Social media research faces a fundamental tension between internal and ecological validity. Vignette experiments often isolate single posts, which fail to capture how users browse feeds where content competes for attention. Field studies offer realism but are subject to algorithmic interference, which threatens internal validity. Against this backdrop, we introduce Digital In-Context Experiments (DICE), an experimental paradigm that bridges this gap and offers experimental control in scrollable feeds that mimic digital environments. Researchers can manipulate entire feed compositions (rather than individual posts in isolation). DICE provides researchers with post-level dwell times as behavioral proxies for (in)attention and combines these unobtrusive measures with traditional survey responses. In this workwhop, we will discuss the type of research questions DICE is more (less) suited for and give guidance on how to use the tool.