iiinova — where your digital self keeps evolving

Eli Holder @elibryan.bsky.social · 16/03/2026

Population change is scary! The fear that "our people" are disappearing fuels "great replacement" myths and shows up in charts. Demagogues divide "us vs. them" for power and clicks. We all risk contributing “data demagoguery” when we use social division in charts to win the attention game.

Mussolini saying stupid s*** “The entire white race, the Western race, could be submerged...” overlaid on a collage showing threatening population change charts.

020

Eli Holder @elibryan.bsky.social · 16/03/2026

12,000 years of "legitimizing myths" say hard work makes you rich (and :. poverty means lazy). These beliefs shape how we read charts. Highlighting systemic influences, e.g. Heiserman et al's 2020 parallel coordinate "ladders," can help bust these myths, calibrating fairness judgments.

A billionaire with money to burn... and outraged capuchin monkeys, with the question “How is [inequality] okay?”

A slide with king solomon, martin luther, john calvin, and ben franklin. It shows a quote from Franklin’s Poor Richard’s Almanac “God helps those who help themselves.”

A slide with outraged monkeys overlaid over a parallel coordinate “ladder” plot showing the “low mobility” stimulus from Heiserman et al 2020’s “Judgments of Economic Fairness Are Based More on Perceived Economic Mobility Than Perceived Inequality”

221

Eli Holder @elibryan.bsky.social · 16/03/2026

Charts are like inkblots, letting people “see what they want” in the data. So charts reinforce the status quo when we see them as evidence supporting our existing cultural beliefs.

Eugenicist Carl Brigham in front of a chart splattered in inkblot ink with text that says “Charts work like inkblots, and citations for Xiong Bearfield et al 2022 “Seeing What You Believe or Believing What You See? Belief Biases Correlation Estimation” and Luo & Zhao 2019 ”Motivated Attention in Climate Change Perception and Action”

120

Eli Holder @elibryan.bsky.social · 16/03/2026

“Why can’t we have nice things?” A few highlights from my talk for Dartmouth’s “Data in Action” symposium, exploring civic dataviz and democracy. What makes collective action so difficult in the United States? How could that possibly depend on charts 📊?! What can we do?

title slide asks “Why can’t we have nice things?” It shows a grumpy George Washington standing in front of Library building and a bunch of population / social outcome charts

Talk outline. We can’t have nice things because of misbeliefs, complacency, stereotypes, and threat, each of which seem to interact with how we make sense of charts and graphs visualizing populations and social outcomes.

230

Eli Holder @elibryan.bsky.social · 25/01/2026

From reddit, 144° difference in perceived temperature. 🥵🥶

Weather map of US with arrows point between Minnesota and Florida. It says "144° in feels like temp"

120

Eli Holder @elibryan.bsky.social · 03/09/2025

Fun sketches → Fun charts! 📊🎉

Pencil sketch of a diverging area chart with annotations

Pencil sketch of some kind of jitter plot

091

Eli Holder @elibryan.bsky.social · 24/08/2025

What can we do here? @elijfinkel.bsky.social speed dating work suggests a much less bleak dating world, vs Tinder's 80/20 dynamic: doi.org/10.70400/NEL... @jamieamemiya.bsky.social & co suggest counterfactuals to blame systems, not people: doi.org/10.1177/1745... Which maybe looks like this...

A quote describing Eli J. Finkel’s “Actor, Partner, Relationship” model of attraction. It reads: “Rather than having the 10s match with the 10s... most of us can find a partner who is something close to a 10 for us.”

There’s also a figure loosely illustrating this, with lots of people matching with lots of other people.

a bar chart showing swiping / liking behavior on tinder between different gender groups

text says: “Better: Blame tinder, not women” a play on “Blame systems, not people”

The original bar chart showing swiping / liking behavior on tinder between different gender groups, paired with a counterfactual bar chart showing liking behavior in speed dates for different gender groups.

The title says “Tinder Distorts Dating Behavior”

000

Eli Holder @elibryan.bsky.social · 24/08/2025

Genetic determinism also shows up in incel belief systems, tangled up with some wild takes on online dating statistics... (e.g. women's suspected urge to "trade up" beyond their "looksmatch"). Preston & co: doi.org/10.1177/1097... Vallerga & Zurbriggen: doi.org/10.1111/asap...

