Imagining Replications: Graphical Prediction & Discrete Visualizations Improve Recall & Estimation of Effect Uncertainty

A systematic review of measurement uncertainty visualizations in the context of standardized assessments

This systematic review summarized findings of 29 studies evaluating visual presentation formats appropriate for communicating measurement uncertainty associated with standardized clinical assessment instruments. Studies were identified through systematic searches of multiple databases (Medline, Embase, PsycInfo, ERIC, Scopus, and Web of Science). Strikingly, we found no studies which were conducted using samples of clinicians and included clinical decision-making scenarios. Included studies did however find that providing participants with information about measurement uncertainty may increase awareness of uncertainty and promote more optimal decision making. Formats which visualize the shape of the underlying probability distribution were found to promote more accurate probability estimation and appropriate interpretations of the underlying probability distribution shape. However, participants in the included studies did not seem to benefit from the additional information provided by such plots during decision-making tasks. Further explorations into how presentations of measurement uncertainty impact clinical decision making are needed to examine whether findings of the included studies generalize to clinician populations. This review provides an important overview of pitfalls associated with formats commonly used to communicate measurement uncertainty in clinical assessment instruments, and a potential starting point for further explorations into promising alternatives. Finally, our review offers specific recommendations on how remaining research questions might be addressed.

Multiple Forecast Visualizations (MFVs): Trade-offs in Trust and Performance in Multiple COVID-19 Forecast Visualizations

The prevalence of inadequate SARS-COV-2 (COVID-19) responses may indicate a lack of trust in forecasts and risk communication. However, no work has empirically tested how multiple forecast visualization choices impact trust and task-based performance. The three studies presented in this paper ( N=1299) examine how visualization choices impact trust in COVID-19 mortality forecasts and how they influence performance in a trend prediction task. These studies focus on line charts populated with real-time COVID-19 data that varied the number and color encoding of the forecasts and the presence of best/worst-case forecasts. The studies reveal that trust in COVID-19 forecast visualizations initially increases with the number of forecasts and then plateaus after 6-9 forecasts. However, participants were most trusting of visualizations that showed less visual information, including a 95% confidence interval, single forecast, and grayscale encoded forecasts. Participants maintained high trust in intervals labeled with 50% and 25% and did not proportionally scale their trust to the indicated interval size. Despite the high trust, the 95% CI condition was the most likely to evoke predictions that did not correspond with the actual COVID-19 trend. Qualitative analysis of participants' strategies confirmed that many participants trusted both the simplistic visualizations and those with numerous forecasts. This work provides practical guides for how COVID-19 forecast visualizations influence trust, including recommendations for identifying the range where forecasts balance trade-offs between trust and task-based performance.

Testing Bayesian models of belief updating in the context of depressive symptomatology

OBJECTIVES

Predictive processing approaches to belief updating in depression propose that depression is related to more negative and more precise priors. Also, belief updating is assumed be negatively biased in comparison to normative Bayesian updating. There is a lack of efficient methods to mathematically model belief updating in depression.

METHODS

We validated a novel performance belief updating paradigm in a nonclinical sample (N = 133). Participants repeatedly participated in a non-self-related emotion recognition task and received false feedback. Effects of the feedback manipulation and differences in depressive symptoms on belief updating were analysed in Bayesian multilevel analyses.

RESULTS

Beliefs were successfully manipulated through the feedback provided. Depressive symptoms were associated with more negative updating than normative Bayesian updating but results were influenced by few cases. No evidence of biased change in beliefs or overly precise priors was found. Depressive symptoms were associated with more negative updating of generalised performance beliefs.

CONCLUSIONS

There was cautious support for negatively biased belief updating associated with depressive symptoms, especially for generalised beliefs. The content of the task may not be self-relevant enough to cause strong biases. Further explication of Bayesian models of depression and replication in clinical samples is needed.

