Towards Perceptual Optimization of the Visual Design of Scatterplots

Abstract

Abstract-Designing a good scatterplot can be difficult for non-experts in visualization, because they need to decide on many parameters, such as marker size and opacity, aspect ratio, color, and rendering order. This paper contributes to research exploring the use of perceptual models and quality metrics to set such parameters automatically for enhanced visual quality of a scatterplot. A key consideration in this paper is the construction of a cost function to capture several relevant aspects of the human visual system, examining a scatterplot design for some data analysis task. We show how the cost function can be used in an optimizer to search for the optimal visual design for a user's dataset and task objectives (e.g., "reliable linear correlation estimation is more important than class separation"). The approach is extensible to different analysis tasks. To test its performance in a realistic setting, we pre-calibrated it for correlation estimation, class separation, and outlier detection. The optimizer was able to produce designs that achieved a level of speed and success comparable to that of those using human-designed presets (e.g., in R or MATLAB). Case studies demonstrate that the approach can adapt a design to the data, to reveal patterns without user intervention.

Information Visualization Evaluation Using Crowdsourcing

“…The number of papers reporting crowdsourcing experiments for InfoVis changed from one in 2009 to 13 in 2017 ( Figure 1). While papers in the early rise of crowdsourcing had to dedicate an entire section to justify the use of crowdsourcing for their empirical studies (e.g., [MDF12]), recent papers simply cite previous InfoVis papers to justify the use of crowdsourcing for evaluating data visualizations (e.g., [MPOW17]). In this section, we report on the findings from our analysis of the 190 experiments with respect to the aspects and dimensions discussed in Section 3.…”

“…For instance, the text could ask the user to click somewhere else (e.g., on a specific region of the visualization) and not the provided response buttons (e.g., [RSC15]); participants that click on the buttons are considered inattentive and their data is excluded from analysis. One paper presented its experiments as micro-games [MPOW17]: participants were initially informed that a reward will be granted only if a minimum of 30% of the questions were answered correctly, and then shown the score halfway through the experiment.…”

“…One paper investigated how users learn about data visualization on online framework [KL16], while another paper assessed how annotations on visualizations can help crowdworkers produce better textual explanation of the visualized information [WHA12]. Crowdsourcing was also used to first calibrate the weights of a cost function that captured the effectiveness of a scatterplot design for different data analysis tasks, and then empirically compare the optimized designs to a number of commonly used default designs [MPOW17]. ..............................................................................................................................…”

“…Decrease in participants' engagement also reduces the likelihood that the worker completes the experiment, thus slowing down data collection and study completion. If experimenters cannot reduce the experiment length and still wish to use crowdsourcing, then either the payment is increased [HB10], or incentives to maintain participants' engagement and interests are devised (e.g., deploying the experiment as a game [MPOW17]), or data for repetitions of combination of study factors is collected from a larger and more diverse pool of participants. Studies that require participants with specific requisites (e.g., full color vision) could, in some cases, take advantage of the large volume of crowdworkers and increase the size of their participants pool, such that the uncertainty introduced by participants not meeting the requisites is reduced.…”

Exploring the Effects of Aggregation Choices on Untrained Visualization Users' Generalizations From Data

“…Some researchers have suggested ensemble perception as an alternative to aggregation, where a person perceives distributional properties from groups of visual items spread over space or time [CG15, CH17, HRA15, KNKH18, SHGF16]. Accumulating evidence suggests that the visual system is capable of quickly, automatically and accurately extracting mental representations of ensembles [AO07, Alv11, HW09, HZ84, HWB15, LKW16, MPOW17]. Visualization researchers have examined how well people can infer statistics from encodings that vary in their level of aggregation for particular types of data, like time series [ARH12, ACG14], multi‐class scatterplots [GCNF13] or hierarchical data sets [EF10].…”