verify possible correlations with the use of technological tools and school attendance. The questionnaire in Appendix B.1, which maps the perception and use of AI, was considered one of the archived documents analyzed.

The data analysis was divided into qualitative and quantitative approaches. The qualitative analysis, of documentary content, was applied to pre-existing written records. Pedagogical reports from class councils and teacher observation records were examined to identify patterns regarding the behavior and challenges of using AI in the school environment, using the thematic axes technique. The information was triangulated among different documentary sources to confirm the observed trends.

The quantitative analysis, of a descriptive and inferential statistical nature, was applied to the numerical information obtained from the databases. Means, frequencies, and percentages of engagement and performance were calculated. To compare academic performance before and after the use of technology, the Wilcoxon Test for paired samples was employed, due to the non-normality of the data. Correlation analysis was also performed to examine the connection between digital engagement and variations in recorded grades. All procedures ensured the absolute anonymity of the data, in accordance with the ethics protocol.

3. Results and Discussion

The initial analysis of the research revealed the widespread presence of Artificial Intelligence (AI) in students’ daily lives, with approximately 96.2% of participants stating they had already used some AI-based technology in their learning process. Among the most commonly employed tools, educational chatbots and adaptive platforms stood out. This predilection for technologies that promote interactivity and personalization suggests a transformation in the educational paradigm, where students assume a more proactive role in the pursuit of knowledge. The demand for immediate feedback and for alternatives that optimize study time indicates that technological mediation is established as a relevant complement to traditional classroom teaching.

The frequency of use of these technologies, however, showed variations: 53.2% of respondents use them rarely, 30.4% sometimes per week, and 16.5% daily. The main contribution of AI in learning, according to the participants, lies in the ability to quickly answer questions, indicated by 62% of them, followed by improved understanding of content, mentioned by 27.8%. The general perception of AI’s impact on learning was significantly positive, with about 65.4% of students recognizing its relevance. This scenario of use and positive perception is corroborated by Sosa-Alonso et al. (2025), who highlight the essential role of Information and Communication Technologies (ICT) in education after the COVID-19 pandemic, which accelerated the digital transition at all educational levels.

Although AI is widely perceived as a factor that improves learning, with 81% of participants indicating improvement between “a lot” and “a little”, and as an effective aid for quickly resolving doubts, its use is still mostly rare for 53.2% of the sample. A significant ethical barrier identified was the reliability of information, with 69.6% of users reporting having encountered inaccurate data. Despite valuing the technology, the majority of respondents, 40.5%, consider the teacher essential and irreplaceable. There is strong interest in integrating AI into education as a complement, with personalized teaching being the most desired aspect, but skepticism persists regarding AI’s ability to understand human emotions like a real teacher.

The research also revealed that the majority of respondents, approximately 48.1%, expressed the desire for the school to use more AI in chemistry teaching, while 26.6% consider it would be very useful in content approaches. The areas where AI can intervene in improving teaching, according to students, include personalized learning, automatic grading of assignments, assistance in study organization, increased student engagement, and the development of cognitive skills. However, a worrying factor is the demand for ICT in schools without adequate planning for future impacts, which can lead to cognitive atrophy if Digital Educational Resources (DER) are used solely to automate functions. Skills such as handwriting, in-person debate, and physical experimentation, especially in areas like Physics and Chemistry, are fundamental for knowledge retention and cognitive development.

The quantitative analysis of the 40 academic performance records demonstrated a clear and significant evolution. An increase of approximately 46.6% was observed in the overall performance average, which rose from 4.93 before the AI intervention to 7.23 after its implementation. The standard deviation remained low and very similar at both times, with 1.14 before and 1.10 after the intervention, indicating that the group evolved in a relatively uniform manner, without a drastic increase in disparity among students. Notably, the minimum value recorded at the post-intervention time, which was 6.0, surpassed the average of the pre-intervention time, which was 4.93. This suggests that even participants with lower initial performance were able to overcome the previous average threshold.

The raw data of individual scores, which ranged between 3.0 and 7.0 before AI and between 6.0 and 9.0 after AI, presented an oscillatory distribution with deep valleys in the pre-intervention period, and a more stable trend with high peaks in the post-intervention period, with a visual average of approximately 5.0 and 7.5, respectively. The Shapiro-Wilk normality test for the “pre AI” and “AI Intervention” variables indicated that the assumption that the data follow a normal distribution was violated, with a p-value of 0.002, lower than the 0.05 threshold. This violation justified the application of non-parametric tests, such as the Wilcoxon Test, which is more robust for non-normal data. The Wilcoxon signed-rank test confirmed the statistical relevance of the change, with a p-value of 0.001, indicating that the observed improvement was not a casual event, but a direct consequence of the AI implementation.

Additionally, the linear regression model applied to the data sought to understand the influence of the “pre-AI” performance level on “post-AI” success. The model explained only 10.9% of the variance (R² = 0.109), suggesting that the “AI Intervention” introduced new change factors that do not depend purely on the user’s previous level. Although the “pre-AI” level proved to be a significant predictor (p = 0.037), its influence was moderate, with a coefficient of 0.312. The line equation, Post AI = 5.746 + (0.312 × Pre AI), indicates that the intercept of 5.746 represents the expected base value, while the coefficient of 0.312 shows the average increase in the intervention per additional unit in “Pre AI”. The residuals of the regression model showed normality (p = 0.109), validating the statistical soundness of the model for predictions.