Visualizing Uncertainty in Probabilistic Graphs with Network Hypothetical Outcome Plots (NetHOPs)

Probabilistic graphs are challenging to visualize using the traditional node-link diagram. Encoding edge probability using visual variables like width or fuzziness makes it difficult for users of static network visualizations to estimate network statistics like densities, isolates, path lengths, or clustering under uncertainty. We introduce Network Hypothetical Outcome Plots (NetHOPs), a visualization technique that animates a sequence of network realizations sampled from a network distribution defined by probabilistic edges. NetHOPs employ an aggregation and anchoring algorithm used in dynamic and longitudinal graph drawing to parameterize layout stability for uncertainty estimation. We present a community matching algorithm to enable visualizing the uncertainty of cluster membership and community occurrence. We describe the results of a study in which 51 network experts used NetHOPs to complete a set of common visual analysis tasks and reported how they perceived network structures and properties subject to uncertainty. Participants' estimates fell, on average, within 11% of the ground truth statistics, suggesting NetHOPs can be a reasonable approach for enabling network analysts to reason about multiple properties under uncertainty. Participants appeared to articulate the distribution of network statistics slightly more accurately when they could manipulate the layout anchoring and the animation speed. Based on these findings, we synthesize design recommendations for developing and using animated visualizations for probabilistic networks.

Examining Effort in 1D Uncertainty Communication Using Individual Differences in Working Memory and NASA-TLX

As uncertainty visualizations for general audiences become increasingly common, designers must understand the full impact of uncertainty communication techniques on viewers' decision processes. Prior work demonstrates mixed performance outcomes with respect to how individuals make decisions using various visual and textual depictions of uncertainty. Part of the inconsistency across findings may be due to an over-reliance on task accuracy, which cannot, on its own, provide a comprehensive understanding of how uncertainty visualization techniques support reasoning processes. In this work, we advance the debate surrounding the efficacy of modern 1D uncertainty visualizations by conducting converging quantitative and qualitative analyses of both the effort and strategies used by individuals when provided with quantile dotplots, density plots, interval plots, mean plots, and textual descriptions of uncertainty. We utilize two approaches for examining effort across uncertainty communication techniques: a measure of individual differences in working-memory capacity known as an operation span (OSPAN) task and self-reports of perceived workload via the NASA-TLX. The results reveal that both visualization methods and working-memory capacity impact participants' decisions. Specifically, quantile dotplots and density plots (i.e., distributional annotations) result in more accurate judgments than interval plots, textual descriptions of uncertainty, and mean plots (i.e., summary annotations). Additionally, participants' open-ended responses suggest that individuals viewing distributional annotations are more likely to employ a strategy that explicitly incorporates uncertainty into their judgments than those viewing summary annotations. When comparing quantile dotplots to density plots, this work finds that both methods are equally effective for low-working-memory individuals. However, for individuals with high-working-memory capacity, quantile dotplots evoke more accurate responses with less perceived effort. Given these results, we advocate for the inclusion of converging behavioral and subjective workload metrics in addition to accuracy performance to further disambiguate meaningful differences among visualization techniques.

The Science of Visual Data Communication: What Works

Effectively designed data visualizations allow viewers to use their powerful visual systems to understand patterns in data across science, education, health, and public policy. But ineffectively designed visualizations can cause confusion, misunderstanding, or even distrust-especially among viewers with low graphical literacy. We review research-backed guidelines for creating effective and intuitive visualizations oriented toward communicating data to students, coworkers, and the general public. We describe how the visual system can quickly extract broad statistics from a display, whereas poorly designed displays can lead to misperceptions and illusions. Extracting global statistics is fast, but comparing between subsets of values is slow. Effective graphics avoid taxing working memory, guide attention, and respect familiar conventions. Data visualizations can play a critical role in teaching and communication, provided that designers tailor those visualizations to their audience.

A Survey of Domain Knowledge Elicitation in Applied Machine Learning

Eliciting knowledge from domain experts can play an important role throughout the machine learning process, from correctly specifying the task to evaluating model results. However, knowledge elicitation is also fraught with challenges. In this work, we consider why and how machine learning researchers elicit knowledge from experts in the model development process. We develop a taxonomy to characterize elicitation approaches according to the elicitation goal, elicitation target, elicitation process, and use of elicited knowledge. We analyze the elicitation trends observed in 28 papers with this taxonomy and identify opportunities for adding rigor to these elicitation approaches. We suggest future directions for research in elicitation for machine learning by highlighting avenues for further exploration and drawing on what we can learn from elicitation research in other fields.

