Post

One question I always ask of a choropleth map is whether it is any more helpful than a table. Is there anything inherently spatial about the pattern? And do people know what's where? Maybe enough people have a sense of which states are where. Probably less so for counties within a state.

It is a strategic argument to show each map, and strategic to call each "misleading." I show these 4 images in class. Same visual "arguments," but people flip their opinion on whether it's an ok argument based on political leaning (top is election, bottom is women's suffrage from 1915).

Show any untrained audience a map of the US with nothing but state boundaries. No colors, no additional information. Then ask: Can you tell from this map alone which state has the largest population? How about the highest GDP? Or the highest average age, or the most expensive housing?

Dissociating the size of geographical regions from every other linked quantitative variable associated with them, except area itself, might be a surprisingly simple way to prevent a choropleth from being mistaken for a mental exercise in comparing colored areas.

I think this depends on dialect and audience To me there's a definition of misleadingness (or maybe correctness) where if there are perceptual operations on a chart your audience will tend to perform and those operations do not correspond to valid data operations, arguably the chart is incorrect

This relates to some topics in my next book. When we design a (communicative) chart, we imbue it with intentionality—fancy word for "aboutness",—we stipulate what it means, so to speak. But on the other side of the cycle there's interpretation. A mismatch between the two is what causes confusion 1/2

A separate discussion is whether certain representational structures persistently lead people to certain (mis)interpretations. Scatter plots are a good case: even if you keep reminding yourself "don't infer causation"... 2/3

But, (and yeah, it's funny), I can sympathize with Alberto's opinion these days. I now think "interpretation" must exist somewhere in the theory of what makes a vis good. Our 2014 paper neglect of that notion reflects how little we know about how to make a good "perceptual interpretation math"

Yeah, 100% - @sheng-long.bsky.social brought that in as "task" in some of our papers (though personally I hate vis tasks and think we should abandon them, but my 2 page rant about that was not accepted as a "visions" talk for vis, lol)

It’s true that election county vote maps that are choropleth are accurate based on the information they are sharing for voting results, but it’s equally true that these maps ARE cognitively misleading when viewing them and/or useful for false narratives etc etc. The Modifiable Areal Unit Problem …

With election maps is a long time interesting cartographic problem that during election times in the best decade, has brought out really great and interesting election maps across journalism (should out to soooo many cartographers across many news sources!)

Fair point, and "misleading" was a shortcut. The map answers "who won where" perfectly well. My issue is that it's mostly used to answer "who's winning the House", and for that people read the size of the area as importance. Hence a companion view, not a replacement: the map morphs between the two.

Yes, you're correct. I love what you did, don't get me wrong. It's just that it worries me that people may be inclined to dismiss choropleth maps thinking that they aren't helpful

I lean toward greater designer responsibility rather than reader. Choropleth maps are designed for areal unit sizes approx constant wrt thematic attribute. Where that relation breaks down, we get noise. Where geographic size systematically relates to interest (urban-rural) we get problematic bias.

Related analogy: Mercator projected map of world countries could be argued is "a map of which country is where, nothing else", but we know the systematic bias in area-attribute relation is a problem. Most accept a better alternative is to reproject the map rather than require reader to undo bias.