Christiane Caneva

@ccaneva.bsky.social

substack.com/profile/4654... An organisation may approve an AI tool based on a good result and overlook the problem. The time spent reading increases, drop-out rates fall, and satisfaction rises. Comprehension remains at the same level. Engagement measures one thing. Learning measures another.

Christiane Caneva, PhD (@christianecaneva)

An organisation may approve an AI tool based on a good result and overlook the problem. The time spent reading increases, drop-out rates fall, and satisfaction rises. Comprehension remains at the ...

substack.com

The better the AI notes, the lower the test scores. In a 2025 study, automated note-taking produced the worst comprehension outcomes — and participants still preferred it. The tool that generates notes for you may also take the understanding with them. canevac.substack.com/p/the-better...

The better the AI notes, the less you understand. A study on the assistance dilemma.

A 2025 study on AI-assisted note-taking and learning outcomes: what the research shows about automated tools, cognitive engagement, and comprehension in higher education — and what it can't yet tell.

canevac.substack.com

On peut enfin détecter si un texte a été produit par une IA. Pourtant, ça ne répond pas à la question qu'on croit. Un filigrane ne dit pas qui a écrit quoi — seulement qu'une IA a participé. Ce que ça change pour la rentrée académique/scolaire 2026 : christianecaneva.substack.com/p/on-peut-en...

On peut enfin savoir qui a utilisé l’IA! Ou presque.

Filigrane des textes générés par IA, obligatoire depuis le 2 août 2026 : pourquoi l’AI Act peut faire accuser sans jamais prouver qui a écrit un texte.

christianecaneva.substack.com