Two graphs walk into a bar: Readout-based measurement reveals the Bar-Tip Limit error, a common, categorical misinterpretation of mean bar graphs

How do viewers interpret graphs that abstract away from individual-level data to present only summaries of data such as means, intervals, distribution shapes, or effect sizes? Here, focusing on the mean bar graph as a prototypical example of such an abstracted presentation, we contribute three advances to the study of graph interpretation. First, we distill principles for Measurement of Abstract Graph Interpretation (MAGI principles) to guide the collection of valid interpretation data from viewers who may vary in expertise. Second, using these principles, we create the Draw Datapoints on Graphs (DDoG) measure, which collects drawn readouts (concrete, detailed, visuospatial records of thought) as a revealing window into each person's interpretation of a given graph. Third, using this new measure, we discover a common, categorical error in the interpretation of mean bar graphs: the Bar-Tip Limit (BTL) error. The BTL error is an apparent conflation of mean bar graphs with count bar graphs. It occurs when the raw data are assumed to be limited by the bar-tip, as in a count bar graph, rather than distributed across the bar-tip, as in a mean bar graph. In a large, demographically diverse sample, we observe the BTL error in about one in five persons; across educational levels, ages, and genders; and despite thoughtful responding and relevant foundational knowledge. The BTL error provides a case-in-point that simplification via abstraction in graph design can risk severe, high-prevalence misinterpretation. The ease with which our readout-based DDoG measure reveals the nature and likely cognitive mechanisms of the BTL error speaks to the value of both its readout-based approach and the MAGI principles that guided its creation. We conclude that mean bar graphs may be misinterpreted by a large portion of the population, and that enhanced measurement tools and strategies, like those introduced here, can fuel progress in the scientific study of graph interpretation.

Revealing Perceptual Proxies with Adversarial Examples

Data visualizations convert numbers into visual marks so that our visual system can extract data from an image instead of raw numbers. Clearly, the visual system does not compute these values as a computer would, as an arithmetic mean or a correlation. Instead, it extracts these patterns using perceptual proxies; heuristic shortcuts of the visual marks, such as a center of mass or a shape envelope. Understanding which proxies people use would lead to more effective visualizations. We present the results of a series of crowdsourced experiments that measure how powerfully a set of candidate proxies can explain human performance when comparing the mean and range of pairs of data series presented as bar charts. We generated datasets where the correct answer-the series with the larger arithmetic mean or range-was pitted against an "adversarial" series that should be seen as larger if the viewer uses a particular candidate proxy. We used both Bayesian logistic regression models and a robust Bayesian mixed-effects linear model to measure how strongly each adversarial proxy could drive viewers to answer incorrectly and whether different individuals may use different proxies. Finally, we attempt to construct adversarial datasets from scratch, using an iterative crowdsourcing procedure to perform black-box optimization.

A Bayesian cognition approach for belief updating of correlation judgement through uncertainty visualizations

Understanding correlation judgement is important to designing effective visualizations of bivariate data. Prior work on correlation perception has not considered how factors including prior beliefs and uncertainty representation impact such judgements. The present work focuses on the impact of uncertainty communication when judging bivariate visualizations. Specifically, we model how users update their beliefs about variable relationships after seeing a scatterplot with and without uncertainty representation. To model and evaluate the belief updating, we present three studies. Study 1 focuses on a proposed "Line + Cone" visual elicitation method for capturing users' beliefs in an accurate and intuitive fashion. The findings reveal that our proposed method of belief solicitation reduces complexity and accurately captures the users' uncertainty about a range of bivariate relationships. Study 2 leverages the "Line + Cone" elicitation method to measure belief updating on the relationship between different sets of variables when seeing correlation visualization with and without uncertainty representation. We compare changes in users beliefs to the predictions of Bayesian cognitive models which provide normative benchmarks for how users should update their prior beliefs about a relationship in light of observed data. The findings from Study 2 revealed that one of the visualization conditions with uncertainty communication led to users being slightly more confident about their judgement compared to visualization without uncertainty information. Study 3 builds on findings from Study 2 and explores differences in belief update when the bivariate visualization is congruent or incongruent with users' prior belief. Our results highlight the effects of incorporating uncertainty representation, and the potential of measuring belief updating on correlation judgement with Bayesian cognitive models.

