EBSCOlearning has used AWS GenAIIC to tackle the challenge of scaling assessment creation efficiently with GenAI technology.

EBSCOlearning needed to enhance the creation of multiple-choice questions (MCQs) for its extensive learning content. The traditional approach, requiring manual input from subject matter experts (SMEs), ensured quality but was costly, slow, and hard to scale. With a growing library of educational materials, EBSCOlearning required a faster, scalable solution to maintain its high standards.

Using Anthropic’s Claude 3.5 Sonnet via Amazon Bedrock, the GenAIIC team developed an automated pipeline for generating MCQs. The system creates up to seven questions per content piece, with detailed explanations for each answer choice. Sophisticated prompt engineering ensures the questions align with strict guidelines, including clarity, inclusivity, and educational relevance. The questions are also evaluated for adherence to quality metrics through a three-tiered review: AI-based guideline evaluation, rule-based checks, and holistic review of question sets.

The key benefit lies in scalability and time efficiency. This GenAI-driven solution produces questions at a fraction of the time and cost of manual efforts, while maintaining—and sometimes surpassing—the quality of human-generated questions. Intelligent revision capabilities, driven by AI, ensure continuous improvement. Any flagged questions undergo multiple iterations until they meet set standards, with unresolved cases referred for human review.

Since implementation, EBSCOlearning reports significant improvements in the speed and consistency of question and assessment creation. The solution is scalable and adaptable for evolving educational content. Future plans include integrating more complex content, personalized assessments, and diverse question types. EBSCOlearning’s GenAI approach significantly reduced the time and cost associated with manual question generation, offering consistent quality across various subjects. This innovation has allowed them to scale assessments efficiently, enhancing the learning experiences with diverse and engaging content. Feedback suggests GenAI questions are often on par with, or even superior to, human-generated ones.