How to Design Graphs That Will Be Ignored?
Statistical charts and information graphics are used to summarise surplus data and information concisely. In recent years, the use of these graphs has increased significantly to the point that these information graphics are surplus, and they surround us. We often see some graphics and don’t even realize that we are doing it or don’t have time to see it for more than a few milliseconds, like, comprehending details from smartwatch graphs with a glance of around 300 milliseconds. If you feel there is a strong probability that your graphs might not engulf the viewer’s entire attention, then you might want to read ahead.
Formation of Gist or Ensemble of a Graph
When a graph or an image is in the viewer’s foveal vision for a concise span of time (subconsciously or consciously), these graphs form a gist or an ensemble of that image in our brains. This gist or ensemble might be perceived as the picture on the left (above if you are reading this article on the phone).
By looking at this picture, we can presume some details about the scene. Similarly, when a viewer sees a graph for a short time (as less as 16 milliseconds), our brains comprehend some details from it, which plays a vital role in forming your opinion, just like this scene in the movie ‘Focus’.
Comprehending a graph in 16 ms
One might still wonder if humans can comprehend anything from the graphs in such a small duration! To your surprise, some studies have confirmed that people could create an ensemble, comprehend and summarize natural images when shown for as less as 20 milliseconds.
Now, you might think that natural scenic images are very different from the monotonous graphs. Yes, they are very different, so I thought to perform research to identify if it is actually possible for us to comprehend the details from a graph in a single glance of 16 milliseconds or more. I created an experiment where for each participant, different types of graphs with different characteristics (line graphs, pie charts, bar charts, node-link diagrams, and scatter plots) were presented for the display time of 16, 32, or 96 milliseconds at random (similar to the video). Two generic questions (referred to as question 1 and question 2 in later parts) about each graph assessed the graph comprehension in the research experiment.
Research Experiment findings
Upon collecting the data and analyzing from my research, I discovered that the average error in participants' results was significantly low (below graphs). These errors were compared with the average probability of a participant choosing the incorrect answer randomly, 80%. Upon using inferential statistics and the Kruskal-Wallis test, I could mathematically confirm that we humans are capable enough to comprehend the details when a graph is displayed for 96, 64, 32, or 16 milliseconds.
Now that we are aware of the fact that viewers can fetch the details from these graphs. The question lies in how we can design the graph to increase the comprehensibility of such a graph. To explain the process of graph comprehension in normal circumstances, we can rely on the justification of Graph comprehension models like Pinker's model or the Knowledge-based model. However, I could also verify from my experiments that these models do not justify the graph’s comprehension in a glance of few milliseconds. Characteristics that affect the comprehension of graphs in normal scenarios do not follow the same behavior.
Display characteristics affecting graph comprehension
My research also identified that display characteristics, like color, type of graphs, and data complexity, do not affect graph comprehension.
Effect of Color: Unlike the conventional models of graph comprehension, This study identified that a colored graph does not improve the graphs' comprehensibility when shown for a quick glance of a few milliseconds. Viewer’s accuracy does not get affected irrespective of the fact that the graph is colored or monochrome.
Data Complexity: My study used 4 levels of complexity in the form of a number of elements in a graph. It confirmed that the complexity of a graph's underlying data does not really impact graph comprehension. Participant’s accuracy was a little low when the graph was elementary or had extremely high complexity.
Type of Graphs: This study has shown that all five types of graphs, i.e., line graphs, pie charts, bar charts, node-link diagrams, and scatter plots, are equally comprehensible. There was a certain trace of evidence in the study group indicating higher accuracy in identifying the elements in bar charts and slightly lower accuracy in identifying the pie chart sections. This behavior could be attributed to the shape of the graphs.
Graph Literacy: There is a huge generic class of comprehension models based on Pinker’s model of graph comprehension. Pinker’s model emphasizes that graph schema or graph literacy improves graph comprehension. However, in this group of studies, graph schema does not influence the comprehension of graphs. This result possibly indicated the influence of the bottom-up process in comprehending the graphs when displayed for a few milliseconds.
In Conclusion
From the experiment's overall results, it is likely that existing graph comprehension models do not justify the comprehension of the gist of a graph. Hence before designing the graphs, you need to be careful of the aforementioned facts if you feel the graph might not get the proper attention of the viewers until the time there is a new graph comprehension model that can explain the graph comprehension process and substantiate the relationship between the extraneous characteristics and comprehension of the gist of the graph.
