Interactive Mascot: New grammar lets devs directly manipulate visualization scene components

Traditional interaction abstractions for data visualizations rely on event streams, signals, and data selections, creating a mismatch with scene-centric systems that expose semantic components like marks, encodings, layouts, and axes as first-class objects. Zhicheng Liu's new paper introduces Interactive Mascot, a grammar designed to bridge this gap. It extends static scene-centric representations by framing interactive behavior as information flow among four core components: trigger, responder, evaluator, and updater, along with two forms of context (event context and state context). This approach allows developers to specify interactions more naturally using the same semantic building blocks already present in the visualization scene.

Interactive Mascot has been implemented as a JavaScript library (http://this http URL) and evaluated for expressiveness, performance, and usability. It covers the full interaction design space of Vega-Lite while adding capabilities for stateful interactions, direct manipulation of scene components, and freeform selection. Runtime performance is comparable to Vega-Lite, and a qualitative user study indicates the grammar is both learnable and usable for authoring interactive visualizations. The work opens new possibilities for building rich, manipulable charts without losing the semantic clarity of scene-centric design.