ExPeerience: Towards AI-Assisted Learnersourcing to Bridge Conceptual Understanding and Problem Solving in Database Programming Education

Overview

This research addresses a fundamental tension in learnersourcing—an educational paradigm in which students generate instructional resources—between prioritizing learning outcomes through cognitive scaffolding and producing reusable, standardized content. The study proposes an AI-assisted learnersourcing model that distributes task responsibilities between human learners and large language models, enabling students to engage in cognitively meaningful work while LLMs manage mechanical and structural tasks. The implementation, ExPeerience, focuses on database programming education and tasks students with collaboratively creating worked-out examples for SQL problem-solving.

Methods and approach

ExPeerience was designed using user-centered design principles to structure AI collaboration across three distinct roles: ideation partner, co-creator of artifacts, and evaluator of student inputs. The system scaffolds students through a workflow for generating contextualized worked-out examples in database programming. Evaluation involved 24 participants who used ExPeerience to co-create SQL problems and solutions. Performance and engagement were compared against a baseline condition using the Gemini chatbot without the integrated collaborative workflow. Assessment metrics included content quality, diversity of problem contexts, depth of student engagement with AI-generated components, and the extent to which participants attempted independent problem-solving.

Key Findings

ExPeerience participants produced SQL problems situated in more diverse and personally meaningful contexts compared to baseline users. Participants actively evaluated, edited, and refined AI-generated components rather than passively accepting outputs. Most ExPeerience users authored their own SQL solutions, whereas baseline participants largely accepted AI outputs without modification and did not attempt independent problem-solving. The worked-out examples generated through ExPeerience demonstrated greater contextualization, variability, and thoughtful construction overall. These outcomes indicate that role-structured AI collaboration improved both learning engagement and content quality, successfully addressing the original tension between learning and sourcing objectives.

Implications

The AI-assisted learnersourcing paradigm demonstrates feasibility as a mechanism for balancing competing educational objectives in student-generated content systems. By assigning distinct collaborative roles to AI agents, educators can maintain cognitive scaffolding necessary for learning while simultaneously ensuring that produced artifacts meet standardization and reusability requirements. This approach suggests that carefully designed role distribution—rather than full automation or full manual creation—optimizes both pedagogical value and resource scalability. The findings support further exploration of role-structured AI collaboration in learnersourcing contexts beyond programming education, with attention to how role definitions and boundaries affect learning engagement and content outcomes.