Four Perspectives on Human Bias in Visual Analytics

Wall, Emily; Blaha, Leslie M.; Paul, Celeste Lyn; Cook, Kristin A.; Endert, Alex

doi:10.1007/978-3-319-95831-6_3

Cited by 39 publications

(30 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For this reason, bias has increasingly become a topic of study across a variety of disciplines, including machine learning and visual analytics (e.g. [8,9,55]). Much of the focus in the visualization community has focused on characterizing and addressing cognitive biases [5,53,54].…”

Section: Biasmentioning

confidence: 99%

Selection Bias Tracking and Detailed Subset Comparison for High-Dimensional Data

Borland

Wang

Zhang

et al. 2019

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

Fig. 1. Exploratory cohort selection in high-dimensional datasets can lead to selection bias-unintended side-effects in variable distributions-that may go unnoticed by the user. Our selection bias tracking system and detailed cohort comparison visualizations, deployed in a medical temporal event sequence visual analytics tool, include (a) a cohort provenance tree to keep track of created cohorts and indicate when selection bias may have occurred, (b-d) a suite of high-dimensional cohort comparison visualizations that employ hierarchical aggregation to display the differences between two cohorts in detail, and (e) data-type dependent comparisons of individual variable distributions.Abstract-The collection of large, complex datasets has become common across a wide variety of domains. Visual analytics tools increasingly play a key role in exploring and answering complex questions about these large datasets. However, many visualizations are not designed to concurrently visualize the large number of dimensions present in complex datasets (e.g. tens of thousands of distinct codes in an electronic health record system). This fact, combined with the ability of many visual analytics systems to enable rapid, ad-hoc specification of groups, or cohorts, of individuals based on a small subset of visualized dimensions, leads to the possibility of introducing selection bias-when the user creates a cohort based on a specified set of dimensions, differences across many other unseen dimensions may also be introduced. These unintended side effects may result in the cohort no longer being representative of the larger population intended to be studied, which can negatively affect the validity of subsequent analyses. We present techniques for selection bias tracking and visualization that can be incorporated into high-dimensional exploratory visual analytics systems, with a focus on medical data with existing data hierarchies. These techniques include: (1) tree-based cohort provenance and visualization, including a user-specified baseline cohort that all other cohorts are compared against, and visual encoding of cohort "drift", which indicates where selection bias may have occurred, and (2) a set of visualizations, including a novel icicle-plot based visualization, to compare in detail the per-dimension differences between the baseline and a user-specified focus cohort. These techniques are integrated into a medical temporal event sequence visual analytics tool. We present example use cases and report findings from domain expert user interviews.

show abstract

Section: Biasmentioning

confidence: 99%

Selection Bias Tracking and Detailed Subset Comparison for High-Dimensional Data

Borland

Wang

Zhang

et al. 2019

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

show abstract

“…• Abstract, unstructured, and complex business data hard to visualize  Visualization of large and abstract business data is difficult (Dilla et al, 2010) • Lack of visual metaphors fit for business applications  Difficulty selecting effective visual metaphor  Difficulty in visualizing business data (Al-Kassab et al, 2014;Edmunds and Morris, 2000)  Behavioral habits and cognitive biases lower effectiveness of visualization (Sacha et al, 2016) • Difficulty to represent some business phenomena with the available visual methods  Using visualization requires special skills and training (Berinato, 2016;Wall et al, 2018)  Difficulties collaborating via visual forms (Martinez-Maldonado et al, 2019) Advances in interfaces pertinent to data science: • Rich variety of tools for data visualization and vocalization  Data visualization amplifies cognition (Fekete et al, 2008;Keenan and Jankowski, 2019)  Visual exploration simplifies analysis of very large data sets (Keim, 2001(Keim, , 2002  Visual metaphors make data easier to understand (Cybulski et al, 2015)  Physical representations of data are now possible, for example, 3D printing (D'Aveni, 2015)…”

