Authoring and Publishing Interactive Articles

Thesis (Ph.D.)--University of Washington, 2021Interactive articles—a dynamic medium that combines narrative text with interactive graphics and computational simulations—are important for education, journalism, and scientific publishing because they can improve learning outcomes by promoting active engagement in readers. However, interactive articles are difficult and costly to produce: they are created using complex general-purpose programming tools and authors receive little guidance on effective design, due in part to a lack of empirical evidence of usage patterns in real-world situations. This research supports the design, creation, publication, and evaluation of interactive articles by presenting analysis of interactive article design, three authoring tools that lower the threshold to create interactive articles, and systems that facilitate the collection and analysis of article usage data. This dissertation starts with an analysis of sixty interactive articles, in which we connect interactive article design patterns to research in multimedia learning, digital journalism, and human-computer interaction. We identify a set of five key affordances of the medium for educational and journalistic usage and discuss implications for authoring tools. Building on this foundation, we present Idyll, an expressive domain-specific markup language that reduces the amount of code and effort needed to create and publish interactive articles through a reactive document model and standard component library. Since our initial deployment, Idyll has been used by educators, journalists, and researchers to, for example, create interactive materials for an undergraduate physics curriculum, communicate the results of detailed urban development models, and publish interactive data-driven analyses in college newspapers. However, the system is too technical and low-level for some users. We extend Idyll with two higher-level authoring systems: Idyll Studio is a WYSIWYG-style structured editor that lowers the threshold for authors by replacing general-purpose programming tools with an easy-to-learn interface centered on direct and instrumental manipulation; Fidyll is a simplified markup language that can be used to rapidly generate interactive articles as well as static PDFs, slideshows, and videos from a single source file, improving the portability, archivability, and accessibility of interactive articles while limiting the amount of markup needed overall. Leveraging Idyll’s reactive document model, we develop a system that automates the detailed instrumentation of interactive articles and publish an open dataset of more than 50,000 session logs from in-the-wild usage of three interactive articles. To facilitate learning from such data, we develop two corresponding visual analysis tools that aid in inspecting usage data at both micro- and macro-levels. Taken together, this suite of tools serves as a set of modular building blocks for authoring and publishing interactive articles that communicate information effectively. The software in this dissertation has been used by educators, journalists, and researchers to create articles that have resonated with hundreds of thousands of readers, that were covered in digital design publications, that are among the top online search results for their respective mathematical and statistical topics, that have gone viral online, and that received honor at the IEEE VIS Workshop on Visualization for AI Explainability. The code is open-source and can be found at https://github.com/idyll-lang. Examples can be found at https://idyll-lang.org/gallery