Our ability to learn the causal structure of the world is a fundamental aspect of human intelligence. This causal knowledge allows us to predict future events and perform actions to generate desired outcomes [1-2]. However, acquiring such knowledge directly from the world is costly in both time and effort [3-4]. Thus, it is vital to transmit already acquired causal knowledge effectively, in order to facilitate quick decision-making and advance collective understanding [5-6].
The present study explores visualizations as a means overcoming these limitations by conveying causal knowledge in a condensed, exportable format over large gaps in space and time. We capitalize on prior work that has examined how the spatial layout of elements in well-designed visualizations mirrors the placement and relations of elements in the physical world, making it easier to grasp the correspondence among elements in each domain [7-10]. Visualizations have also been shown to promote inference of abstract relations by leveraging a small set of relational symbols, such as lines and arrows [11-12].
In particular, participants will draw simple machines (gears, levers, and pulleys), which comprise a unique class of objects that can be rich in both visual appearance and causal information. The study thus explores how prompts to generate visual explanations (e.g., of how a machine functions) or visual depictions (e.g., of what a machine looks like) may affect what participants choose to include in their drawings. Specifically, the study probes drawing behavior by examining how factors such as ink usage, number of strokes, and drawing time differ between conditions.