Section: Limitationsmentioning

confidence: 99%

Data science in organizations: Conceptualizing its breakthroughs and blind spots

Cybulski

Scheepers

2021

Journal of Information Technology

View full text Add to dashboard Cite

The field of data science emerged in recent years, building on advances in computational statistics, machine learning, artificial intelligence, and big data. Modern organizations are immersed in data and are turning toward data science to address a variety of business problems. While numerous complex problems in science have become solvable through data science, not all scientific solutions are equally applicable to business. Many data-intensive business problems are situated in complex socio-political and behavioral contexts that still elude commonly used scientific methods. To what extent can such problems be addressed through data science? Does data science have any inherent blind spots in this regard? What types of business problems are likely to be addressed by data science in the near future, which will not, and why? We develop a conceptual framework to inform the application of data science in business. The framework draws on an extensive review of data science literature across four domains: data, method, interfaces, and cognition. We draw on Ashby’s Law of Requisite Variety as theoretical principle. We conclude that data-scientific advances across the four domains, in aggregate, could constitute requisite variety for particular types of business problems. This explains why such problems can be fully or only partially addressed, solved, or automated through data science. We distinguish between situations that can be improved due to cross-domain compensatory effects, and problems where data science, at best, only contributes merely to better understanding of complex phenomena.

show abstract

“…There is universal agreement among scholars and practitioners that visuals should be designed ethically with the intent to inform, not mislead, the audience [12,13]. Literature on the deceptive potential of data visualizations to mislead has been around for some time [13][14][15][16][17], as have ideas for how to detect such deception [6]. Scholars have also discussed best practices for how developers of data visualizations can create not only functional graphs but also pay attention to ethical representations of information to avoid misleading the reader [8,11,[18][19][20].…”

Section: Literature Reviewmentioning

confidence: 99%

“…The deceptive potential of data visualizations has been discussed by practitioners and theorists [2][3][4][5], and they have made recommendations for how developers can avoid inadvertently deceiving those to whom they are communicating data [2,6,7]. Such recommendations include starting y-axes at zero, framing data broadly enough to accurately represent trends, avoiding 3-D effects that can misrepresent the size or angle of data points, and avoiding randomly spaced data points; though there is still some debate over whether these recommendations should be adhered to at all times, or whether there are instances in which they can be discarded depending on the focus of the data being communicated [8][9][10][11].…”

Section: Introductionmentioning

confidence: 99%

The Deceptive Potential of Common Design Tactics Used in Data Visualizations

O'Brien

2020

Proceedings of the 38th ACM International Conference on Design of Communication

View full text Add to dashboard Cite

Visualizations effectively communicate data about important political, social, environmental, and health topics to a wide range of audiences; however, longstanding trust of graphs as conveyors of factual data makes them an easy means for spreading misinformation. Scholars in technical and professional communication have not yet conducted needed empirical research into people's perception and comprehension of data visualizations, especially when part of larger information texts [1]. Our study investigated the extent to which people exaggerated the differences between data points when reading graphs about non-controversial topics that used deceptive techniques and/or exaggerated titles. Participants (n=329) were randomly assigned to view one of four treatments for four different graph types (bar, line, pie, and bubble) and then asked to answer a question about each graph. Results show that deceptive techniques used in the graphs (including truncated axes, 3-D exaggeration, and arbitrary sizing), caused participants to misinterpret information in the deceptive vs. control visualizations for all of the graphs regardless of graph type, previous visualization coursework or comfort level with reading graphs. Results also showed that the presence of exaggerated vs. control titles that accompanied each graph did not significantly influence the extent of the misinterpretation. We will discuss the implication of these findings for technical communication as well as avenues of future research already underway that investigate these topics further.

show abstract

Four Perspectives on Human Bias in Visual Analytics

Cited by 39 publications

References 46 publications

Selection Bias Tracking and Detailed Subset Comparison for High-Dimensional Data

Selection Bias Tracking and Detailed Subset Comparison for High-Dimensional Data

Data science in organizations: Conceptualizing its breakthroughs and blind spots

The Deceptive Potential of Common Design Tactics Used in Data Visualizations

Contact Info

Product

Resources

About