VisDoc uses GenAI to reduce cognitive overload in onboarding docs
New research shows AI restructuring cuts cognitive load for newcomers by significant margins.
Researchers from Oregon State University, Northern Arizona University, and Rochester Institute of Technology have introduced VisDoc, a GenAI-powered prototype that restructures open source software (OSS) onboarding documentation to reduce cognitive overload. Drawing on Cognitive Theory of Multimedia Learning (CTML), the pipeline segments dense, fragmented onboarding materials into task-based units, infers workflows, eliminates redundant content, and generates multimodal explanations (e.g., text + diagrams). An expert evaluation (N=4) confirmed VisDoc's completeness, accuracy, and potential for adoption.
A between-subjects experiment with 14 newcomers compared VisDoc's restructured docs against original OSS onboarding materials. Participants using VisDoc completed tasks with higher success rates, reported significantly lower cognitive load (measured via NASA-TLX), and rated usability higher on the System Usability Scale. The study provides empirical evidence that cognitively informed restructuring—not just summarization—can improve task performance and reduce frustration, offering a scalable approach for OSS communities to retain contributors.
- VisDoc uses a GenAI pipeline to operationalize CTML strategies for restructuring onboarding documentation.
- Segments docs into task-based units, infers workflows, removes redundancy, and generates multimodal explanations.
- In a 14-person study, VisDoc users achieved higher task success and significantly lower cognitive load vs. original docs.
Why It Matters
AI-driven documentation restructuring could reduce newcomer abandonment in open source by lowering cognitive barriers.