On the 18th, the session “Assessment Practices in the Age of AI: Examples, Tools and Safeguards” took place. The DEEC TALK featured Sérgio Pequito, Vice President for Research, Development and External Relations at DEEC. The aim of this session was to share some of the artificial intelligence tools that can be used in an educational context.

The professor began by addressing the importance of the constructivist perspective, which aligns teaching activities, learning objectives, and assessment, explaining its relationship with Bloom’s taxonomy: a pyramid framework that includes the concepts of remembering, understanding, applying, analyzing, evaluating, and creating. To illustrate this theory, he used cooking as an analogy: the use of different techniques and ingredients ultimately influences the final result. He then discussed the differences between formative and summative assessment, taking into account five quality requirements: validity, feasibility, transparency, ease of implementation, and the importance of the process.

• IAedu Platform (FCT / FCCN): a platform aimed at democratizing access to artificial intelligence in higher education in Portugal;
• Notebook LM: a tool that allows the creation of presentations, infographics, and other materials based solely on documents uploaded to the platform, without relying on external internet sources;
• Nano Banana: a software tool for image editing and creation;
• Gradescope: a platform that enables grading, organizing, and returning assessments (exams, assignments, and quizzes).

However, the use of these tools requires certain precautions, such as submitting appropriate documents and providing clear instructions, including the necessary keywords to correctly guide the AI agent. In this way, the system can function as a tutor, creating opportunities for active learning. As examples, Sérgio Pequito mentioned the possible use of these tools to generate incorrect answers in order to encourage student debate. Additionally, they can help manage expectations by highlighting common mistakes in problem-solving, support content synthesis by summarizing key concepts and formulas, and assist in exam simulation.

Furthermore, artificial intelligence can also be used for writing revision, identifying structural and spelling corrections, which can be particularly useful in producing documents such as academic thesis.

However, the use of these tools raises ethical questions, such as the authorship of results. For this reason, the professor emphasized the importance of considering the entire assessment process, including the in-person defense of results and the promotion of students’ critical thinking. Sérgio Pequito also highlighted that instructors may adopt a “traffic light” system, indicating whether AI can be fully used, partially used, or not allowed at all in their courses.

To ensure transparency, students may also be encouraged to sign a declaration specifying how different tools were used throughout their work.

The researcher also shared a 30–70 model: 30% of the process may involve the use of AI—for example, for idea generation, exploring alternatives, clarifications, and revisions—while the remaining 70%, including the origin of ideas, decision-making, critical evaluation of results, and conclusions, should always rely on the human component. However, users should also actively test their limits by assessing to what extent they can explain results without AI, how the use of these tools has influenced their reasoning, and whether they are able to defend all proposed arguments.

In this context, the professor introduced the concept of cognitive complementarity: learning can be optimized by combining human interpretation, context, and conceptual structuring with the system’s scalability and pattern recognition capabilities.

In conclusion, he also spoke about the evolving roles of teachers and students in light of new processes, curriculum redesign, ethical challenges, social responsibility, and other related issues.

The session also included the participation of several faculty members, who discussed topics such as fostering student curiosity, the structure of academic curricula given the high number of summative assessments, the limits of AI use in courses and the consequences of violating them, among other issues.