IEEE VIS 2022 happened between October 16th and 22nd in Oklahoma City and I was really happy to have been able to attend in person. The very-earnest motto of the conference was “VIS ’22 is about the people you meet,” and for me at least, it really was! I’ll note that my strategy of writing a blog post describing my research and interests, tweeting it ahead of time, and posting it to the Discord really helped with meeting people, so I will definitely be doing that again. That said, the personal connections I made aren’t that interesting to read about, so here is my list of highlights from the conference content. If you prefer a longer recap with more photos, Tamara Munzner’s traditional epic yearly VIS twitter thread is also available. Tamara also went on the Data Stories podcast if you prefer your highlights in audio format.

Sunday

Sunday was the first of two days of “co-located workshops and events:” basically eight parallel tracks of half-day or full-day events. I attended the Visualization for Data Science (VDS) workshop (videos: session 1, session 2) in the morning and really enjoyed the closing keynote by Remco Chang. In the session, he described an interesting idea for a Grammar of Hypotheses which might in the future be used to describe what types of  questions can be answered by a given data set or visualization or asked by an analyst, in the hopes of matching them up.

I sat in on the workshop on Visualization for the Digital Humanities (Vis4DH) (videos: 1, 2, 3, 4) in the afternoon and liked the keynote by Miriam Posner about the differences in approach to “data” between humanists and scientists. I won’t paraphrase it badly here, but some key differences include the idea that replication is not a goal in the humanities as it is in the sciences, and that humanists often seek to question/challenge existing categorizations rather than use/reuse them. “Uncertainty” is often about the idea of the category or its appropriateness, rather than numerical uncertainty about a particular value.

Monday

The keynote for the workshop on evaluation and BEyond – methodoLogIcal approaches for Visualization (BELIV) (if you can “believe” that acronym!) by Casey Fiesler (video) really made me think about the ethics of using data in academic or other work in a deeper way than before (and I like to think I’ve read and thought about this a lot already). She highlights that ethics is not just what your ethics committee checks for or what is legal or what is in the terms of use, and encourages people to think like science fiction writers (think: Black Mirror) and imagine the possible terrible ethical consequences of their research.

The closing panel of the VisInPractice workshop (videos: 1, 2, 3, 4) was on Integrating Research and Products, and I appreciated the comments of Richard Brath from Uncharted about how visualizations that really work for domain experts (rather than novices) often seem to contravene strongly held beliefs researchers have about “the right way” to do visualization, like using charts with 2 (or more! gasp!) y-axes.

Tuesday

The opening day for the conference proper was highlighted by the awards ceremonies (video), including best papers, papers that had stood the test of time, career awards, and others. Richard Brath wrote a nice piece about the opening keynote and its relationship to the closing one: both had lots to say about how text integrates with graphics (video: opening keynote by Marti Hearst).

Title slide from opening keynote by Marti Hearst: “Show It or Tell It?”

The paper that stood out to me the most was “Affective Learning Objectives for Communicative Visualizations” (preview video), which frames some aspects of vis that are less-often researched in a really nice way.

Wednesday

This was the first of two days with four paper sessions each. The first session was the one that most specifically overlapped with my interests: Transforming Tabular Data and Grammars (video). There were some super-charged pivot tables presented, such as HiTailor (preview video), but the “Animated Vega-Lite” (preview video) and “No Grammar to Rule Them All” (preview video) papers were what really got me fired up. I loved the expert-interview evaluation of “Animated Vega-Lite” and how they were able to add really composable/low-viscosity animation abstractions into Vega-Lite. “No Grammar to Rule Them All” made so many good points and framed things so nicely I was nodding along the whole time I was reading it. I really liked the framing of colloquial versus formal models, for example. 

During the Understanding and Modeling How People Respond to Visualizations session (video), a handy Survey of Perception-Based Visualization Studies by Task was presented, which I’ll likely be referring back to as a resource.