Slide showing Sydney Sweeney beside the Jamie Miller character from Netflix’s “Adolescence” series

Beside Sweeney is text: “Sydney Sweeney is genetically superior”
Beside Miller is text “Others are genetically inferior.”

This concept of genetic worth is part of both eugenics and the incel belief system

Slide showing a fake statistical statement about inequality in dating: “80% of women are only interested in the top 20% of men. So 80% of men have to fight for the bottom 20% of women.”

Beside it is a bar chart showing swiping / liking behavior on tinder between different gender groups, which are interpreted among incels as evidence of women’s “hypergamy”

A bar chart showing swiping / liking behavior on tinder between different gender groups. A figure loosely illustrating the presumed lookism 80 / 20 rule “80% of women are only interested in the top 20% of men. So 80% of men have to fight for the bottom 20% of women.”

To the right are different explanations offered for these disparities (and lack of dates)

110

Eli Holder @elibryan.bsky.social · 22/08/2025

What can we do (for the charts)? Part of the misattributions stem from design choices. ❌ Bar charts increased genetic and personal blame for health disparities. ✅ These "geo-emphasized" jitter plots shifted blame toward systemic explanations. Links + videos from our last paper: 3iap.com/mbat/

Slide showing a bar chart of public health disparities with the text “Save lives. Stop bar charts.”

From my deathly dataviz talk: https://3iap.com/rladies2025/

Slide showing the “geo emphasis” chart from our must be a tuesday study, with text “We can do better”

From must be a tuesday study: https://3iap.com/mbat/

How did viewers explain health disparities in our Must Be A Tuesday study (https://3iap.com/mbat/)? It shows that conventional bar charts significantly increased genetic attribution and personal blame. And that the geo-emphasis charts significantly shifted blame from personal toward external attributions.

Slide shows a simplisitc chart of outcomes, hiding variability, which says “less of this”
And a more complete jitter plot of outcomes, showing variability, which says “more of this”

140

Eli Holder @elibryan.bsky.social · 22/08/2025

Similar stories here: Good outcomes = good genes → "Do Black people really need vaccines or anesthesia?" Bad outcomes = bad genes → "There's nothing we can do to help!" These charts are obviously well-intended, but ours + other studies find that viewers often misattribute these disparities.

Demonstrating viewers’ explanations for health disparities based on viewing health disparity charts.

They often blame people (personal attribution) by explaining the disparities in terms of personal behavior or characteristics “Drinking and smoking”, “too much fried food”, “bad genes.” These may make sense at individual levels, but not for populations.

Or you could attribute these disparities to external social determinants of health like “access to care, good food”, “neighborhood pollution”, “structural racism”

quotes are paraphrased from qual responses seen in the “Must be a Tuesday” study https://doi.org/10.1109/VIS55277.2024.00021

Tracing a theory of action for how genetic misattributions can build into other beliefs that ultimately lead to tangible real world outcomes, like Black patients being misdiagnosed when presenting with pain.

From my deathly dataviz talk: https://3iap.com/rladies2025/

RFK Jr’s statement about Black vs White vaccine schedules. It shows his quote: “We should not be giving Black people the same vaccine schedule that’s given to whites, because their immune system is better than ours.”

Two news headlines showing Sen Roger Marshall and Sen Bill Cassidy blaming public health problems on individuals.

They read:
“GOP Senator Blames Americans For Most Of Their Health Problems”
https://www.rollingstone.com/politics/politics-news/roger-marshall-make-america-healthy-again-1235232716/

“Maternal death rate isn’t as bad if you don’t count Black women, GOP senator says”
https://www.businessinsider.com/gop-senator-la-outlier-maternal-death-rate-skewed-black-women-2022-5

130

Eli Holder @elibryan.bsky.social · 22/08/2025

Similar story here: Good outcomes = good behavior (us) → Bad outcomes = bad behavior (others) → "Should we spend so much $$$ vaccinating people who don't deserve it?"

A quote from RFK Jr “It’s exceedingly difficult for measles to kill a healthy individual ...there’s a connection between those who suffer from measles and individuals who lack proper nutrition or don’t engage in regular physical activity.”

A quote from Dr Oz, head of medicaid, “It’s your patriotic duty to be as healthy as you can.”

