Mene Mene Tekel Upharsin: A Trip Report from IEEE VIS 2024
The following is my trip report from IEEE VIS 2024, the premier conference in data visualization research, held virtually, for reasons I will briefly get into. Listen, this was a bit of a rough one. Soundtrack provided below:
The most salient bit here is that this conference was a bit of a Lucy/Charlie Brown football fakeout situation. The conference was originally meant to be held at a beach hotel in St. Pete Beach, only for hurricane Helene to strike and flood the hotel two weeks beforehand, causing a last minute reschedule to downtown Tampa, which was then on course to be hit by hurricane Milton, which one of the conference executive committees (we have three such committees, and I am happy to discuss this seeming organizational surplus at length, but these posts already get a little too inside-baseball for my taste) deemed sufficiently alarming to cause a switch to fully remote at the last minute.
Pivoting a conference twice, changing both venue and modality, all within a two week period, is absolutely wild, and to me shows just how much of conference organizing is a bunch of unpaid low bus-number heroics. It did not make me feel great that the conference was making people, to use a concrete example, take time away from their newborn child in order to set up a bunch of complex and finicky interconnected systems in order to pull together an entire conference with a sort of feverish intensity. As mentioned in an earlier post, I have an intense distrust of heroics, and view their consistent presence in organized systems a sign of trouble.
But anyway, as a result of the last-minute pseudo-cancellation, the Vibes Were Off, conference-wise. I can’t tell how much of this is projection (I’m estimating in the high 80%s), but it felt like folks were demoralized, directionless, and prone to the kinds of hair-trigger scolding that happens when people who thought they were going to be on the beach end up instead waking up early in the morning to stare at youtube streams. Just generally gloomy and defeated (I even wrote “funereal” in my notes) but also with the awareness that it’s kind of a gauche or at the very least kind of silly thing to feel gloomy and defeated about (I didn’t get to spend the university’s money going to the beach to hang out with my friends? I’m sure all of the people who were more directly and horrifyingly impacted by the back-to-back hurricanes must really be sympathizing with me right now).
I also saw the solidification of some trends that I was starting to fret about after my first set of virtual conferences, things that seem obvious in hindsight but still kind of worry me:
- Communication channels are underused and imbalanced. When people talked, it was mostly senior people talking to other senior people. Just as with being first person to ask a question after an academic talk, it takes a certain amount of comfort and chutzpah to be the first person to join a virtual hallway, or chat in a discord channel, or participate in a Zoom breakout room. Especially when you know that all of your academic heroes (or rivals) could be watching you just on the other side of the screen, silently (or not so silently) judging you. I was mostly talking to the same half dozen or so people all week.
- Everything still feels one-off and experimental. This issue was especially exacerbated by the extent to which everybody was in “crisis mode.” Since folks had little time to plan, we reverted to our default solutions from prior hybrid events, which were themselves impromptu ad hoc solutions created by pandemic-driven necessity. Speaking for myself, I was hesitant to even think about exploring the potential of the virtual conference medium because I didn’t want to break something, create more work for already over-worked people, or lose people who were already overwhelmed juggling two or three browser windows worth of stuff to navigate.
But all of this is how the medium made things feel kind of down. I also want to talk about the message. Attendance and submission counts have mostly stabilized and acceptance rates are a bit down. So we’re not really growing, and we’re also being a bit more conservative, but that might be over-indexing on what can be noisy year-to-year trends. But there was actually a moment during the opening keynote (by Bill Pike, the Chief Science and Technology Officer at the Pacific Northwest National Laboratory) where I felt a bit of that pit-of-my-stomach kind of existential despair. Dr. Pike has extensive experience in visual analytics, working both with and as a government funder of scientific research. The first question he was asked was naturally about how to get funders interested in visualization and visual analytics in a world where AI and ML have sucked up so much of the dollars and the enthusiasm. Here’s his reply:
I think it goes back to really trying to respect and assert the role for humans in the work that those funding agencies are trying to support. And I’ll admit that if we start sometimes with the “V”-word, we can stumble. If we start with the “H”-word, the human word, we find ourselves in a much more effective place at convincing funding agencies, whether basic or applied, to support visualization research.
