Who is actually thinking here? New practical article about the HEAR framework, available for free download in Dutch and English
Rules state what is permitted, and training teaches how things work. But between these two lies a space that rarely has language, and it is precisely there that learning happens. In a new four-page practical article, Pascal Mariany, founder of CivAI and senior lecturer at Utrecht University of Applied Sciences, elaborates the HEAR framework for everyone working with generative AI in education. The article is available for free download, in Dutch and in English.
Download the article in Dutch (pdf, 4 pages)
Download the English version, Who is really thinking here? (pdf, 4 pages)
Where the Friction Lies
The friction is not in the technology. It arises when a student receives an answer before their own question has been fully developed, and no longer knows whether they have been helped or overtaken. Research from the past two years shows how real this shift is. Those who outsource their thinking to AI score lower on critical thinking and perform worse once the AI is removed. It is not the tool that determines the learning effect, but the design of the interaction.
Four Layers, One Conversation
HEAR stands for Human Expressive AI Reasoning, the framework from the book The Art of Living with AI. It describes four layers that together determine whether an AI interaction is educational, responsible, and formative.
Human Intent. The intention comes first. What do I want to learn, achieve, or decide here?
Expressive Interaction. AI formulates fluently and assertively, but has no intention. Do I follow the content, or the form?
AI Mediation. Between question and answer, the system selects, organises, and narrows down lines of thought. What choices were already made before I started?
Reasoned Responsibility. Ownership of the judgement, grounded in traceable considerations. Can I explain why this is correct and what is my own?
Two layers lie with the human, two in the interaction with the system. The human is at the beginning and at the end. The risk of AI in education is not that students know less, but that they think less.
What You Will Find in the Article
The article is set up as a practical piece, with two infographics, quotes from the book, and sixteen scientific sources, from Zimmerman and Weizenbaum to recent studies on cognitive outsourcing, trust in AI, and learning with AI tutors.
Figure 1 shows the HEAR model at a glance, figure 2 the shift from task-oriented to learning-oriented AI use for students, teachers, teams, and management.
The four layers of Human Intent, from goal intention to responsibility intention, help to make the question behind the prompt explicit.
Six concrete starting points for teachers and teams, from the intention question before the first prompt to the team agreement on where AI is deliberately kept absent.
HEAR in practice shows how the AI companions in EduGPT are designed based on the framework, such as the Feedback Companion, which does not take over the teacher’s judgement, and the Curriculum Companion, which does not assess the curriculum but puts design questions on the table.
About the Book
The article is based on chapter 6 of The Art of Living with AI (Van Duuren Media, 2026). The English edition, The Art of Living with AI, will be published in October 2026. Those who read about HEAR earlier on civai.eu may know the framework from the article on curriculum thinking from August.
Want to respond or continue the conversation? Email pascal@civai.eu, or share the article within your team. Sharing with source attribution is appreciated.
Image for illustration, created with AI