From Research Outputs to Public-Facing Data Stories: Designing a GenAI Assistant for Evidence-Grounded Scrollytelling

Researchers increasingly face expectations to communicate scientific findings beyond academic venues, yet transforming research outputs into accessible, accurate, and engaging public-facing formats remains difficult. Generative AI (GenAI) may reduce the authoring and implementation effort, but it creates risks of narrative drift, unsupported simplification, and loss of scientific integrity. We address this design problem in the context of data storytelling, a communication approach that combines narrative structure, visual explanation, and guided interpretation. Employing design science research, we design a GenAI-based assistant for evidence-grounded research scrollytelling, a web-based form of data storytelling organized through scrolling-based progression. Grounded in boundary-spanning theory and informed by a researcher focus group (n=8), we instantiate and demonstrate the artifact on six research outputs and evaluate it through formative think-aloud sessions (n=5). We contribute nascent design knowledge for structuring GenAI-supported transformation of research outputs into public-facing data stories while retaining researcher control over evidence and meaning.