Two quotes from Attwell et al 2018’s focus group study of anti-vaxers.

Attwell et al 2018 ‘The Unhealthy Other’ How vaccine rejecting parents construct the vaccinating mainstream. https://doi.org/10.1016/j.vaccine.2018.01.076

News headline announcing cuts to mRNA vaccine research. It reads: RFK Jr cancels $500m in funding for mRNA vaccines for diseases like Covid”

https://www.bbc.com/news/articles/c74dzdddvmjo

110

Eli Holder @elibryan.bsky.social · 22/08/2025

How does this work for the lame jeans/genes ad? What's the argument behind the ad? What makes it funny (to some people)? Where does it lead? Good jeans → Good genes (us) → Bad genes (others) → "haha our genes _are_ great!" → "What if we ranked people based on genetics?!"

Slide with a photo from the American Eagle “good genes” ad with the subtext as actual text: “Sydney Sweeney is genetically superior”

Slide on the rhetoric of the “good genes” american eagle ad. It says “Genetic superiority is definitely a real thing. Hint: It isn’t.” It shows the eugenicist Karl Pearson looking excited and pointing to Sydney Sweeney.

Slide visualizing the eugenic idea that people can be stack ranked based on some simplistic unilinear concept of genetics. On the left it shows “genetically inferior” with an Elphaba-esque woman. On the right it shows “genetically superior” with the american eagle model

Slide with a fake photo of Sydney Sweeney holding up a bottle of bleach spray. It says “Sydney Sweeney supports sterilization... of kitchen surfaces.” The risk of dog-whistle ads is maybe clearer when considering where this kind of rhetoric has led historically, e.g. sterilization programs and genocide.

100

Eli Holder @elibryan.bsky.social · 22/08/2025

What do these have in common? It's not what they "say," it's "what do they suggest?" They suggest a world where "genes are destiny." They're blame-framed, so looks, love, and health can seem like immutable personal characteristics, rather than results of big messy social forces.

applying the "woke" critical analysis framework to the jeans ad. What does the ad claim technically? What does it suggest? How might people act on it?

Applying the rhetorical analysis framework to the health risk charts we tested in “Must be a Tuesday” https://doi.org/10.1109/VIS55277.2024.00021

It shows a bar chart with heart disease mortality for four different groups (which are stand ins for racial outcomes). It asks what does the chart suggest to some people about why heart disease mortality is higher for people in Group B?

applying the framework to a bar chart showing swiping / liking behavior on tinder between different gender groups. What does the chart show technically? What does it suggest? How might people act on it?

Applying the framework to RFK Jr’s statements about Black vs White vaccine schedules. It shows his quote: “We should not be giving Black people the same vaccine schedule that’s given to whites, because their immune system is better than ours.” and asks what does this suggest?

120

Eli Holder @elibryan.bsky.social · 22/08/2025

Things to do at SIGGRAPH: ✅ Behind-the-scenes FX secrets for Superman! ✅ Never-before-seen Oscar-worthy animation shorts!! 🤔 90 minutes with me... on "woke dataviz," eugenics, public health charts, fighting data demagoguery?!? A few new slides from my talk... 📊🧵...

photo from SIGGRAPH 2025 in Vancouver. it shows the word SIGGRAPH in big letters on the expo floor.

title slide from my talk: “Woke Dataviz for white guys.” It shows a flirty girl, a bodybuilder, a t rex, motorcycles, rockets, and a dork in military fatigues

Slide on the rhetoric of the “good jeans” american eagle ad. It says “Genetic superiority is definitely a real thing. Hint: It isn’t.” It shows the eugenicist Karl Pearson looking excited and pointing to Sydney Sweeney.

A collage of well-intended, but accidentally sketchy civic dataviz showing social outcome disparities for health, wealth, education, etc.

2102

Eli Holder @elibryan.bsky.social · 30/07/2025

The plots aren't exactly ggdist QDPs. Wanted a similar concrete story (1 dot ~= 1 person), but at 500 dots they're not exactly countable. Dots are also evenly-spaced on the x-axis, which we needed visually, but @mjskay.com + @racquelf.bsky.social might advise otherwise on precision tradeoffs! 😬

Closeup mockup showing the evenly spaced dots and how dot stacks align between plot rows.

