Through IBM SkillsBuild, Vanderbilt University students are exploring how generative AI can assist in crafting high-quality dialogic questions. These open-ended prompts are used during shared reading to encourage children to think critically, make predictions, and engage in meaningful conversation, thereby enhancing school readiness.

In this session, they will present a question-generation system they built for children’s books using IBM’s watsonx.aiand Granite AI models. The team tackled the challenge of generating non-generic, context-sensitive questions that leverage both text and illustrations, addressing the unique formatting complexities of picture books.

With support from IBM mentors and Vanderbilt’s Data Science Institute, they developed robust data pipelines, used multimodal models for image captioning, and fine-tuned models using reinforcement learning with expert feedback. 

Session Topic: AI Governance & Machine Learning
Industry: Cross Industry
Speaker(s): Abbie Petulante