Keeping abreast of current trends, technologies, and best practices in visualization and data analysis is becoming increasingly difficult, especially for fledgling data scientists. In this paper, we propose Lodestar, an interactive computational notebook that allows users to quickly explore and construct new data science workflows by selecting from a list of automated analysis recommendations. We derive our recommendations from directed graphs of known analysis states, with two input sources: one manually curated from online data science tutorials, and another extracted through semi-automatic analysis of a corpus of over 6,000 Jupyter notebooks. We evaluate Lodestar in a formative study guiding our next set of improvements to the tool. Our results suggest that users find Lodestar useful for rapidly creating data science workflows.

We propose topology-aware space distortion (TASD), a family of interactive layout techniques for non-linearly distorting geometric space based on user attention and on the structure of the visual representation. TASD seamlessly adapts the visual substrate of any visualization to give more screen real estate to important regions of the representation at the expense of less important regions. In this paper, we present a concrete TASD technique that we call ZoomHalo for interactively distorting a two-dimensional space based on a degree-of-interest (DOI) function defined for the space. Using this DOI function, ZoomHalo derives several areas of interest, computes the available space around each area in relation to other areas and the current viewport extents, and then dynamically expands (or shrinks) each area given user input. We use our prototype to evaluate the technique in two user studies, as well as showcase examples of TASD for node-link diagrams, word clouds, and geographical maps.

Visualization interfaces designed for heterogeneous devices such as wall displays and mobile screens must be responsive to varying display dimensions, resolution, and interaction capabilities. In this paper, we report on two user studies of visual representations for large versus small displays. The goal of our experiments was to investigate differences between a large vertical display and a mobile hand-held display in terms of the data comprehension and the quality of resulting insights. To this end, we developed a visual interface with a coordinated multiple view layout for the large display and two alternative designs of the same interface---a space-saving boundary visualization layout and an overview layout---for the mobile condition. The first experiment was a controlled laboratory study designed to evaluate the effect of display size on the perception of changes in a visual representation, and yielded significant correctness differences even while completion time remained similar. The second evaluation was a qualitative study in a practical setting and showed that participants were able to easily associate and work with the responsive visualizations. Based on the results, we conclude the paper by providing new guidelines for screen-responsive visualization interfaces.

Tracking the sensemaking process is a well-established practice in many data analysis tools, and many visualization tools facilitate overview and recall during and after exploration. However, the resulting communication materials such as presentations or infographics often omit provenance information for the sake of simplicity. This unfortunately limits later viewers from engaging in further collaborative sensemaking or discussion about the analysis. We present a design study where we introduced visual provenance and analytics to urban transportation planning. Maintaining the provenance of all analyses was critical to support collaborative sensemaking among the many and diverse stakeholders. Our system, StoryFacets, exposes several different views of the same analysis session, each view designed for a specific audience: (1) the trail view provides a data flow canvas that supports in-depth exploration+provenance (expert analysts); (2) the dashboard view organizes visualizations and other content into a space-filling layout to support high-level analysis (managers); and (3) the slideshow view supports linear storytelling via interactive step-by-step presentations (laypersons). Views are linked so that when one is changed, provenance is maintained. Visual provenance is available on demand to support iterative sensemaking for any team member.

Causality visualization can help people understand temporal chains of events, such as messages sent in a distributed system, cause and effect in a historical conflict, or the interplay between political actors over time. However, as the scale and complexity of these event sequences grows, even these visualizations can become overwhelming to use. In this paper, we propose the use of textual narratives as a data-driven storytelling method to augment causality visualization. We first propose a design space for how textual narratives can be used to describe causal data. We then present results from a crowdsourced user study where participants were asked to recover causality information from two causality visualizations--causal graphs and Hasse diagrams--with and without an associated textual narrative. Finally, we describe CAUSEWORKS, a causality visualization system for understanding how specific interventions influence a causal model. The system incorporates an automatic textual narrative mechanism based on our design space. We validate CAUSEWORKS through interviews with experts who used the system for understanding complex events.

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.