Visualizing data without misleading or stereotyping

Introduction:

In the wake of the presidential elections that revealed a nation that remains highly divided, this bilingual post explores how we might visualize data in a way that doesn’t mislead audiences or stereotype different types of people.

今月初めに行われた大統領選挙では、米国でまだまだ分断が続いていることが明らかになりました。この投稿は、誤解を招かず、ステレオタイプを強化しない形でいかにデータを可視化できるかについて模索しています。英語の本文の後に日本語が続きます。

Earlier this month, we witnessed one of the most dramatic presidential elections in history. These past four years, the U.S. had been unrecognizable to me. As a woman, quasi-immigrant, and minority, I felt that I was unwelcome on all three counts. I was in a state of disbelief as racist remarks and actions were normalized, and many laws that I had been proud to associate with the U.S. were rolled back one by one.

It is a huge relief to have a national leader who seems rational, calm, and mindful of the growing diversity of the U.S. demographic. I am ecstatic that we now have the first woman vice president–who also happens to be Black, Asian, and the daughter of an immigrant. That fact alone allays my concerns about criticism over the president-elect’s treatment of women.

A Divided Country

But as we all know, this was no swift victory. The “blue wave” touted by pundits never came. Instead, we had a handful of swing states that flipped, one by one, from slightly red to barely blue over the course of four days. I kept taking screenshots of the close race (at one point a difference of 1,000 votes, or less than 0.1%!) and sending it to friends. I pored over the news analyzing the developments in each state, from which counties’ votes were being counted first, to why Nevada seemed to take its sweet time, to legendary figures like Stacey Abrams and the late John McCain affecting the outcome in Georgia and Arizona.

The state of the swing states as of the morning of Nov. 6. With 99% of votes reported, Georgia had a difference of 1,000 votes, or less than 0.1%. Screenshot from Google based on results compiled by the Associated Press.

Maps

With all the election results readily available online, it has been really fascinating to be able to zoom into any state and look at the results in each county. Maps like this one (for Virginia) show islands of blue cities in a sea of red.

Election results by county in Virginia. Screenshot from Google based on results compiled by the AP.

But this year, the way maps show election data seemed to undergo an important and fascinating shift. With the slogan “Land doesn’t vote; people do,” several maps came out to show votes in proportional circles based on how people voted in each county, as opposed to coloring in the entire area of each county. Since fewer people live in rural areas, this was a much more accurate representation. Based on how much recognition these newer maps received, I suspect future elections will be represented in this way.

Either way, the fact remains that we are a deeply divided country, mostly reflecting whether we live in urban or rural areas. So how do we heal as a nation? One way, I think, is to avoid stereotyping others as much as we can.

Visualizations that Reinforce Stereotypes

The below illustration is called “What it means to be a typical Democrat or Republican, based on everyday items.” A translation of the words that appear throughout the illustration are in the chart below (all translations are my own).

From this website of The Asahi Shimbun. I added the numbers in purple for the translations provided below.
DemocratsRepublicans
Prius; VolvoCarsHummer; Porsche
MSNBCTV stationsFOX
Comedy; RomanceMoviesWar; Action
Jazz; RapMusicCountry
Tennis; SoccerSportsRodeo; Motor Races
Women: Silky smooth; Men: Long with beardsHairstyleWomen: Voluminous; Men: Short and neat
CasualAttireBusiness Suits
Sushi; VegetarianFoodFried Chicken; BBQ
StarbucksBeveragesCoors Beer

This was apparently first published in The Asahi Shimbun about 10 years ago. It came up on its website this past March (with the explanation that “trends have not changed that much since then”) as part of an article that helps young job applicants / recent college graduates understand current events.

When I first saw this, I couldn’t help but laugh. It’s wonderful that Japanese audiences are paying close attention to the U.S. elections. I think visuals are very important, especially to a younger audience. But I also think we need to be careful not to generalize too much–precisely because we are shaping young minds.

To start with the basics, the data comes from mixed sources. This illustration is apparently based on “data from advertising and research firms, as well as the voices of American voters.” That’s at least three sources that probably use different methodologies, samples, dates, and collection methods. While I don’t expect the entire methodology to be part of the picture, I’d at least like to know the names of the companies that collected this data.