210

Eli Holder @elibryan.bsky.social · 30/07/2025

This is also a (6-month late) response to @datarevelations.com's sagely haiku: "Make everything gray, except the few things you think should be highlighted" These WIP mocks show earlier color coding attempts–which I still like stylistically!–but monochrome won on function for guiding attention...

Work in progress mockups showing a version with color coded population segments, instead of the monochrome look that made it into the final report

130

Eli Holder @elibryan.bsky.social · 25/06/2025

TBH this is my first encounter with a "mosaic plot," but the same pairing effect works for showing inequality. Still compelling as a quick gut read, but it takes some studying to read precisely. e.g. still not sure what's happening with 1st+women?! en.wikipedia.org/wiki/Mosaic_...

Mosaic plots from wikipedia of expected titanic survival rates where "yes" (light gray). This shows proportionate outcomes where equal portions of all groups survive.

Mosaic plots from wikipedia of actual titanic survival rates where "yes" (light gray). This shows disproportionate outcomes where equal portions of all groups survive.

000

Eli Holder @elibryan.bsky.social · 25/06/2025

Student growth charts from @damianbetebenner.bsky.social. I like how seeing them as a pair (left is baseline, right is after intervention) gives an immediate gist sense of changed proportions. dbetebenner.github.io/NCSA_2025_Co...

Growth as an Equity Indicator (Baseline Scenario). The mosaic plot depicts percentages of students demonstrating different levels of growth based upon the growth needed by the student.
The widths of the columns represent the current percentage of students in each category (in this figure 50%, 25% and 25% have growth goals of > 1 Year’s Growth, 1 Year’s Growth, and < 1 Year’s Growth, respectively.
The heights of the rectangles represent the percentage of students in each row category demonstrating that amount of growth (uniform distribution of growth by prior achievement).

This shows proportionate results.

Same chart as a "good scenario"

This figure shows a good scenario with regard to equity.
The growth associated with the students needing the equity boost to catch-up shows a preponderance of students demonstration a least a year’s worth of learning.
Decreasing the percentage of students demonstrating less than a year’s worth of learning is key.

120

Eli Holder @elibryan.bsky.social · 20/06/2025

PSA: This button on Google Cloud costs $16 per click. Ask me how I learned...

Screenshot of Google cloud's Vertex AI / "Media Studio" / Veo 2 tool annotated with big red letters saying "$16 per click" and pointing to a submit button. If you follow the page's prompt to "Try this: "Pancakes with maple syrup, cinematic food advertisement b-roll, dynamic."" It will spit out four 8-second clips of oddly rendered pancakes and bill you $16, which is quite a bit more expensive than actual pancakes.

020

Eli Holder @elibryan.bsky.social · 23/05/2025

I didn't want to like it, but recent "vibe coding" models are disturbingly impressive?! And kinda fun for prototyping?! 😬 I'm not new to glsl / three.js / etc, but I was surprised at how much it sped up my learning curve for testing new algorithms / approaches?

to chat gpt: "can you write glsl to show a surface (a rectangle) that's being twisted almost like a dna helix?"

Not quite what I wanted, but I didn't realize it could spit out complete, working GLSL?

to chatgpt: "let's use glsl to draw 10 lines on the screen, starting from different points on the bottom, but converging toward the top center. Their paths wobble a little bit, but they never overlap."

The "non-overlap" solution it gave was surprising and pretty clever.. it said "By giving each line a different phase (here phase = i * 0.7) and choosing AMPLITUDE small compared to the horizontal spacing (1/(NUM_LINES−1)), they never cross."

A bit more involved with Windsurf + Gemini, but this was the fastest I've ever gotten three.js up and running. I asked it to generate a foldy flat oval surface and it gave an algorithm for potato chip geometry using ParametricGeometry. It struggled with cell shading, but to be fair, so did I... Then maybe most impressive was how I could give a super vague description of the composition and it generated something reasonably close... with parameters that could be tweaked more to produce something actually reasonable ("okay let's change the layout again and add more of the petals and have them randomly fall down the screen vertically like someone was above the screen dropping them")

Asked the robot to turn a noisy black and white image into a flow field then trace the paths without intersecting. This took lots of manual tweaking and a bit of prior knowledge, but still much faster than I could've done on my own.

The biggest help was letting me test a "topology" approach that absolutely didn't work, but instead of a day-long rabbit hole, I was able to rule it out in like 20-30 minutes...

