top of page

What we learned (so far) in the AI and media literacy intervention for socioeconomically vulnerable Portuguese vocational students




Between March and June 2026, we ran the “first phase” of the Portuguese PRODIGI AI literacy pedagogical intervention, with vocational students who receive school social support.

The previous post showed how we designed the training program.

Now this post shares how it worked in practice in five professional schools in the Lisbon Metropolitan Area: what students told us, what worked, what we had to change along the way - and what we suggest in the future.


How it worked

The five modules of 90 minutes each were held in person at schools over five to six weeks.

Each session had one or two researchers, plus a facilitator: a master’s student in Communication Sciences who has worked with young people at risk of social exclusion. He was close in age to the participants and also helped design and run the sessions.

We mixed short explanations with activities. We used Mentimeter polls, let students experiment live on their own phones, held guided discussions, and ran small-group tasks.



First-module session

In every school, students from different courses sat together, many for the first time. That mix brought different experiences and ways of speaking into the room.


Who took part

Aiming for 150 youth participants in total, in these five schools we asked each school for a group of about 25 students who received social support and were interested in the topic. However, in the end, only 59 students completed the programme, meaning they filled in the questionnaire before and after and attended at least three sessions: 32 girls and 27 boys, most aged 15 to 17, from 14 different courses.


What students told us

At the end of the training, students filled in a paper evaluation sheet on what had been completely new, what they wanted to know more about, and anything else they wanted to say.

What  was completely new… After the environmental cost of AI, the next surprise was the people behind AI. One student said it was new to hear about the engineers, technicians, and annotators who work on it, because AI seemed like a kind of magic. Another was surprised by “the quantity and variety of professions derived from AI”. Others wrote that ChatGPT “is not 100% true”, that platforms store our data to base their answers on it, and that “there are people who pour their hearts out to ChatGPT”.

They wanted to know more about… When we asked what they wanted to know more about, many asked about money and work. How can I use AI to earn money? How do you use it to program? How might it change jobs in my field, such as health? Others asked different questions. One wondered how AI makes people dependent and “lazy”. Another asked: “Did the person who created it think we are very stupid and lazy?”.

Anything else? Most students (spontaneously) expressed high satisfaction with the training. One wrote that they got “a lot of information and space to share their opinion”. Another said the training was important to understand something “which we think we know, but in reality we don’t know that well”, and that it was taught in a “fun and creative” way. A few found the sessions tiring. Older participants often wrote very short comments, such as “more AI”.


What got them talking

In Module 2, students argued about Kate Crawford’s sentence that AI is neither intelligent nor artificial. One said AI is intelligent “because it answers correctly to whatever you ask it”. Some read “artificial” literally and thought of artificial grass. In one school, a student looked up the definition on her phone and read it to the class. In another, a student asked ChatGPT “Do you know if you exist?” and read the answer aloud.

In Module 4, we showed a news story from the newspaper Público about a woman who married an AI chatbot. Students checked on their phones whether it was true. One pointed out that the bot “must have been agreeing with her all the time”. Another said: “For her to marry, he also had to accept!”



We also asked students to type either “I am happy” or “I am sad” into a chatbot. To “I am happy”, the chatbot gave a short answer, and one student sighed that it was “because he doesn’t care about me”. To “I am sad”, students received long replies with advice and questions. Students noticed that the chatbot seems to want to keep the conversation going, “to get more of our data” and to push people towards a paid plan.

Generated ChatGPT images for “a Portuguese family” showed white and middle-class people, two parents, two children, all smiling and mainly outside home. In one school, students rejected it straight away. They said “nobody actually lives with grandparents”, that “there are Black Portuguese people” and “lower social classes”, and that “not everyone likes to smile like that”; some tried “Brazilian families” on their phones to compare. In another school, the chatbot drew a working-class couple with two children, indoors, with a plaque on the wall bearing the Estado Novo dictatorship’s slogan, “Deus, Pátria, Família” (God, Homeland, Family). At first, students took the flag, the slogan, the clay jugs and the clothes as “proof” that the family was Portuguese. Then the facilitator asked whether they had any of these at home. They did not (“I have God, but not the rest,” said one), and when asked what a stereotype is, a student answered that it is when you see someone dressed a certain way and immediately associate them with a specific group. The group then asked for a “Cape Verdean family”, and one student predicted a family “with too many kids and no flip-flops”. It did not happen: the chatbot gave a polished indoor portrait, much like the “Portuguese family” ones.

In a school, the question “Have you ever encountered fake information?” turned into a lively discussion about television news, social media, and news media trust.

 

To each group, its own dynamics

Every group was different. In one school, Acting and Legal Services students engaged more critically, while a group of boys in sports courses were more detached. In another, students from a Manicure course had little confidence with digital tools, and several refused to do the digital production. Some topics also worked better with some students than others. Football examples of misinformation worked well with boys, while girls more often mentioned celebrities, influencers, and musicians. Politics rarely sparked interest.

Many students had a limited vocabulary and some words that we considered accessible were new to them. Some were embarrassed to read aloud, although in several groups they gradually gained the confidence to do it.


What we changed along the way

●      We had planned to compare Character.AI and ChatGPT, based on a comment from a Portuguese teenager. Most students had never heard of Character.AI, so we removed it to avoid making anyone feel behind.

●      A group task on the components of AI ended with students copying text without understanding it. We replaced it with images and visual aids.

●      In the “Google vs. ChatGPT” battle, students used Google’s AI instead of regular search. We then asked them to use standard Google search.

●      A few examples didn’t work well. For instance, many students did not recognixe and name the recently elected President of the Republic.

●      We reinforced short written and visual instructions, and explained beforehand what we sensed as difficult words.

●      We dropped the deepfake exercise we had previously  planned, because of technical problems and because some students suggested making images to mock classmates and teachers. The media literacy game Bad News was cut for lack of time.


What did not go as planned

In each school, a “contact point”, such as a course coordinator or a psychologist, had to present the project to students, collect consent (including from parents of minors) and protect the schedule and classroom. In two schools, this worked very well, but in the others, some students arrived without knowing why they were there and with no memory of signing anything. Several left after the first session and never came back, even though others were enthusiastic about the activities.

Calendar pressures added to this. The sessions fell during the Easter break, national holidays and, in most schools, the last weeks of the school year. Internships and off-site training took students away, in one school to the extent that only four students completed the final questionnaire. Some parents did not give consent because of final exams or worries that their children would fall behind.

There were also practical problems, such as thick walls blocking Wi-Fi, a school server down for part of a session, and students without mobile data, battery, or phone storage. Some students could not do tasks such as making a presentation or sending an email from their phones, and in some groups only a few students did most of the work.


What we suggest for others

●      Keep the core of the programme, but update the examples often.

●      Build trust. Plain language, small groups and a facilitator close in age made a difference.

●      Plan an alternative for every activity, and check students’ digital skills before assuming everyone has them.

●      Give each group member a clear task, so teamwork is not left to the most confident student.

●      Involve the school from the start. Teachers and coordinators need to plan for and provide close monitoring of the intervention preparation process, while also ensuring sustained support for participant motivation.


What is happening now

To reach the target of 150 students, we launched a call for professional courses in public schools that integrate the Network of Schools of Excellence (ESCXEL), in the Lisbon Metropolitan Area. We trained more young people to work as facilitators, and we are concluding the “second phase” of fieldwork in October 2026.


More on the results soon!





Opmerkingen


01_znak_siatka_podstawowy_kolor_biale_tlo (002).png
eu_co_funded_en.jpg

PRODIGI is co-funded by the European Union, under Grant Agreement no. 101182849. Views and opinions expressed are however, those of the author(s) only and do not necessarily reflect those of the European Union. The European Union cannot be held responsible for them.

White and Black Minimalist Virtual Assistant LinkedIn Article Cover Image.png
bottom of page