Argus: Interactive a priori Power Analysis

A key challenge HCl researchers face when designing a controlled experiment is choosing the appropriate number of participants, or sample size. A priori power analysis examines the relationships among multiple parameters, including the complexity associated with human participants, e.g., order and fatigue effects, to calculate the statistical power of a given experiment design. We created Argus, a tool that supports interactive exploration of statistical power: Researchers specify experiment design scenarios with varying confounds and effect sizes. Argus then simulates data and visualizes statistical power across these scenarios, which lets researchers interactively weigh various trade-offs and make informed decisions about sample size. We describe the design and implementation of Argus, a usage scenario designing a visualization experiment, and a think-aloud study.

Bayesian-Assisted Inference from Visualized Data

A Bayesian view of data interpretation suggests that a visualization user should update their existing beliefs about a parameter's value in accordance with the amount of information about the parameter value captured by the new observations. Extending recent work applying Bayesian models to understand and evaluate belief updating from visualizations, we show how the predictions of Bayesian inference can be used to guide more rational belief updating. We design a Bayesian inference-assisted uncertainty analogy that numerically relates uncertainty in observed data to the user's subjective uncertainty, and a posterior visualization that prescribes how a user should update their beliefs given their prior beliefs and the observed data. In a pre-registered experiment on 4,800 people, we find that when a newly observed data sample is relatively small (N=158), both techniques reliably improve people's Bayesian updating on average compared to the current best practice of visualizing uncertainty in the observed data. For large data samples (N=5208), where people's updated beliefs tend to deviate more strongly from the prescriptions of a Bayesian model, we find evidence that the effectiveness of the two forms of Bayesian assistance may depend on people's proclivity toward trusting the source of the data. We discuss how our results provide insight into individual processes of belief updating and subjective uncertainty, and how understanding these aspects of interpretation paves the way for more sophisticated interactive visualizations for analysis and communication.

Uncertain About Uncertainty: How Qualitative Expressions of Forecaster Confidence Impact Decision-Making With Uncertainty Visualizations

When forecasting events, multiple types of uncertainty are often inherently present in the modeling process. Various uncertainty typologies exist, and each type of uncertainty has different implications a scientist might want to convey. In this work, we focus on one type of distinction between direct quantitative uncertainty and indirect qualitative uncertainty. Direct quantitative uncertainty describes uncertainty about facts, numbers, and hypotheses that can be communicated in absolute quantitative forms such as probability distributions or confidence intervals. Indirect qualitative uncertainty describes the quality of knowledge concerning how effectively facts, numbers, or hypotheses represent reality, such as evidence confidence scales proposed by the Intergovernmental Panel on Climate Change. A large body of research demonstrates that both experts and novices have difficulty reasoning with quantitative uncertainty, and visualizations of uncertainty can help with such traditionally challenging concepts. However, the question of if, and how, people may reason with multiple types of uncertainty associated with a forecast remains largely unexplored. In this series of studies, we seek to understand if individuals can integrate indirect uncertainty about how "good" a model is (operationalized as a qualitative expression of forecaster confidence) with quantified uncertainty in a prediction (operationalized as a quantile dotplot visualization of a predicted distribution). Our first study results suggest that participants utilize both direct quantitative uncertainty and indirect qualitative uncertainty when conveyed as quantile dotplots and forecaster confidence. In manipulations where forecasters were less sure about their prediction, participants made more conservative judgments. In our second study, we varied the amount of quantified uncertainty (in the form of the SD of the visualized distributions) to examine how participants' decisions changed under different combinations of quantified uncertainty (variance) and qualitative uncertainty (low, medium, and high forecaster confidence). The second study results suggest that participants updated their judgments in the direction predicted by both qualitative confidence information (e.g., becoming more conservative when the forecaster confidence is low) and quantitative uncertainty (e.g., becoming more conservative when the variance is increased). Based on the findings from both experiments, we recommend that forecasters present qualitative expressions of model confidence whenever possible alongside quantified uncertainty.