030

Eli Holder @elibryan.bsky.social · 05/05/2025

🎆📊 Big win for civic dataviz! NAEP now shows student group scores as full distributions. Even for mass-audience reports, we're not stuck w/ conventional (misleading) bars. Proud to have a tiny role in this (+ @lesvillegs.bsky.social @rsavoie.bsky.social) www.nationsreportcard.gov/reports/math...

Postcard from NAEP with icon of a histogram. It says "2024 nation's report card distribution charts. View how NAEP performance varies within specific student groups. NAEP: National Assessment of Educational Progress"

1111

Eli Holder @elibryan.bsky.social · 28/04/2025

Had a blast last week at @ncme38.bsky.social #NCME25. Felt a bit surreal presenting now that "equity" research is forbidden knowledge, but was great to see the measurement community engaging with this and the overwhelmingly positive feedback was reassuring.

talk summary slide. It says "Improving Score Reporting of Student Group Results to Better Support Educational Equity" and lists the panelists.

Teaser slide from my talk, asking "Does reporting on achievement gaps motivating closing them?" It shows a screenshots of a news report highlighting racial disparities for a local school district, from Quinn 2020 Experimental Effects of “Achievement Gap” News Reporting on Viewers’ Racial Stereotypes, Inequality Explanations, and Inequality Prioritization

Teaser slide from my talk, asking"Do school rankings help families find their way to better schools?" It shows a screenshot of a school ranking website.

Teaser slide from my talk on the guarantees made by conventional dataviz guidelines

030

Eli Holder @elibryan.bsky.social · 27/02/2025

I am excited to announce I have been selected to be a judge for this year's Information is Beautiful Awards - a celebration of global excellence in data visualization! I look forward to reviewing your finest jitter plots. www.informationisbeautifulawards.com/news/640-mee...

IIB awards judge postcard. It says "I'm a Judge! Information is Beautiful Awards 2024"

050

Eli Holder @elibryan.bsky.social · 14/02/2025

Move fast and break EVERYTHING.

Screenshot of an article that says "Can Tech Get Nuclear Power to Move at AI Speeds?"

000

Eli Holder @elibryan.bsky.social · 12/02/2025

A "mostly fine" Civilization would be such a relief at this point...

screenshot of Civ VII review. It says "Civilization VII is mostly fine on Steam Deck, which is great news"

030

Eli Holder @elibryan.bsky.social · 10/02/2025

Most frustrating: Pretending that polls reflect Trump's actions completely ignores how these attitudes are actively manufactured and, conveniently, ignores how they're shaped by the media. Even reporting on polling could shape attitudes that might show up in the next round of polling.

quote from Isaac Chotiner's interview with Senator Brian Schatz: "And one of the things that Republicans do well is not allow their pollsters to tell them everything. They invent whole stories and then the polling follows."

https://www.newyorker.com/news/q-and-a/do-democrats-have-a-plan-brian-schatz

100

Eli Holder @elibryan.bsky.social · 10/02/2025

Vibes-y Qs also invite misinterpretation. e.g. "is he keeping his promise?" is 🤷 without also asking "What promises do you think he made? What actions do you think he's taken?" So CBS' headline isn't just shaky, it invites Rorschach hot takes like "See! Trump's doing what he promised TO ME."

Headline from CBS' article saying 'Trump has positive approval amid "energetic" opening weeks; seen as doing what he promised

100

Eli Holder @elibryan.bsky.social · 10/02/2025

His Raleigh office is nearby, so went in the morning with my daughter. Called the number on the door in the lobby and got an answering machine. Expected I guess? Went later that day with a larger group, maybe 100-200 people? Big crowd for a Raleigh Tuesday afternoon www.mobilize.us/indivisible/...

People protesting outside Sen Tillis Raleigh office last Tuesday

110

Eli Holder @elibryan.bsky.social · 04/02/2025

Wow, big crowd for a random Raleigh weekday?!

Crowd outside Sen Tillis and Budd's Raleigh NC offices, protesting Trump / Musk / Vought / Project 2025 takeover.

000

Eli Holder @elibryan.bsky.social · 31/01/2025

We look at @kstrump.bsky.social and @arielrwhite.bsky.social's clever study showing how depressing inequality charts aren't always as depressing as theory might predict. The chart on the right should lead to resignation and rationalization, but it doesn't?! doi.org/10.1017/XPS....