👏 emoji reacts in the discord, etc. etc. He’s certainly not wrong here, and I generally appreciate the sentiment. We should be putting the people and their goals first, and the technology second — we shouldn’t be technologists who just have one hammer and so see every problem as a nail (or at least not if we want to actually solve real problems; although then there’s the issue that graduate school seems structurally organized around generating single-hammer researchers). But the immediate knee-jerk response I had was: “sure, but AI people don’t have to be quite so coy, do they?” I don’t see ML/AI folks hiding the terminology of their entire field as a dirty word that will make people think of them as boring or outdated or unhelpful. Well, not at the current moment, anyway (sparing a prayer for all of the symbolists and expert systems folks from the last couple of AI winters). Again, I don’t think he’s wrong here, and there are positive ways of viewing this statement as a sign of the field’s success rather than failure (hey, maybe it means that people feel generally empowered with good tools and good guidelines to solve their own data visualization problems without having to go begging academics for help: that’s not so bad of a way to be less trendy, as a research field). But it didn’t feel great to me, in the moment.
Okay, with the acknowledgment of the stormy weather in my personal headspace at the time, let’s get to the content. Normally I either use this particular prelude to either produce a theme or apologize for not having noticed a theme, but I think if you read all of the following with a mental “hey, lighten up, buddy” you’ll get into the right mood without me having to do any additional thematic clustering.
Talks
“The Golden Age of Visualization Dissensus”
Alberto Cairo
To fulfill one of my course requirements in college, I took the intro to biology course for non-majors. I forget the exact course title because everybody generally called it “Disney Bio”, the connotation here being that this course skipped over the gnarlier and more complicated or ambiguous bits of biology to leave you with happy animals, singing birds, and simplified lessons that the whole family could enjoy.
What Alberto lays out in this talk is that our conception of data visualization is likewise a bit “Disney-fied, ” sanitizing or downplaying the (occasionally sordid and turbulent) genealogy of our field, and flattening and simplifying the actual practice of data visualization to focus on a narrow set of use cases and analytical goals. His provocation here is based on the Alfred North Whitehead bon mot “the safest general characterization of the European philosophical tradition is that it consists of a series of footnotes to Plato”, asserting that data visualization often feel like “footnotes to Tufte,” resulting in a very “Apollonian” view of data visualization (as being about simplicity and authority and “clarity” and so on) as opposed to a more “Dionysian” visualization (that is more about hedonics and aesthetics and affect etc.)†.
I had just given a talk 24 hours earlier where I had argued for a similar need to quit thinking of visualization in limiting “Tufteist” and “anti-Tufteist” terms but instead seek a new synthesis. And we had a paper at CHI this year that likewise argues for the need for counter-histories of data visualization. So I was receptive to the message here. But I did wonder how much of this is just computer science’s fault for gettings its grubby hands on data visualization research, rather than some intrinsic deficiency in data visualization pedagogy or practice. I also had an extended conversation with Danyel Fisher about how much of this Apollonian/Dionysian dichotomy is really just disguising the very real differences in goals, looks, feels, and pedagogy in visualizations designed for data journalism as opposed to business intelligence, statistical graphics, or data art.
Visualizing Inequality: What we can learn from grassroots data activism
Catherine D’Ignazio
Speaking of data art… Catherine’s conference capstone talk I think was an extremely nice bookend to Alberto’s framework of Apollonian/Dionysian visualization. The provocation in her talk is that we often look at visualizations like the ones associated with the New York Times’ “Extensive Data Shows Punishing Reach of Racism for Black Boys” that show massive gender or racial or class inequalities, and asked if these visualizations are “doing the work that we want them to do? And what is it, precisely, that we want them to do?” To strip the nuance from her argument here, there’s a critique that these visualizations highlight difference and disfunction, potentially portraying the groups they are focusing on as damaged or in need of patronizing help by an exterior “savior” rather than providing agency for these groups or providing a clear path forward for meaningful change.
I think there’s something to this critique. In line with Alberto’s “Apollonian” data visualization, I feel like there’s a strong distaste in data visualization circles for taking the steps beyond just “spreading awareness,” I presume for fear of being “biased.” For instance, an assumption that if we simply just show people enough charts about climate change, it will make the needed societal shifts just happen automatically after there’s a sufficiently large set of people “aware” that the world is warming. An assumption that the goal of a data visualization designer is, or should only be, “just showing the data,” but stopping a step short of advocating for what should be done now that the audience has seen what the data shows. I think, for many of these issues, “awareness” is not the stumbling block to meaningful reform or change: it’s combatting (often entrenched) interests, allocating (or re-allocating) resources, and stirring people to specific action.
A large chunk of the remainder of the talk was highlighting specific examples of data-driven advocacy and protest which were generally pretty cool, so I had to a do a little work to build up the right amount of unease, but I did get there in the end. What I worried myself about is: why would any of the people doing actual, impactful data advocacy give a single solitary shit about our research community? Are we giving them any useful information, either directly or after a few levels of sci-comm translation or communication? Are we giving them useful tools or resources, critical support or large platforms? I mean, maybe, sometimes. But not very often.
