skip to main content


Title: A Bayesian Cognition Approach to Improve Data Visualization
People naturally bring their prior beliefs to bear on how they interpret the new information, yet few formal models exist for accounting for the influence of users' prior beliefs in interactions with data presentations like visualizations. We demonstrate a Bayesian cognitive model for understanding how people interpret visualizations in light of prior beliefs and show how this model provides a guide for improving visualization evaluation. In a first study, we show how applying a Bayesian cognition model to a simple visualization scenario indicates that people's judgments are consistent with a hypothesis that they are doing approximate Bayesian inference. In a second study, we evaluate how sensitive our observations of Bayesian behavior are to different techniques for eliciting people subjective distributions, and to different datasets. We find that people don't behave consistently with Bayesian predictions for large sample size datasets, and this difference cannot be explained by elicitation technique. In a final study, we show how normative Bayesian inference can be used as an evaluation framework for visualizations, including of uncertainty.  more » « less
Award ID(s):
1749266
NSF-PAR ID:
10096233
Author(s) / Creator(s):
Date Published:
Journal Name:
Proceedings of 2019 ACM Computer-Human Interaction (CHI)
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Narrative visualization is a popular style of data-driven storytelling. Authors use this medium to engage viewers with complex and sometimes controversial issues. A challenge for authors is to not only deliver new information, but to also overcome people’s biases and misconceptions. We study how people adjust their attitudes toward (or away from) a message experienced through a narrative visualization. In a mixed-methods analysis, we investigate whether eliciting participants’ prior beliefs, and visualizing those beliefs alongside actual data, can increase narrative persuasiveness. We find that incorporating priors does not significantly affect attitudinal change. However, participants who externalized their beliefs expressed greater surprise at the data. Their comments also indicated a greater likelihood of acquiring new information, despite the minimal change in attitude. Our results also extend prior findings, showing that visualizations are more persuasive than equivalent textual data representations for exposing contentious issues. We discuss the implications and outline future research directions. 
    more » « less
  2. Two people looking at the same dataset will create different mental models, prioritize different attributes, and connect with different visualizations. We seek to understand the space of data abstractions associated with mental models and how well people communicate their mental models when sketching. Data abstractions have a profound influence on the visualization design, yet it’s unclear how universal they may be when not initially influenced by a representation. We conducted a study about how people create their mental models from a dataset. Rather than presenting tabular data, we presented each participant with one of three datasets in paragraph form, to avoid biasing the data abstraction and mental model. We observed various mental models, data abstractions, and depictions from the same dataset, and how these concepts are influenced by communication and purpose-seeking. Our results have implications for visualization design, especially during the discovery and data collection phase. 
    more » « less
  3. People have expectations about how colors map to concepts in visualizations, and they are better at interpreting visualizations that match their expectations. Traditionally, studies on these expectations ( inferred mappings ) distinguished distinct factors relevant for visualizations of categorical vs. continuous information. Studies on categorical information focused on direct associations (e.g., mangos are associated with yellows) whereas studies on continuous information focused on relational associations (e.g., darker colors map to larger quantities; dark-is-more bias). We unite these two areas within a single framework of assignment inference. Assignment inference is the process by which people infer mappings between perceptual features and concepts represented in encoding systems. Observers infer globally optimal assignments by maximizing the “merit,” or “goodness,” of each possible assignment. Previous work on assignment inference focused on visualizations of categorical information. We extend this approach to visualizations of continuous data by (a) broadening the notion of merit to include relational associations and (b) developing a method for combining multiple (sometimes conflicting) sources of merit to predict people's inferred mappings. We developed and tested our model on data from experiments in which participants interpreted colormap data visualizations, representing fictitious data about environmental concepts (sunshine, shade, wild fire, ocean water, glacial ice). We found both direct and relational associations contribute independently to inferred mappings. These results can be used to optimize visualization design to facilitate visual communication. 
    more » « less
  4. Abstract

    In visual communication, people glean insights about patterns of data by observing visual representations of datasets. Colormap data visualizations (“colormaps”) show patterns in datasets by mapping variations in color to variations in magnitude. When people interpret colormaps, they have expectations about how colors map to magnitude, and they are better at interpreting visualizations that align with those expectations. For example, they infer that darker colors map to larger quantities (dark-is-more bias) and colors that are higher on vertically oriented legends map to larger quantities (high-is-more bias). In previous studies, the notion of quantity was straightforward because more of the concept represented (conceptual magnitude) corresponded to larger numeric values (numeric magnitude). However, conceptual and numeric magnitude can conflict, such as using rank order to quantify health—smaller numbers correspond to greater health. Under conflicts, are inferred mappings formed based on the numeric level, the conceptual level, or a combination of both? We addressed this question across five experiments, spanning data domains: alien animals, antibiotic discovery, and public health. Across experiments, the high-is-more bias operated at the conceptual level: colormaps were easier to interpret when larger conceptual magnitude was represented higher on the legend, regardless of numeric magnitude. The dark-is-more bias tended to operate at the conceptual level, but numeric magnitude could interfere, or even dominate, if conceptual magnitude was less salient. These results elucidate factors influencing meanings inferred from visual features and emphasize the need to consider data meaning, not just numbers, when designing visualizations aimed to facilitate visual communication.

     
    more » « less
  5. Educators can leverage a variety of process models to scaffold students from beginning designer practices to practices aligned with more experienced designers. The Center for Socially Engaged Design at the University of Michigan developed a Socially Engaged Design (SED) Process Model to explicitly emphasize important aspects of design that are often underemphasized or not included in commonly-used design process model visualizations, including, for example, designers embracing the limitations of their own perspective and acknowledging the power they hold, the benefits of integrating contextual considerations, and the use of prototypes throughout a design process rather than as single phase in a design process. To better understand the role of design process models, broadly, and the perceived value of process models that emphasize the importance of people and context in design work, specifically, we investigated upper-level mechanical engineering students' perceptions of this SED Process Model’s visualization. Our findings from this initial exploratory study showed both variability and several consistent themes in participants’ perceptions, for example, there were several interpretations of relationships between different aspects of the model, iteration in design was salient to all participants, and while this SED Process Model’s visualization does have recommendations, several participants noted it does not specify exactly how to achieve those recommendations. Understanding engineering students’ perceptions of this SED Process Model’s visualization can help us (1) iterate on the process model’s visualization and (2) better understand how to leverage multiple process model visualizations in engineering curricula. 
    more » « less