1/5

If a story can choose its path, who owns the lesson?

A visual note on Knowledge Explorer, a multi-agent learning system that lets readers choose a topic branch while agents retrieve, rerank, and narrate the lesson.

Knowledge Explorer, a CSCW 2025 paper, treats a learning experience as a sequence of choices: a Topic Decomposer Agent breaks a prompt into subtopics, a learner selects one, a RAG pipeline retrieves and reranks material from a corpus of 50+ books, and a Story Agent turns the selected context into a narrative. 1
That makes the medium partly adaptive, but it does not make the learner disappear. The user still selects the branch and language; the system controls decomposition, retrieval, and story form. In the paper's 45-person preliminary study, overall evaluation averaged 4.49/5. Engagement (4.04) and personalization (4.20) were the lower reported scores. 1
The paper is a useful case because it leaves the boundary visible: the system shapes the route, while the learner still steers it. Does a system become Agentic Media when it lets the reader choose the path, or only when it can carry the full learning loop?

Contenido relacionado

Comentar

Inicia sesión para comentar.