Technology
July 10, 2026
AI in education: the challenge of responsibility
Smart technologies require governance, teacher training, and critical thinking to avoid deepening inequalities
Artificial intelligence has been consolidating itself as one of the most relevant forces in contemporary digital transformation, and education has come to occupy a central place in this movement. If before the discussion revolved around the digitalization of content and the expansion of remote learning, now the debate advances to systems capable of generating feedback, adapting learning paths, supporting pedagogical planning, and automating administrative tasks (Bond et al., 2024).
In practical terms, this means that AI is no longer just a peripheral resource: it is beginning to interfere with the very organization of the educational process, with teacher mediation, and with the way knowledge is produced, evaluated, and distributed. International organizations such as Unesco (United Nations Educational, Scientific and Cultural Organization) have been warning that this advancement needs to be guided by a vision centered on the human being, and not just by promises of efficiency (Miao; Holmes, 2023).
Part of the enthusiasm around AI in education stems from an age-old problem: the difficulty of offering more responsive teaching in systems marked by profile heterogeneity, budget limitations, and scarcity of pedagogical time. In this context, artificial intelligence tools offer something historically desired by schools and universities: the possibility of personalizing learning at scale.
Recent reviews show that the AIED (Artificial Intelligence in Education – something like “Artificial Intelligence in Education”) area has grown rapidly and now encompasses everything from intelligent tutoring systems to generative assistants, educational analytics, automated assessment, and teaching support. The literature also indicates that this ecosystem should no longer be seen as a passing trend, but as part of a broader reconfiguration of educational practices (Wang et al., 2024).
Personalization
One of the most promising uses of AI is precisely in personalization. Intelligent tutoring systems can identify error patterns, adjust the difficulty level, recommend exercises, and offer immediate feedback according to the student’s performance.
This is an important change, because traditional teaching often operates at a single pace for very diverse classes. Classic meta-analyses on intelligent tutoring systems already indicated moderate positive effects on learning in higher education, showing that these systems tend to outperform methods such as traditional teaching, reading materials, autonomous activities, and computer-assisted instruction, although they do not yet fully replace human tutoring.
These findings reinforce that AI can improve educational outcomes when designed with a consistent pedagogical foundation and integrated into teaching work, and not just as an isolated technological solution. More recently, systematic reviews continue to point to gains in engagement and performance, although they highlight the need for more robust experiments and greater attention to the implementation context (Steenbergen-Hu; Cooper, 2014; Kulik; Fletcher, 2016).
In the context of teaching work, AI can also represent a relevant change. According to Unesco, artificial intelligence can assist teachers in both pedagogical activities and learning management, allowing for content personalization, performance pattern analysis, automation of repetitive tasks, and support for educational decision-making processes.
For the entity, these technologies can enhance learning experiences, contribute to lesson preparation, support assessment processes, and favor individualized student monitoring, in addition to strengthening the continuous professional development of teachers. However, the document emphasizes that AI should not replace the teacher, but act as a tool to support pedagogical practice, preserving human interaction, critical thinking, teacher mediation, and ethical responsibility as central elements of the educational process (Unesco, 2024a).
Evaluation
Another important field is assessment. With the rise of generative AI, educational assessment has ceased to be merely a space for measurement and has also become a terrain for epistemological dispute. Recent literature indicates that generative AI tends to profoundly transform higher education assessment processes, while simultaneously creating opportunities and challenges.
Among the opportunities, the possibility of offering immediate feedback, supporting student self-assessment, favoring learning personalization, and stimulating more autonomous study practices stand out. However, the use of tools like ChatGPT also raises concerns related to academic integrity, as these systems can produce texts, answers, and activities with a high degree of apparent quality. In view of this, Xia et al. (2024) argue that assessments need to be redesigned, shifting the focus from simple content reproduction to more complex competencies, such as critical thinking, creativity, problem-solving, self-regulated learning, and ethical use of AI.
The authors also highlight the need for teacher training in assessment, digital literacy, and artificial intelligence, as well as the review of institutional assessment policies, with greater emphasis on innovative, interdisciplinary practices aligned with new educational demands.
There is also a frequently underestimated dimension in the debate on AI and education: inclusion. In theory, artificial intelligence-based technologies can enhance accessibility, support students with disabilities, offer automatic translation, assisted reading, textual simplification, clear language feedback, and more flexible and personalized learning paths. This potential becomes even more relevant in contexts of mass education, where individualized support tends to be limited.
However, the GEM Report, from Unesco, warns that technology is not an automatic synonym for equity. The expansion of digital solutions can deepen existing inequalities when problems of connectivity, insufficient infrastructure, absence of institutional policies, and low levels of digital literacy persist.
The report highlights, for example, that a large part of schools in the world still do not have adequate internet access and that many countries lack clear regulations for the use of educational technologies. Thus, the inclusive impacts of AI depend less on the technology itself and more on the social, organizational, and pedagogical conditions that structure its implementation in the educational environment (Unesco, 2023).
Challenges
The challenges, therefore, are as relevant as the potentialities. The first of these is the risk of superficialization of learning. If misused, AI can encourage quick answers without deep understanding, weaken intellectual authorship, and reduce the cognitive effort needed to consolidate knowledge. Another critical point is the problem of hallucinations, biases, and factual errors.
Generative models can present false information with high verbal fluency, which is particularly concerning in environments where students are still forming conceptual repertoires and validation criteria. Recent articles on the use of large language models in education insist that the pedagogical value of these tools depends on the simultaneous development of skills in verification, curation, prompt formulation, and critical reading (Kasneci et al., 2023).
The issue of governance also deserves attention. Education deals with sensitive data, individual trajectories, historical inequalities, and asymmetric power relations. In this scenario, adopting AI without clear parameters regarding data protection, traceability, decision-making criteria, and institutional responsibility would be a strategic and ethical error.
Unesco has been arguing that educational systems need frameworks that articulate rights, safety, human oversight, and AI competency development for both teachers and students. It is not just about regulating a tool, but about defining which educational project it is intended to strengthen with it (Miao; Holmes, 2023; Unesco, 2024a; Unesco, 2024b).
AI’s great contribution to education may not be in automating teaching, but in pressuring schools, universities, and managers to rethink what truly matters to learn. In a context where ready-made answers become abundant, capabilities such as argumentation, judgment, creativity, contextual interpretation, ethics, and complex problem-solving gain even more value.
Technology can be a powerful ally to scale, personalize, and support teaching, but it will hardly replace the formative role of the pedagogical relationship. The central point, therefore, is not to choose between accepting or rejecting AI. It is to define, with clarity, which uses strengthen learning and which impoverish it. In education, as in few other fields, innovation without criteria can be mere automation of what already worked poorly. With criteria, it can become a concrete opportunity to qualify teaching, management, and human development.
| To access the references of this text click here |
Who wrote this column
Renato Máximo Sátiro
