Why Education Is Important

Find a Child Therapist

CollaBot combines retrieval-based accuracy with generative AI's flexibility to support group work.

Students using CollaBot wrote significantly better essays than those using task scripts alone.

Some students found its scaffolding helpful, others found the time-tracking feature stressful.

Co-authored by Kodie Curran, Hannah Feeney, Jessica Murphy, and Michael Hogan.

Collaborative learning has been shown to promote greater academic achievement, more positive attitudes towards learning, and increased persistence in the classroom (Springer, 1999). However, orchestrating collaborative learning effectively remains a significant challenge for teachers (Dillenbourg, 2013).

It’s easier to lecture one group of silent students than to open Pandora’s box of collaborative learning. We all recognize the scenario: The teacher strives to give equal attention to each group working together in the classroom – yet one group struggles to move past the first task; another can’t agree on a shared focus; and another relies on a single student to carry the group’s workload. As the deadline approaches to present their results, the students are stressed, and the teacher is fatigued. These disruptions to collaborative learning — cognitive, socio-emotional, and participation challenges — are precisely what CollaBot, designed by Hu et al. (2025), was built to mitigate.

When conversational agents were first introduced to this process as a helping hand for teachers and students, they lacked personalization and flexibility. Being able only to retrieve pre-defined responses from a static knowledge base, they struggled to meet the varied, shifting needs of a live collaborative group. Then came generative AI, rich in flexibility, but prone to inaccurate, invented content. CollaBot operates on a retrieval-augmented generation system (RAG) that combines the benefits of these two AI system design architectures to give........

© Psychology Today