Increasing Interpretability of Bayesian Probabilistic Programming Models Through Interactive Representations

Bayesian probabilistic modeling is supported by powerful computational tools like probabilistic programming and efficient Markov Chain Monte Carlo (MCMC) sampling. However, the results of Bayesian inference are challenging for users to interpret in tasks like decision-making under uncertainty or model refinement. Decision-makers need simultaneous insight into both the model's structure and its predictions, including uncertainty in inferred parameters. This enables better assessment of the risk overall possible outcomes compatible with observations and thus more informed decisions. To support this, we see a need for visualization tools that make probabilistic programs interpretable to reveal the interdependencies in probabilistic models and their inherent uncertainty. We propose the automatic transformation of Bayesian probabilistic models, expressed in a probabilistic programming language, into an interactive graphical representation of the model's structure at varying levels of granularity, with seamless integration of uncertainty visualization. This interactive graphical representation supports the exploration of the prior and posterior distribution of MCMC samples. The interpretability of Bayesian probabilistic programming models is enhanced through the interactive graphical representations, which provide human users with more informative, transparent, and explainable probabilistic models. We present a concrete implementation that translates probabilistic programs to interactive graphical representations and show illustrative examples for a variety of Bayesian probabilistic models.

Illusion of Causality in Visualized Data

Students who eat breakfast more frequently tend to have a higher grade point average. From this data, many people might confidently state that a before-school breakfast program would lead to higher grades. This is a reasoning error, because correlation does not necessarily indicate causation - X and Y can be correlated without one directly causing the other. While this error is pervasive, its prevalence might be amplified or mitigated by the way that the data is presented to a viewer. Across three crowdsourced experiments, we examined whether how simple data relations are presented would mitigate this reasoning error. The first experiment tested examples similar to the breakfast-GPA relation, varying in the plausibility of the causal link. We asked participants to rate their level of agreement that the relation was correlated, which they rated appropriately as high. However, participants also expressed high agreement with a causal interpretation of the data. Levels of support for the causal interpretation were not equally strong across visualization types: causality ratings were highest for text descriptions and bar graphs, but weaker for scatter plots. But is this effect driven by bar graphs aggregating data into two groups or by the visual encoding type? We isolated data aggregation versus visual encoding type and examined their individual effect on perceived causality. Overall, different visualization designs afford different cognitive reasoning affordances across the same data. High levels of data aggregation by graphs tend to be associated with higher perceived causality in data. Participants perceived line and dot visual encodings as more causal than bar encodings. Our results demonstrate how some visualization designs trigger stronger causal links while choosing others can help mitigate unwarranted perceptions of causality.

In Pursuit of Error: A Survey of Uncertainty Visualization Evaluation

Understanding and accounting for uncertainty is critical to effectively reasoning about visualized data. However, evaluating the impact of an uncertainty visualization is complex due to the difficulties that people have interpreting uncertainty and the challenge of defining correct behavior with uncertainty information. Currently, evaluators of uncertainty visualization must rely on general purpose visualization evaluation frameworks which can be ill-equipped to provide guidance with the unique difficulties of assessing judgments under uncertainty. To help evaluators navigate these complexities, we present a taxonomy for characterizing decisions made in designing an evaluation of an uncertainty visualization. Our taxonomy differentiates six levels of decisions that comprise an uncertainty visualization evaluation: the behavioral targets of the study, expected effects from an uncertainty visualization, evaluation goals, measures, elicitation techniques, and analysis approaches. Applying our taxonomy to 86 user studies of uncertainty visualizations, we find that existing evaluation practice, particularly in visualization research, focuses on Performance and Satisfaction-based measures that assume more predictable and statistically-driven judgment behavior than is suggested by research on human judgment and decision making. We reflect on common themes in evaluation practice concerning the interpretation and semantics of uncertainty, the use of confidence reporting, and a bias toward evaluating performance as accuracy rather than decision quality. We conclude with a concrete set of recommendations for evaluators designed to reduce the mismatch between the conceptualization of uncertainty in visualization versus other fields.