Papers
We Don’t Know How to Assess LLM Contributions in VIS/HCI
Anamaria Crisan
For the past few sets of these conference trip reports, I have often given the caveat that I won’t be discussing much about ML or related papers, mostly because I feel like I have been largely (and unproductively) repeating myself on the subject (why I feel this way about ML but not on all the other subjects which I have been unproductively repeating myself is left as an exercise to the reader and/or a mental health professional). This constraint has been increasingly felt as ML and LLMs grab increasing shares of the oxygen in all of the rooms they are in (my back of the envelope calculation from bidding on papers for CHI 2025 was that something like slightly over a third of submissions had an AI angle of some sort, and I suspect that’s an undercount). I have felt equally uncomfortable when asked to review these papers, both because I’m not particularly in deep in the ML world, but also for similar reasons why I felt uncomfortable (okay, pissed off) about the glut of COVID visualization papers in 2020–2022: I am skeptical of research that just rides the coattails of current trends without doing the deep thinking or interdisciplinary engagement needed to have a real impact. Ian Arawjo in particular is skeptical of the surge in “LLM Wrapper papers” where you just find an HCI task, stick a text prompt that looks into an LLM in the interface somewhere, and call it a day.
But this position paper from Ana I think goes in a much more interesting direction than my usual instinct to make tut-tutting noises towards the trend-followers while continuing on with my day, which is to think about, when we do have interesting LLM contributions in visualization, are we actually set up to even know that we’re on to something? It’s easy enough to blame some abstract Other that writes an LLM paper without looking at the prior work or assessing the actual HCI contribution, but what about the people reviewing these papers and either being wowed by novelty, or (perhaps just as bad) being instinctually dismissive? For every “doing something with the new version of Chat GPT that came out a month or two ago just because it’s there”, there just might be an equal and opposite “rejecting a paper for not doing something with the new version of Chat GPT that came out a month or two ago.”
The paper focuses more on guidelines for reviewers, but a related issue I am worried about is who is actually doing all of this work? Assessing the contributions of a Vis+AI or Vis+LLM paper would seem to require somebody with a good background in both, and those are a bit of a rara avis in visualization conference reviewing pools at the moment. Again, anchoring on next year’s CHI, it seems like people with even a tentatively suggested expertise in AI or ML got bombarded with more review requests than they could handle, and even the lowest effort “LLM Wrapper” papers require non-trivial effort to correctly sort and identify. Somebody has to make that effort balance out somehow, and I’m not certain we’re well-situated as a field to make that happen.
The Effect of Visual Aids on Reading Numeric Data Tables
Yongfeng Ji, Charles Perin, Miguel A. Nacenta
Somebody close to me was preparing a presentation that included some suggestions for improving the UX of some health records data. They had tables with zebra striping and got some pushback and so asked me, as nominally a data visualization person with an interest in spreadsheets and tables, to dig up some of the assuredly extensive lit they could use as ammunition for their case. “No problem”, I said, and went on google scholar. My internal monologue was something like “well, I know tables aren’t too fashionable to study now, but surely there was a bunch of work in the early days of InfoVis, or, failing that, some of the earlier UX stuff from the 90s…” The result: bupkis.
This lacuna is wild to me. Are you telling me we’ve run who knows how many studies on people’s bespoke analytics tools that might never have more than one or two users, but we hadn’t done more than a couple of cursory explorations of spreadsheets and tables, the data analytics tools so prevalent that they make the entire field of visualization look like a rounding error? It was enough to make me pass around a link to Robert Kosara’s “An Empire Built On Sand: Reexamining What We Think We Know About Visualization” and raise an eyebrow.
So that’s why I’m highlighting this paper here, not so much for their actual findings (although I think they are interesting: zebra striping seems to help a bit for some tasks, highlighting helps with other tasks, which is enough to put me in the “tentatively pro-zebra” camp even if it’s not a slam dunk or unequivocal finding), but to give plaudits for somebody in the community actually, you know, looking into this stuff after so many years. Maybe in a few years the visualization community will look into pivot tables, console statements, slide decks, or any of the other ways that actual people tend to get their data rather than all the wacky ad hoc techniques we designing.