The integration of Artificial Intelligence and digital platforms in the educational ecosystem reshapes teaching-learning dynamics across multiple dimensions, as pointed out by Maurer and Huwer (2026). On the pedagogical and cognitive level, adaptive tutoring systems, especially in Chemistry education, transform learning by stimulating reasoning and offering customized support for solving complex problems (Journal of Chemical Education, 2026). This approach aligns with discussions on personalized learning paths driven by AI, which enhance intrinsic motivation and strengthen student autonomy, making them active agents of their own learning pace.

However, the effectiveness of this technological personalization critically depends on the process of teacher appropriation. Findings published in Computers & Education (2025) and by Sosa-Alonso et al. (2025) emphasize that the consolidation of digital educational resources requires overcoming the barrier of fleeting fashion towards adoption based on pedagogical conviction, ensuring that technology expands rather than limits student protagonism. The triangulation of this evidence, including the study from Technology in Society (2025) which demonstrates how continuous computational support mitigates technological anxiety and generates measurable academic gains, reinforces that AI, when implemented in a personalized manner and integrated into conscious teaching practice, reduces emotional learning barriers and enhances human capital development.

In summary, the research demonstrated that the integration of Artificial Intelligence in high school, particularly in Chemistry, promoted a positive and statistically significant impact on students’ academic performance, evidenced by the increase in the overall average and the validation of the Wilcoxon Test. Although AI offers remarkable potential for personalization and learning autonomy, the neuroscientific perspective warns of the risk of cognitive atrophy if its use is restricted to automating functions. Thus, the effectiveness of AI as adaptive support is intrinsically linked to the qualified intervention of the teacher, who must mediate the process to ensure critical, consolidated, and humanized learning, aligning with the objective of understanding the real impacts of these technologies in the educational environment.

4. Conclusion

The present study aimed to analyze the impact of technological mediation by Artificial Intelligence on students’ academic performance, exploring the intersection between neuroscience and AI in the learning process, with a focus on chemistry. A positive and statistically significant impact was found, evidenced by an increase of 46.6% in the overall performance average, which rose from 4.93 to 7.23 after the intervention with AI tools, as confirmed by the Wilcoxon test (p=0.001). It was observed that most students already used AI, mainly chatbots and adaptive platforms, perceiving benefits in quickly resolving doubts and understanding content. Linear regression indicated that the intervention introduced new change factors, independent of prior knowledge. However, the neuroscientific perspective warned about the duality of Digital Educational Resources, which, while promoting autonomy, can lead to cognitive atrophy if used solely to automate functions. The main contribution lies in demonstrating that AI represents valuable adaptive support, whose effectiveness is intrinsically linked to the teacher’s qualified intervention to ensure critical, consolidated, and humanized learning.

Despite the robust findings, the study presented limitations arising from the secondary data collection methodology, which prevented direct interaction with participants to deepen perceptions about information reliability and frequency of use. The sample of 40 records, although sufficient for the statistical analysis employed, may limit the generalization of the results. For future studies, it is suggested to investigate pedagogical strategies that integrate AI in a way that preserves and stimulates higher cognitive skills, such as handwriting, in-person debate, and physical experimentation, which are essential for knowledge retention. It is crucial to develop research that addresses the ethical and pedagogical planning of AI integration in the educational environment, ensuring that technology complements teaching without replacing the irreplaceable role of the teacher in mediating learning and citizen formation.

Bibliographic References

Catania, L. J. Chapter 3 – The science and technologies of artificial intelligence (AI), Healthcare and Bioscience, 2021, Pages 29-72.

Ferreira, A; München, S; Wirzbicki, S. M. Educação do Campo e o Ensino de Ciências da Natureza e Suas Tecnologias no Ensino Médio: um panorama das pesquisas brasileiras. Brazilian Journal of Rural Education. Universidade Federal da Fronteira Sul – UFFS, 2024.

Maurer, N.; Huwer, J. Teaching and Learning about Artificial Intelligence (AI) in Secondary Education Chemistry Classes. Chemical Education Research| August 11, 2026

Moreno, E. L. e Heidelmann, S. P. Recursos Instrucionais Inovadores para o Ensino de Química. Quím. nova esc. São Paulo-SP, BR. Vol. 39, N° 1, p. 12-18, FEVEREIRO 2017

Nja; Obi, C.; Uwe; Edet, U. & Nkereuwem; Inwang, V. Artificial Intelligence Tools of Personalized Learning and Intelligent Tutoring System as Correlates of Students Motivation in Chemistry. AJSTME, Volume. 10, Issue 1. 27-32, January, 2024. ISSN: 2251-0141 (Print), 2971-6233 (Online).

Sosa-Alonso, J. J.; RIVERO, V. M. H.; Mesa, A. L. S.; Aguilar, A. B. Adoption of digital educational resources by early childhood education teachers: a fad or a conviction? Computers & Education 238 (2025) 105396.

Tian, J.; Zhang, Y. Does artificial intelligence help in improving human capital based educational development? Evidence from 29 countries. Technology in Society 83 (2025) 103004

Article originating from a Final Course Work on Neuroscience and Learning in Education

To learn more about the course, click here and access the MBX Academy platform