Keywords

Decision-Based Learning (DBL), AI-Assisted Instructional Design, Faculty Training and Adoption, Cognitive Load Reduction, Educational Technology Evaluation, Qualitative Product Evaluation, Six Sigma Evaluation, Instructional Workflow Optimization, AI in Higher Education, DBL Model Creation

Description

This project evaluated a AI Assistant built inside ChatGPT to support faculty in creating Decision-Based Learning (DBL) expert decision models in the Conate DBL software. The evaluation addressed a persistent training bottleneck: faculty experienced high cognitive load and significant time loss when navigating a documentation-heavy workflow that required repeated movement between training guides, ChatGPT, and the DBL platform, including extensive copy-and-paste steps. The AI Assistant was developed to replace much of the manual prompting process by providing structured protocols that guide model creation (e.g., exploring possible DBL models, creating paths, generating scenarios, and drafting learning modules) while faculty still build and finalize the model in the DBL software.

The evaluation was conducted during DBL faculty trainings in Peru (Universidad Nacional de San Agustín) and Ecuador (Universidad Nacional de Chimborazo), with the primary data informing this report drawn from the Peru implementation. Using a mixed qualitative and process-improvement approach, data were collected through trainer observations, semi-structured faculty interviews, surveys, comparative analysis of faculty-created models, and Six Sigma tools, including affinity diagrams and statistical process control charts.

Findings indicate that the AI Assistant significantly reduced perceived cognitive load, improved workflow efficiency, and enabled faculty to create more complex DBL models with a greater number of decision points in less time. Faculty reported that the Assistant simplified navigation, reduced reliance on external documentation, and supported deeper engagement with DBL pedagogy. However, the evaluation also identified adoption constraints related to AI access, variability in faculty AI literacy, and the need for clearer scaffolding and asynchronous training supports.

The results informed actionable recommendations for refining the AI Assistant, improving faculty training design, and guiding Conate’s strategic decisions regarding AI integration and resource allocation.