Image shows two line charts, each labeled "United States: Income Inequality Over Time." Two stimuli charts used in "Does Inequality Beget Inequality? Experimental Tests of the Prediction that Inequality Increases System Justification Motivation.” Both of these charts use the same data. In fact, they're identical except for one key difference:
Flatter Trendline shows the full y-axis (0–100).
Steeper Trendline shows a compressed y-axis (35-50).

110

Eli Holder @elibryan.bsky.social · 07/01/2025

Excited to share my latest with @rladiesrome.bsky.social on Thursday, wherein I'll: – poopoo John Snow's cholera map – trace through implications of equity-oriented public-health charts – offer some practical tips toward more equitable, expressive ways to plot group health outcomes See you then!

Cover slide for my "deathly dataviz" talk. It says "a morbid introduction to equity impacts for visualizing population health"

A zoomed in screenshot of John Snow's map of cholera cases around London's broad street pump

121

Eli Holder @elibryan.bsky.social · 03/01/2025

Think it depends on recipient's stage. If material is "just bad" implies "new to this" then leaning positive and limiting critique might actually motivate more improvement over the long run. Some random studies on feedback balance (via dashboard design, of course 🤦): 3iap.com/dashboard-ps...

blog post quote:

Balancing commitment and contrast matters when choosing between positive and negative feedback. For example:

If you’re teaching a child to play piano, you might lean toward positive feedback. Encouraging and rewarding their successes builds their commitment, whereas focusing on mistakes may cause them to give up before they have a chance to improve (children, dolphins and husbands have this in common).

However, if your student already plays for the philharmonic, negative feedback might be more effective. If they’re already committed and confident in the instrument, highlighting their mistakes surfaces the gap between their current playing and latent virtuosity, helping them improve.

010

Eli Holder @elibryan.bsky.social · 03/01/2025

e.g. one of my EffAff charts before/after Gabby feedback. Thought the first was fun, but I love the second. Wouldn't have gotten there on my own. Of course, neither look as good as her work 😅, but the process improved my sensitivity for "could this be less busy?" www.effaff.com/health-risk-...

Example chart showing experiment effect. It's visually very busy with lots of different font combos, overly chunky elements, etc.

Example chart with slightly more information density, but less overwhelming visually.

110

Eli Holder @elibryan.bsky.social · 03/01/2025

Exception is maybe "punching up" at institutional viz, which acts more like accountability? (eg Phil Meyer re: attacking NYT). But even this isn't clearly effective? Only mild success I've had moving immovable practices started w things like invited (solicited) talks, people finding my work, etc. 🤷

Philip Meyer quote: "The most efficient way to improve the standards of journalism, therefore, is to attack the New York Times.

https://www.aapor.org/wp-content/uploads/2022/11/Public-Opin-Q-1990-MEYER-451-9.pdf

000

Eli Holder @elibryan.bsky.social · 03/01/2025

Tired: Criticizing others' work. Wired: Criticizing your own embarrassing work even harder than the critics. www.reddit.com/r/slatestarc...

Screenshot of a reddit thread criticizing an old chart / blog post.

Them: "It's all just shoddy work"
Me: "Aha! It's actually even shoddier than you think!"

150

Eli Holder @elibryan.bsky.social · 30/12/2024

Portfolio is overdue for an update, so exploring marginalia and sidebars. Still a bit messy, but kinda fun! 3iap.com/work/nightin...

030

Eli Holder @elibryan.bsky.social · 02/12/2024

Polling makes lawmakers more responsive to the electorate, they said. (Original voter ID referendum only passed by 55%.)

