Research on the effect of visual conventions on perception and inference

Abstract

Visual conventions are perceptually efficient graphic agreements with common-sense like referents and are commonly used in what we interact with in daily life. It becomes a studying-worthy issue on whether such conventions can enhance performances and reduce cognitive load when we perceive and reason about new knowledge. Furthermore, whether the visual conventions can affect experts, who have prior knowledge and design experience about different visual encoding principles the same degree as novice who have no background knowledge in this area, is another research focus in this study. Our research is carried out according to action features when we read visualizations. Four task features are extracted, based on which behavioral and eye-tracking measurement were conducted, that is data localization, simple and complex data comparison, and knowledge inference. Both expert and novice participants were enrolled in our experiment. The result indicates that conventional elements in visualizations can hugely improve performances in more complex tasks involving higher-level cognition, like making comparisons and reasoning about new knowledge. The performance improvement can be seen from shorter response time on achieving conclusions and higher accuracy rates. Meanwhile, cognitive load, which can be measured from shorter total fixation duration and fewer fixation counts in AOIs, is reduced through applying visually conventional features. No statistically significant difference is found in comparing perceptual and inferential outputs of expert and novice group. We draw conclusions that visual conventions in visualizations can better performance in relatively complex activities, and it can be equally perceived and acquired regardless of the user’s knowledge background is. What we conclude in this study can be extended to areas of dynamic data visualization and layout design in the digital interface domain.