Introduction:

As it is shown in several articles, scientific visualization of the data is among the best ways to expand the ideas and prepare the framework for future analysis and discussions. The main point is that understanding the perception of the visualized data can improve the quality and quantity of the analysis. Visual features assigned to data attributes must have some properties. For example, it should take the advantage of the strength of the visualized system. It should also be properly be aligned to the analysis needed for the viewers. One other important aspect is that it should avoid the visual inferences that can mask the information. This idea is shown better in figure 1.

Figure 1- The flowchart of a perception in visualization

Figure 1- The flowchart of a perception in visualization

As you see in this figure, trying to find the insight directly from data is usually a tie consuming and not efficient process while having a properly managed visualized data can help us to find the best perception, which will lead to a cognition for the end user. To explain the idea more, four papers which include the applications of this property are discussed in this article.

Part 1- Theory of graphical perception by Clevelan and McGill

In this research, the authors are evaluating the perception in data visualization from two points of view: first identifying a set of elementary perceptual tasks and second how accurately people can perform them. They first started with some basic drawings for example a circle that was about to reflect the area and length or some colors that might reflect some information. Then, they expand the idea to some distributions such as scatter plot, which shows the rate of murder in an area or pie charts and divided bar charts, which show the position-angle experiments or statistical maps with shadings that turn out some information about different state of US.
By sharing these graphs to participants, they hoped to get their insights about 10 elementary tasks. Some of them are length align or nonaligned a common scale, length, direction, angle, volume, area, etc. The distribution of log error of all the experiments shows that the result depends on the type of the visualization. For example, position aligned a common scale returns the least error (most of the participants could identify it correctly) while the shading or color saturation had the highest amount of error in experiments. Altogether, a theory was found out through experiments (position judgements were more accurate than length judgement). This will also result to another theory which is replacing some of the visualization methods such as dot charts and framed-rectangle charts with those which were more visible for investigators such as bar charts or pie charts.

Part 2- Crowdsourcing graphical perception by Heer and Bostock

The main goal of this paper (this section) is to prove what we saw in previous section and also evaluate some more facts about visualization perception. In this paper, the authors tried to show the effectiveness of perception through crowdsourcing. In olden days, this was a time-consuming process to hire people to do the researches and evaluate their work. Nowadays there are services that can easily handle this type of researches like what the authors did and recruited some mechanical Turk from amazon. They paid them $$$0.01 to $$$0.1 per task and used their insights to investigate the crowdsourced experiments on different graphical experiments. They divided their job into three general categories.

  1. The effectiveness of their technique compared to old experiments (Cleveland & McGill work)

  2. Demonstrating the use of crowdsourcing to generate new perception results

  3. Analyzing the cost of hiring mechanical Turks from Amazon

They first replicated the Cleveland experiments whereas the mechanical Turks were about to make some judgements based on the graphs. These graphs were three positions encoding along a common scale, two length encoding, one angle (such as pie charts) and one circular area (bubble chart). The result of the experiments shows that the more complex the graphs are, the higher the error is. It turns out in most of the cases when the participants are trying to analyze a bar chart whereas they need to find the lengths, it is easier for them compared to when they need to deal with bubble charts or angles.

On second set of experiments, the authors wanted to examine the separation and layering of the graphs through luminance contrast. The participants were asked to parametrize the display of chart gridlines over a plot area. They were about to change the lightness of the graph as much as it is viewable for them. Using the output results, the authors found some level of significant effect for the plots. This can be helpful for the default value for future graphs of the software. Finally, they showed that the price of hiring these participants considering the amount of the jobs that they can handle, and their accuracy is valuable for this research.

Part 3- More papers

Dastani (2002) analyzed the data and the perceptual information of the visualized data and to classify the data from perceptual point of view knowing the information of visualizations. Their model for the effective visualization is depicted in figure 2.

Figure 2- Process model of effective data visualization

Figure 2- Process model of effective data visualization

This is one step ahead of what we evaluated in previous papers. After knowing that a more graphical visualization (such as bar chart compared to scatter bar chart) works better in terms of human intelligibility, we now want to see how much details should be considered in graphs. The author used the sale information of some automobile manufacturers in different countries as a base line. The comparison between the graphs show that whenever more information is revealed on the graphs, it is more readable and perceptually understandable for the end users. A comparison is shown in figure 3. After several experiments, it is shown that the left-hand side graph which has more information, is a better choice compared to right hand side graph.

Figure 3- Visualization of Auto sale in different countries

Figure 3- Visualization of Auto sale in different countries

Lam et. al (2012) continued the work of Heer to find some empirical studies in information visualization. In this research, they used the information visualization, which contains not only the visualization itself but also the complex processes that a tool is meant to support. The scope of their evaluation contains pre-design, design; prototype, deployment and redesign of the model which intends to answer a research question. Finally, the important and interesting result in their work is that how to evaluate the human judges. To achieve this goal, they evaluated some of the paper goals and tried to see if they could achieve the goal through visualizations. Finally, they showed that more visualization in research papers are achieved in current years, which result in more understandable research papers.

Conclusion:

Several researches show the importance of perception in data visualization. While some of the earlier researches show the effectiveness of visualization of bar charts compared to scatter bar charts, the latter researches proved the same thing through more advanced evaluating tools. Rather than that, it shown that the more readable and described the graph is, the more perceptually understandable for the end-user it will be. This shows the importance of visualized data (type of visualization) for companies, which deal with end user.

References:

Cleveland, W. S., & McGill, R. (1984). Graphical perception: Theory, experimentation, and application to the development of graphical methods. Journal of the American statistical association, 79(387), 531-554.

Heer, J., & Bostock, M. (2010, April). Crowdsourcing graphical perception: using mechanical turk to assess visualization design. In Proceedings of the SIGCHI conference on human factors in computing systems (pp. 203-212). ACM.

Dastani, M. (2002). The role of visual perception in data visualization. Journal of Visual Languages & Computing, 13(6), 601-622.

Lam, H., Bertini, E., Isenberg, P., Plaisant, C., & Carpendale, S. (2012). Empirical studies in information visualization: Seven scenarios. IEEE transactions on visualization and computer graphics, 18(9), 1520-1536.