A Framework for Externalizing Implicit Error Using Visualization

This paper presents a framework for externalizing and analyzing expert knowledge about discrepancies in data through the use of visualization. Grounded in an 18-month design study with global health experts, the framework formalizes the notion of data discrepancies as implicit error, both in global health data and more broadly. We use the term implicit error to describe measurement error that is inherent to and pervasive throughout a dataset, but that isn't explicitly accounted for or defined. Instead, implicit error exists in the minds of experts, is mainly qualitative, and is accounted for subjectively during expert interpretation of the data. Externalizing knowledge surrounding implicit error can assist in synchronizing, validating, and enhancing interpretation, and can inform error analysis and mitigation. The framework consists of a description of implicit error components that are important for downstream analysis, along with a process model for externalizing and analyzing implicit error using visualization. As a second contribution, we provide a rich, reflective, and verifiable description of our research process as an exemplar summary toward the ongoing inquiry into ways of increasing the validity and transferability of design study research.

Visualizing Uncertain Tropical Cyclone Predictions using Representative Samples from Ensembles of Forecast Tracks

A common approach to sampling the space of a prediction is the generation of an ensemble of potential outcomes, where the ensemble's distribution reveals the statistical structure of the prediction space. For example, the US National Hurricane Center generates multiple day predictions for a storm's path, size, and wind speed, and then uses a Monte Carlo approach to sample this prediction into a large ensemble of potential storm outcomes. Various forms of summary visualizations are generated from such an ensemble, often using spatial spread to indicate its statistical characteristics. However, studies have shown that changes in the size of such summary glyphs, representing changes in the uncertainty of the prediction, are frequently confounded with other attributes of the phenomenon, such as its size or strength. In addition, simulation ensembles typically encode multivariate information, which can be difficult or confusing to include in a summary display. This problem can be overcome by directly displaying the ensemble as a set of annotated trajectories, however this solution will not be effective if ensembles are densely overdrawn or structurally disorganized. We propose to overcome these difficulties by selectively sampling the original ensemble, constructing a smaller representative and spatially well organized ensemble. This can be drawn directly as a set of paths that implicitly reveals the underlying spatial uncertainty distribution of the prediction. Since this approach does not use a visual channel to encode uncertainty, additional information can more easily be encoded in the display without leading to visual confusion. To demonstrate our argument, we describe the development of a visualization for ensembles of tropical cyclone forecast tracks, explaining how their spatial and temporal predictions, as well as other crucial storm characteristics such as size and intensity, can be clearly revealed. We verify the effectiveness of this visualization approach through a cognitive study exploring how storm damage estimates are affected by the density of tracks drawn, and by the presence or absence of annotating information on storm size and intensity.

Hypothetical Outcome Plots Help Untrained Observers Judge Trends in Ambiguous Data

Animated representations of outcomes drawn from distributions (hypothetical outcome plots, or HOPs) are used in the media and other public venues to communicate uncertainty. HOPs greatly improve multivariate probability estimation over conventional static uncertainty visualizations and leverage the ability of the visual system to quickly, accurately, and automatically process the summary statistical properties of ensembles. However, it is unclear how well HOPs support applied tasks resembling real world judgments posed in uncertainty communication. We identify and motivate an appropriate task to investigate realistic judgments of uncertainty in the public domain through a qualitative analysis of uncertainty visualizations in the news. We contribute two crowdsourced experiments comparing the effectiveness of HOPs, error bars, and line ensembles for supporting perceptual decision-making from visualized uncertainty. Participants infer which of two possible underlying trends is more likely to have produced a sample of time series data by referencing uncertainty visualizations which depict the two trends with variability due to sampling error. By modeling each participant's accuracy as a function of the level of evidence presented over many repeated judgments, we find that observers are able to correctly infer the underlying trend in samples conveying a lower level of evidence when using HOPs rather than static aggregate uncertainty visualizations as a decision aid. Modeling approaches like ours contribute theoretically grounded and richly descriptive accounts of user perceptions to visualization evaluation.