Precision farming and AI in schools with the PariPasso project
A photograph of a leaf, a fruit or a patch of soil can serve as a starting point for learning to observe, classify and organise data and, above all, to verify the results provided by artificial intelligence. This is the approach proposed by Michele Fiore in the webinar From images to data. AI for precision agriculture, the latest event in the PariPasso series – Rizzoli Education’s project dedicated to educational inclusion, organised by the Fondazione Mondo Digitale.
A computer science teacher, engineer and digital facilitator, Fiore has used precision agriculture as a practical example to demonstrate a method that can also be applied to other subjects. The starting point is simple: with AI, it is not enough simply to ask a question and receive an answer. The educational process follows three steps: observe, analyse, decide. And the final step remains the responsibility of the individual. ‘Ultimately, it is always the human who must have the final say on what the artificial intelligence has produced,’ Fiore pointed out during the session.
Learning by training a machine
One of the most interesting parts of the webinar was devoted to Teachable Machine, the Google tool that allows users to create simple machine learning models without any knowledge of programming languages.
Students can, for example, create categories such as ‘healthy leaf’, ‘damaged leaf’ and ‘yellowed leaf’, upload photographs, train the model, test it with new images and analyse any errors. If the result isn’t right, they go back, check the categories and refine the training.
This is precisely where the workshop becomes particularly interesting from an educational perspective: the pupils discover that an AI system depends on the examples it receives and the quality of the data it is trained on. And they become, in Fiore’s words, “programmers without the knowledge of a programmer”.
Apple, banana and tomato: what does AI really see?
The workshop then got into full swing with three simple photographs: an apple, a banana and a tomato. The same images were fed into various AI systems with a very specific instruction: to describe only what is actually visible — colour, shape, surface, distinctive features — and to explicitly state what cannot be determined. A seemingly simple distinction that opens up an important exercise in critical thinking. From a photograph, we can observe the colour or a surface imperfection; we cannot determine the internal quality of a fruit, the presence of chemical residues or its exact nutrient content. Furthermore, by comparing the responses produced by different tools, the students discover that the outputs can vary. At that point, the most important task is not to obtain the machine’s answer, but to explain the differences, verify and justify their own conclusions. ‘Review is the fundamental part of the entire teaching process,’ emphasises Fiore.
Change the photograph, change the response
Even the original image must be critically examined. Light, distance, angle, shadows and background can alter what the system recognises and increase the likelihood of error. One activity that is easily replicated in the classroom involves, for example, photographing the same object over several days or under different conditions and comparing the AI’s responses. The result can take the form of a photographic diary, a table, a graph or the starting point for a class discussion. It is a lesson that extends far beyond agriculture: checking the input helps us understand the output, and a convincingly formulated answer is not necessarily correct.
A workshop even for those who do not teach agriculture
Precision agriculture provides the context for the webinar, but the proposed method is intentionally interdisciplinary. Italian can focus on description and argumentation; maths on the organisation and representation of data; science on observation; technology on how tools work; and citizenship on sustainability and the responsible use of AI. ‘Artificial intelligence within a structured UDA can be used for Italian, for history, for a whole host of subjects: it is cross-curricular,’ explained Fiore. The webinar also presents a 15-hour UDA for an agricultural college, which brings together crop production, IT, maths, agricultural chemistry and Italian, culminating in the creation of a dossier and an authentic task. However, Fiore also proposes a more streamlined version, divided into three lessons, which can be adapted for lower secondary schools. The programme can therefore be useful not only to teachers at agricultural and technical colleges, but also to teachers of science, maths, technology, computing, Italian and civic education, to support teachers and to anyone interested in designing interdisciplinary activities on AI.
Everyone taking part in the same experience
The visual approach also offers opportunities for inclusion. Starting with concrete and easily observable elements allows the complexity of the activities to be adjusted and the same experience to be offered through different methods. The aim, as Fiore explained, is “to enable everyone to take part in the same experience in ways suited to their own needs”. The message running through the entire webinar is simple: AI can recognise features, organise information and flag anomalies, but the crucial skill remains that of critically interpreting data, verifying it and contextualising it. For those who were unable to watch the live stream, the recording allows you to catch up on the entire workshop, the prompts used, the examples with Teachable Machine and the UDA proposal.
And even if you don’t teach agriculture, it’s worth watching: photography is merely the starting point; the real workshop is about how we learn to observe, question and verify artificial intelligence.
The next session
REGISTER
