PariPasso with Rizzoli Education. The first course on augmented teaching has concluded
Not an AI that does students’ homework for them, but an educational ally capable of guiding, supporting and helping them understand their mistakes. The first series of sessions in the course on augmented teaching, part of the PariPasso programme by Rizzoli Education, concluded yesterday.
The third webinar, entitled Designing with AI. Strategies and tools for personalised learning, took teachers through a practical experimental workshop, demonstrating how artificial intelligence can support personalised learning without ever replacing the teacher’s role or the student’s active engagement.
Featuring Gaetano Manzulli and Daniela Pieraccini, the session brought together two complementary dimensions: on the one hand, the design of subject-specific tutors built on the teacher’s materials; on the other, a critical understanding of the mechanisms driving AI systems and the biases that can compromise their results.
AI as a bridge to inclusion
At the outset, Gaetano Manzulli highlighted a crucial premise: inclusion does not concern only certain categories of students, but the entire class. Artificial intelligence, he explained, can support truly personalised teaching because it allows activities, explanations and feedback to be tailored to learning paces, specific needs and even the development of talents. Not just SLDs or SEN, therefore, but a broader focus on the differences found in every class. At the heart of the proposal was RAG technology, which allows teachers to upload selected materials (PDFs, notes, handouts) and instruct the chatbot to respond based on those sources. This is by no means a minor step: in this way, the AI does not merely generate probable answers, but works on a knowledge base defined and verified by the teacher.
From prompt to learning tutor
The most practical part of the webinar demonstrated, step by step, how to build a learning space in School AI. From defining the prompt to sharing it with students, Manzulli illustrated how to create a tutor capable of guiding study, suggesting small steps, helping to identify an error or structure an assignment. The logic, however, remained clear throughout: do not delegate cognitive effort to the machine, but use AI to support autonomy. When, during the demonstration, the tool was asked to write a report in the student’s place, the response was consistent with the set of instructions given by the teacher: no completed assignments, but assistance with organising work, understanding and revision.
When AI gets it wrong: the problem of bias
In the second part, Daniela Pieraccini took participants behind the scenes of artificial intelligence, explaining how machine learning works in an accessible way. Through experiments carried out with Teachable Machine, she showed that a system does not learn abstract rules, but recurring patterns in the data it receives. And this is precisely where the problem arises.
If the data is scarce, lacks variety or reflects social stereotypes, the output will also be distorted. The most striking example came from image recognition: the system classified Alan Turing as a scientist, but Marie Curie as a nurse. This was not a simple random error, but the result of a training dataset incapable of adequately representing the diversity of the scientific world.
When an algorithm struggles to recognise a female scientist as such, the problem is not just with the machine: it speaks to us of the stereotypes that continue to influence data, language and social representations. This is why the webinar also links to Coding Girls & Women, the programme through which the Fondazione Mondo Digitale works to make STEM pathways more accessible, equitable and inclusive.
A second case demonstrated how, at times, bias does not directly concern gender but certain recurring elements in images, such as the bob haircut associated with the figure of the engineer. A valuable demonstration for teachers, as it highlights just how much AI systems are influenced by the choices made at the outset.
The first series of webinars ‘Augmented Teaching. From words to data, from data to knowledge’
- From text to data. Exploring texts with artificial intelligence
- with Francesca Sabatini and Felicia Bitetti
- The Inquiry method. Designing a research pathway with AI
- with Annamaria Bove
- Designing with AI. Strategies and tools for personalised learning
- with Gaetano Manzulli and Daniela Pieraccini