Quoted text: "this is a very highly popular measure, and was polled earlier this year and received support of 70% to 29%" Daniel said

000

Eli Holder @elibryan.bsky.social · 24/11/2024

Rorschach chart 📊 in action... Spike in new businesses might be: - "Bidenomics in action" - "Having a boss effing sucks" - "PPP Loan Fraud" - "Uber drivers setting up LLCs for tax purposes" - "Remote work and advances in LLMs" - "Creative destruction" - "Ageism"

A fake chart with a big inkblot over the plot

060

Eli Holder @elibryan.bsky.social · 22/11/2024

This is the funniest 7 minutes you can spend learning about ethical dataviz: www.youtube.com/watch?t=1471... Relatedly: TIL @mcnuttandrew.bsky.social coined "Visualization Villainy" three years before my / Gabby's outlier talk... AND led the way on shitpost ethics talks. But credit where due! 🤦🤦🤦

Diagram from McNutt, Huang, Koenig's "Visualization for Villainy" VIS paper, covering the "design space of harm that visualizations can do"

Harm can be described along two axes, either Non-Physical←→Physical and Indirect←→Direct.

https://arxiv.org/pdf/2109.06007

081

Eli Holder @elibryan.bsky.social · 18/11/2024

Ranks = arbitrary given small differences. So a bit of noise can move one country up / down the list in a way that's disproportionate. Also arbitrary given metric chosen for sorting. #1 is based on % proficiency(?), but why not avg/p50? Newer PISA report captures this dynamic by calling out ties...

Table I.2.1 from PISA 2022 results report, showing that even though you can order countries by scores, but score diffs between neighboring countries might just be noise

https://www.oecd.org/en/publications/pisa-2022-results-volume-i_53f23881-en.html

100

Eli Holder @elibryan.bsky.social · 18/11/2024

WIP updates to a cringey / toxic 📊 from a few years ago. Ideas: - Unit charts! - Avoid stigmatizing w/ red + butterfly - Ranks are aggro/arbitrary, instead contrast US results vs composite benchmark - Category % bars exaggerate differences, try orig. scale for X instead via 3iap.com/us-numeracy

For PIAAC Numeracy scores, the US ranks #28 of 38 countries surveyed in terms of % adults scoring at level 3 or above. e.g. for US, ~37% of respondents had proficient scores (level 3+), vs 63% for Japan and 58% for Finland.

Reproduced from

https://web.archive.org/web/20201024031753/https://www.oecd.org/skills/piaac/publications/countryspecificmaterial/PIAAC_Country_Note_USA.pdf

Each horizontal bar represents proportion of test takers for each PIAAC level (without showing non-responses due to language barriers). These proportions (seem to?) cause bars to shift left or right moreso than you might expect if you were just comparing score means.

Ranking also seems to exaggerate the differences. Average US score is pretty close to the overall average, but 28 out of 38 feels worse?

Work in progress alternative.

Each row shows numeracy score distributions for each country, with labels to summarize % proficiency. # of blue figures correspond to # of "proficient" testers (per 100). Can still see 64% for Japan, 39% for US. But now the height of each person stack shows the proportions for each level.

Notes:

Previous chart + PIAAC website combine different schemes for counting non-responses, plus also different testing periods. This should be more consistent?

Showing something closer to the real distributions. Not exactly, but close. Quantile dot plot might be truer to the data, but would involve some math to estimate the distribution from the published dataset. Unclear if any (perceived) accuracy gain would actually change viewers' gist reads enough to justify the extra effort? So this is a hybrid, where each level bucket is filled up with rows of people in proportion to the published percents for each level (via https://nces.ed.gov/surveys/piaac/ideuspiaac/dataset.aspx). Because the middle four level buckets are evenly spaced, this has a similar visual effect to a distribution.

The "shadow figures" are meant to highlight the delta between the lower rows and Japan on top. This is a convention I like when using smoothed area curves, but with these little figure stacks, not confident this would be read as a separate benchmark distribution, so didn't stick with this.

Main risk with this is that it's still doing hard country-vs-country comparisons to show the opportunity gap for the US (e.g. "here's where we should be") and create a sense of urgency. Since Japan has the best results in the study, this might evoke model minority stereotypes or suggest too much of a culture gap such that viewers might discount this as a fair benchmark ("That might work in Japan, but it won't work in the US!"). Related issue: Creating this sense of urgency is maybe not even necessary in the context of the article...

Landed on this one to replace the previous spiteful red butterfly chart in the blog post. Still may tweak it further.

This is mostly the same as the previous WIP chart, with one big change: Instead of using specific countries as the benchmark, it's using a composite of the countries with the 3 highest average scores (Japan, Finland, Netherlands). The idea here is to frame the US results in terms of a more aspirational results, but avoid specific baggage related to Japan / Finland / Netherlands individually. This is probably still unnecessary for the goal this chart serves in the blog post. And even for the sake of benchmarking, it's probably better to use an internal geography for the benchmark (e.g. CT or VT I think had highest numeracy scores) instead of other countries. This would involve more data pre-processing though, and might make it more "inspiring" but I don't think would impact risks of misperceptions, so safe to punt on that.

