The integration of artificial intelligence in education promises to support teachers, but its effectiveness faces hurdles in the lack of infrastructure and in assessment biases.
The adoption of artificial intelligence (AI) in schools and universities raises expectations about personalized learning. AI can act in report generation and scenario simulation, alleviating the daily tasks of teaching. This use requires continuous monitoring to avoid inaccuracies that demotivate students, as pointed out by the analysis from Stanford experts.
The systemic integration of technology demands public policies built for it. These strategies must be aligned with the specific context of the country, focusing especially on massive training.
(Note: The next references in this text come from preprints, preliminary research publications that have not yet undergone peer review.)
For the adoption to work, it needs to start from clear pedagogical principles. Technological tools must reinforce human reasoning instead of replacing it, ensuring that the educational mission remains central.
Despite the huge potential recognized in content creation and lesson planning, implementation faces structural technical unpreparedness. A survey conducted with basic education teachers in Brazil shows that these barriers manifest themselves, for example, in the scarcity of information technology infrastructure in schools.
In the context of measuring and grading essays, the approach used impacts the result. The application of agentic workflows surpasses the consistency of language models operating independently for this task.
However, automating assessments can raise ethical and social justice questions. Applying technology in educational measurement hinders the transparency of decisions and can end up reinforcing the inequalities already present in assessments.
The perception of students about these tools also affects teaching effectiveness. In tests with college students, when undergraduate students do not know who graded the activity, they tend to evaluate the machine-generated feedback better. However, upon revealing the technological authorship, students demonstrate prejudice and trust plummets immediately.
This student trust is complex and goes beyond the tool’s accuracy metric. It is heavily mediated by social interactions and the prior affective reactions the student had with technology while studying.
Technology in education must focus on expanding human capabilities. Public managers and schools need to align enthusiasm for innovation with the guarantee of adequate infrastructure and transparent assessment practices.