Visualization designers increasingly use diverse types of visualizations, but assistive technologies and education for blind and low vision people often focus on elementary chart types. We explore textual explanation as a more generalizable solution. We define three dimensions of explanation strategies based on education theories: comparing to a familiar chart type, describing how to draw one, and using a concrete example. We develop a prototype system that automatically generates text explanations from a given chart specification. We conduct an exploratory study with 24 legally blind people to observe both the effectiveness and the perceived effectiveness of the strategies. The findings include: description of visual appearance is overall more effective than instructions for drawing, effective strategies differ by each chart type and by each participant, and the user's perceived effectiveness does not always lead to better performance. We demonstrate the feasibility of an explanation generation system and compile design considerations.
https://doi.org/10.1145/3544548.3581139
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2023.acm.org/)