The Short Papers session on “Visualization Systems and Graph Visualization” (video) included a paper on VegaFusion which I co-authored, but I also enjoyed hearing about the NL4Vis (Natural Language for Visualization) improvements in the “Facilitating Conversational Interaction in Natural Language Interfaces for Visualization” paper. I also of course really enjoyed seeing the Plotly-Resampler being presented at VIS.

Slide from Eli Holder’s thought-provoking presentation on “Dispersion vs Disparity.”

The final session of the day before the amazing banquet at the First Americans Museum was about Uncertainty (video). The paper that stuck with me from that session (and really the whole conference) was the striking “Dispersion vs Disparity: Hiding Variability Can Encourage Stereotyping When Visualizing Social Outcomes” (preview video), which shines a light on a visualization pitfall I hadn’t thought of before but now can’t unsee.

Thursday

During the Interactive Dimensionality (High Dimensional Data) session (video), I was a little puzzled by some of the papers on “steering” embeddings, but “VERTIGo: A Visual Platform for Querying and Exploring Large Multilayer Networks” (preview video) seemed like a solid UI for querying multilayer graphs.

At the Graphs and Networks session (video), I really liked the presentation on Taurus: Towards A Unified Force Representation and Universal Solver for Graph Layout (preview video), which seemed to really neatly unify a bunch of algorithms into one equation, propose a new algorithm based on that equation, and then propose a single efficient way to solve it. Visualizing Graph Neural Networks with CorGIE: Corresponding a Graph to Its Embedding (preview video) also looks like a clever way of understanding Graph Neural Networks, and I loved the idea of mapping a CIELAB a/b slice to the x/y coordinates of an embedding so as to provide a consistent node color across views!

During the Reflecting on the Field session (video), the paper “Visualization Design Practices in a Crisis: Behind the Scenes with COVID-19 Dashboard Creators” (preview video) talked about COVID dashboards as “boundary objects”— which were the site of negotiations between many parties during a crisis and reminded me of the video I made by stringing together 470 static official Quebec dashboards. This paper made me wonder about the design decisions that drove them.

The final session I attended on Thursday was on Provenance and Guidance (video), and I thought that the “Medley: Intent-based Recommendations to Support Dashboard Composition” (preview video) and “GEViTRec: Data Reconnaissance Through Recommendation Using a Domain-Specific Visualization Prevalence Design Space” (preview video) papers were nice, complementary approaches to semi-automated dashboard creation.

Friday

There was only one paper session before the capstone address and closing, on Comparisons (video). The View Composition Algebra for Ad Hoc Comparison (preview video) is a simple yet powerful concept based on the idea of comparison as a basic interaction.

Title slide from capstone keynote by Kerry Magruder: “Galileo’s Telescopic Discoveries.”

The capstone by Kerry Magruder on Thinking Visually and Galileo’s Telescopic Discoveries was just wonderful (video). Edward Tufte talks about Galileo’s illustrations in his work, but this was a much more in-depth exploration of this topic. Apparently there was a tour of the History of Science library at Oklahoma University the night before where some folks got to see these manuscripts, but I didn’t realize it was happening until after.

Takeaways

It’s really hard to summarize a week-long conference with multiple tracks like this, but the main takeaway for me from this conference is that the IEEE VIS community is full of really fascinating, open, and excited people doing really interesting, multi- and cross-disciplinary research. I would recommend anyone interested in visualization to check out some of the videos above and maybe consider attending either remotely or in-person if it’s nearby enough to travel (next year it will be in Melbourne!). 

Coda

I didn’t attend the alt.VIS workshop, but looking at this blog post, the paper entitled “* (Name Of This Paper Can Be Automatically Generated)” presented there is pretty spot-on!

 | Website

Nicolas Kruchten builds data visualization and data analysis tools. He maintains Plotly.py and is the creator of Plotly Express as well as react-pivottable and pivottable.js. He is currently a graduate student at the École de Technologie Supérieure in Montreal, Canada and posts videos of talks and neat visualization projects at https://nicolas.kruchten.com/.