Entanglements for Visualization: Changing Research Outcomes through Feminist Theory
Derya Akbaba, Lauren Klein, Miriah Meyer
I was actually going to highlight another paper that Derya was a co-author on this year (Discursive Patinas: Anchoring Discussions in Data Visualizations by the dream team of Tobias Kauer, Derya Akbaba, Marian Dörk, and Benjamin Bach) because that patina work hit on a bunch of topics I think are really exciting whenever they pop up in these kinds of conferences over the years (naturalistic metaphors like patinas and desire paths for exploring interaction histories or just generally making visualizations feel less “lonely” and and meaning more mutually constructed), but it felt remiss not to highlight this not-entirely-unrelated work instead, if for no other reason than it’s very squarely in my wheelhouse. This paper uses Karen Barad’s notion of entanglements to argue for a re-thinking of epistemology in visualization.
You don’t need to know what “epistemology” is in general, or even what makes an epistemology feminist rather than non-feminist, to get some value out of this paper. I think just having an idea that all of the important parts of data visualization (like the “data” itself, but also the “insights” from the data, or the final visualization design) are not these independent things that you can scientifically poke at in isolation, but “tentacularly” interconnected phenomenon, provides some value. And all of these things are also “soft” in a way: able to be disputed, re-evaluated, or modified. I’m actually not too mad about how “positivist” visualization methods can be in experimental design or whatever, but, as I alluded to when discussing Catherine D’Ignazio’s capstone earlier, I am downright peeved when visualization designers think of data as this fixed and immutable thing and the job of the designer is to simply visualize the stuff in as “hands off” a manner as possible, without critical engagement: it is a profoundly incurious and ethically troubling way to work with data, and we can and should do better.
The other thing that I want to use this paper for (and, in fact, already have started doing in conversations), is as evidence that the data visualization community is not (if it ever was, even in the most uncharitable histories), a bunch of computer scientist with cargo cult theories methodologies borrowed from half-remembered psychology lectures, but, you know, a full-fledged and diverse field of people with backgrounds ranging from graphic design to perceptual psychology, science and technology studies to applied mathematics. A strawman “naïve scientist” conception of the field has been used by senior people to justify all kinds of stultifying conservatism (“we know that a quantitative analysis is not appropriate here, but we know reviewers expect a study, so…” or “we thought we made an interesting tool and we learned interesting things from it but there were no novel visualization techniques employed, so there’s no way we’d get it published”, etc. etc.), and I’m tired of it (and I’m likewise tired of the opposite trend of folks jumping in from outside the field to try to score easy points about how naïve and blinkered we all are, and isn’t it a shame none of us have read any critical theory). A paper on honest-to-Haraway feminist epistemology got the best paper this year! You really can lean on the diversity and accommodation of the field and take a few chances! More on how I feel academics use the existence of a Lacanian “little Other” (who does not necessarily exist as such, or exists only as a projection of the ego) as a flimsy excuse to not push boundaries or launder the “anticipatory compliance” that is inherent in academia coming in a follow-up post once I figure out what I’m actually mad at and how to be mad in a productive way.
Wrapup
One of the silver linings of the switch to virtual is that most of the content is trickling its way onto YouTube, so you really don’t have to take my word for any of this and can dig through the material yourself. We seem to perhaps be swinging the pendulum in the other direction, however, and the organizers of next year’s conference in Vienna (following up from last year’s in Melbourne, which I didn’t attend because it was super far and super expensive and I had just started a new gig) have stated that they will not support the remote (or even hybrid) participation that I’ve gotten used to over the years. I believe it was Miriah Meyer who started this discussion on Discord, but I wonder if these contradictory desires to both be inclusive and global but also to support the kinds of in-person direct serendipitous mixing that everybody seems to admit isn’t really working in virtual venues is a (likely doomed) effort to try to recapture some sort of pre-pandemic (or even earlier, if this imagined past even really existed) “golden age” of academic conferences, where the field was small enough that you could run into old friends at every coffee break, be conversant about all of the emerging work with new friends, and have an appropriate but not unseemly amount of fun. And, well, I don’t know if we (as a field, as an organization, or even as parts of academia as a whole) are structurally able to produce those kinds of experiences anymore.
But I could be wrong; I’m happy to be pleasantly surprised.
* Bonus song:
† My eyes lit up when Alberto made this reference because we were assigned Thomas Mann’s Death in Venice in high school (I think sophomore year? So something like age 13–14 for me) and one of the assignments was to make a piece of art that dealt with the themes of the book. I made this painting that was meant to be a quasi-cubist interpretation of the narrator’s struggle with the Apollonian versus the Dionysian and my teacher hated it and I think I got a C or something, which, okay, it’s not great, but my parents kept it and so it was hanging in my bedroom throughout my adolescence as a sign of my mediocrity: