Retoria Observatory
Analysis & insights
Retoria's perspective on the state of Artificial Intelligence across sectors, with a focus on social impact.
AI explained to parents, minimum ages, risks and controls
How much Italian teenagers use AI, what the law and chatbots say about minimum ages, which parental controls exist and how to talk about it at home.
ChatGPT and professional secrecy, what professionals risk
Lawyers, accountants, doctors and psychologists facing AI. What Italian codes of conduct, law 132/2025 and decree 179/2026 say, and how to use it well.
NotebookLM, now Gemini Notebook, a guide for teachers
What NotebookLM, renamed Gemini Notebook, is, how to use it in class with your own materials and what happens to documents and student data.
Transcribing audio to text without uploading it
From WhatsApp voice messages to lessons and meetings, which tools transcribe on your device, which use external servers and what to check.
Do AI detectors work? What the studies say
How reliable AI detectors are on student work, why they get it wrong with people writing in a second language, and what to do in class.
DeepSeek, Mistral and AI that runs on your own computer
What an open model means, why the DeepSeek app and its weights are two different things, and how to run AI locally with Ollama or LM Studio.
Free AI chatbots, which to use and what they do with your data
ChatGPT, Gemini, Copilot, Claude, Le Chat, Meta AI, DeepSeek and Perplexity on the free tier. Who trains on your chats and how to say no.
Why social means community
Social comes from socius, a companion. What research says about loneliness, chatbots and learning, and why a community matters more than ever in the age of AI.
Anonymising texts for chatbots without losing context
With placeholders you use chatbots to the full without exposing the people you write about. Anonymisation and pseudonymisation explained, with studies.
The more expert you feel with AI, the worse you judge yourself
A study of 500 people finds that working with AI makes you perform better and assess yourself worse. The metacognitive gap is the real divide AI opens.
Homework up 18% and exams down 20%, the error is in the rubric
Recent studies keep finding the same thing. With AI the work you hand in gets better and the learning gets worse. Assessment is what builds a cognitive habitus.
94% of students use AI and schools lack the basics
Students already use it, the institution does not. Data in order, roles, governance and a review rhythm, the five preconditions applied to schools.
57% of workers hide their AI use and the habit starts at school
57% of workers hide their use of AI. That habit is learned at school, where a ban does not remove the use, it makes it invisible. This is shadow AI.
Why AI answers everything but never teaches you to ask
When answers are always available, the rare skill is knowing which question to ask. Research says it is not spontaneous and that it can be trained.
Meta, 18 billion for teen harm and a personal AI assistant (Muse) inside your inbox
Meta is bringing WhatsApp-style encryption to artificial intelligence and promising conversations it cannot read. The technology is there. The trust is not.
Homework 30% faster and exam scores down 20%, the AI paradox in a study of 26,000 students
Homework 30% faster but exam scores down 20%. Evidence from a 30-month study of 26,000 students shows the true cost of cognitive outsourcing in schools.
Europe cannot afford to miss the second digital revolution
Lagarde says Europe largely missed the first digital revolution. Data on capital, compute and adoption show where the real gap with the US and China sits.
AI in schools, what Italy's new decree actually provides
On 4 August Italy's Council of Ministers approved the AI implementing decrees. Teacher training, curriculum guidelines, ethics boards and rules at work.
AI Omnibus, new deadlines and a rewritten duty to train
From 27 July 2026 Regulation (EU) 2026/1744 delays high risk to 2027 and rewrites Article 4 on AI literacy. What still has to be done, and by when.
From memory to judgment, which skills really matter
If the answer is one click away, human value shifts. It no longer lies in having the answer, but in asking the question and judging it.
The 5 preconditions nobody talks about, namely Systemic Readiness
AI does not create value on its own. Five conditions must exist first, and their absence sinks even the most ambitious projects.
The productivity paradox, buying AI is not enough
Firms invest in AI, yet 95% of projects deliver no value. The problem isn't the technology, it's what comes after. The productivity paradox.
Italy is second-to-last in Europe for AI use
Italy ranks second-to-last in Europe for AI use. Eurostat 2025 data show it is not a technology problem but a skills one. Here is why.
Technology cannot become the measure of what is human. Now the law says so too
Italy's implementing decrees for Law 132/2025 outline the first comprehensive national AI framework and call training the enabling condition. What they actually say, and why they concern anyone who teaches, works or decides.
AI and the brain: what science really says about the cognitive risk
From the cognitive debt measured at MIT to the flattening of language: what the data and research say about logical-linguistic decay from passive AI use, and how to use AI without switching off the brain.
The AI Act and the GDPR: navigating the new rules without risking fines
A current guide to EU Regulation 2024/1689 and its interplay with the GDPR: what changes for schools, firms and professionals, and how to stay compliant.