Also added better labeling to help orient people, plus average scores for each row since that's probably expected.

180

Eli Holder @elibryan.bsky.social · 18/11/2024

Introduce yourself with four video games

110

Eli Holder @elibryan.bsky.social · 15/11/2024

Absolutely cannot focus this week. Is mindlessly browsing shadertoy one of the five stages of grief? Maybe it should be? There's something reassuring about "Here's an entire galaxy in 41 lines of glsl"... www.shadertoy.com/view/MdXSzS

030

Eli Holder @elibryan.bsky.social · 15/11/2024

lol, yeahh, turning point for me was "how might we render a zillion timepoints, animate them, and make it work on 8 y/o ipads." My transitions never got quite as smooth as y'alls though 🤣

110

Eli Holder @elibryan.bsky.social · 04/11/2024

I always come back to this Philip Meyer speech, where he buckets the value as 1) to "maintain interest" from readers and 2) as calibration for pollsters / political scientists. Neither imply utility for a typical voter though, so "useless" seems pretty safe for most? www.aapor.org/wp-content/u...

Transcript from Philip Meyer's 1990 AAPOR presidential address, highlighting "Why journalists should feel so guilty about their obsessive concern with who is ahead is hard to fathom. The most interesting fact about an election is who wins. The most interesting thing about a campaign at any given moment is who is ahead."

Transcript from Philip Meyer's 1990 AAPOR presidential address, highlighting "Nevertheless, I do not resent all the resources that are devoted to election polls because of two redeeming features. One is that they help maintain interest in an election at a time in our history when maintaining the public's attention is increasingly difficult. Politics has always had a necessary entertainment function. We should not feel guilty about our participation in that aspect."

220

Eli Holder @elibryan.bsky.social · 18/10/2024

The #ieeevis capstone from @kanarinka.bsky.social was 🔥🔥🔥, particularly around the importance of intentionality when visualizing inequality. Theories of change behind popular disparity 📊s are "questionable" (at best!), whereas inspiring data activism uses social data intentionally (strategically).

Slide from Catherine D'Ignazio IEEE VIS Talk, showing a "doubt" about typical visualizations of social outcome disparities. She highlights two implicit, questionable theories of change behind these charts. 1. if you show harms / injustice, someone will magically go and solve it. 2. by quantifying harms, you can make a better case for reparations (unclear if this works though).

Slide from Catherine D'Ignazio IEEE VIS Talk, highlighting 3 lessons we can learn from grassroots data activists. 1. Alternative epistemological frameworks. 2. Embodied and emplaced visual grammars. 3. Strategic political calculations. #3 is highlighted with further open questions to better understand their impacts.

171

Eli Holder @elibryan.bsky.social · 17/10/2024

Health agency implications: 1. Equity reporting can backfire because of how it's presented. :. Urgently need to rethink visual comms strategies for pub health / equity. 2. Equitable designs are healthier designs. They're better for social misbeliefs, and as effective for risk, support, behavior.

A population health bar chart showing disparities in heart disease between social (racial) groups with text: "These charts are bad for our health"

120

Eli Holder @elibryan.bsky.social · 17/10/2024

On the other hand, we designed these charts to remind viewers that health outcomes can depend more on where people live, rather than who they are. Not only does this approach improve misattribution, it's just as effective for viewers' health risk perceptions, behavioral intent, and policy support.

A proposed alternative way to visualize population health disparities, emphasizing geographic variability with annotations and jitter plots.

Results slide comparing conventional bars to the geo emphasis charts, showing that both charts are similarly impactful for increasing risk perception, behavioral intent, and policy support.

Results slide for two (of five) attribution measures, showing where conventional bars and geo-emphasis diverge in impacts. It shows that bar charts significantly increase genetic and personal attributions (blame, stereotyping) relative to participants' baseline responses. On the other hand the geo-emphasized chart is significantly better than conventional bars for minimizing genetic attributions, and significantly better than baselines for shifting attributions toward external explanations instead of personal blame.

130