The illustration is full of points I want to ask more about. For example (and I am also making big generalizations here), the “Republican” category seems to combine several types of people: the military type (short, neat hair), the wealthy type (Porsches), people living in rural areas (fried chicken; country music), etc. More minor examples show weird combinations too, like Starbucks (likely coffee) with sushi for Democrats. It is very confusing because all these mixed data is illustrated in the same picture.

And while the variety in music tastes and hairstyle is certainly interesting, I don’t see how it makes a big difference. The only thing I thought was truly relevant here is the type of media consumed (FOX vs. MSNBC), which other sources also indicate. I would rather know about the difference in opinion on topics like education, immigration, and religion. And, at least in terms of food, there’s evidence that we can’t associate them with political thought: The New York Times recently published a quiz asking readers to look at photos of fridge contents and guess whether they belong to a Trump supporter or Biden supporter. As of today, readers have made 25 million (!) guesses, and were correct 52% of the time–it’s 50/50, even with that huge sample.

Caricatures

To me, the most egregious point that could be corrected is that all four people depicted here are white. According to data compiled by the Pew Research Center, as of 2019, 40% of registered Democrats were non-white (even back in 2010, when this illustration was made, it probably would have been more than 36% (2008)). The Democratic party clearly states that “diversity is a strength,” and its support for immigrants and minorities is clear. So it seems especially odd to represent 100% of the Democrats here as white people.

But I also see how it’s extremely hard to visualize people “correctly.” Take, for example, the controversial NHK video that attempted to explain the BLM movement in June. If Black people or other minorities were added to the Asahi Shimbun visualization of Democrats and Republicans, would it have made things better? I doubt it, because it’s hard to illustrate someone without resorting to caricatures, especially if you do not know them well.

A screenshot (from here) of the controversial NHK video that illustrated Black people who wore tank tops and Afros, lighting the city on fire, and saying that they were resorting to violence because they were angry about the income gap (with no mention of police brutality)

And the truth of the matter is that there’s an inherent difference in illustrating someone who is already in the majority versus someone who is not. The former has already been drawn in many different ways, and one additional illustration is just that–a collection to add to many different images that readers may have in their head. It won’t skew the audience’s minds in either direction. Someone who is rarely drawn, on the other hand, automatically becomes a representative of their entire group because they are rarely seen. It’s similar to how movie characters used to be caricatures. The token Asian characters in older films were stereotypes (Mr. Yunioshi in Breakfast at Tiffany’s or Long Duk Dong in Sixteen Candles), whereas now we are seeing diverse backgrounds and personalities (from Crazy Rich Asians to The Farewell) because there are more films and more characters.

Visualizing Content in Better Ways

A lot of the data in the Asahi Shimbun illustration is interesting, even if not necessarily relevant. I think this could be improved by 1) listing its sources by name, 2) not showing people in the illustrations, and 3) instead of showing the top one or two in the same big picture, perhaps choosing the top five in each category and turning them into separate charts. Illustrations are so powerful, helping us understand and remember things better–but without the full context, they can also be misleading.

Media have to work with quick deadlines, and it’s easy for me to be an armchair critic. But as people pointed out with the BLM video, I believe there are ways to find consultants. On a deeper level, I believe we all need to have a better understanding of each other, so that we don’t stereotype others, and know when we are about to create caricatures.

These take long conversations, better education, more reading, stronger media representation, and so much more. But to circle back to the original discussion, at least we know that we are politically divided. At least we are beginning to learn, in the past six months, how much pain Black people have been experiencing. I hope that we can strive to understand each other. After this election, the only direction to go is onwards and upwards.

ステレオタイプを強化しない形で情報を可視化するには

今回の大統領選挙では、大好きな米国がようやく少し戻ってきた気がします。2016年の選挙以降、移民、日系人、そして女性として、ずっと緊張や不安を抱えてきました。国のリーダーがアジア人に差別的な言葉を使ったり、移民に対する大統領命令を発したり、女性蔑視の発言をするたび、心身ともに疲弊し、いつも少し怯えながら過ごす日々でした。今回、バイデンが大統領となって心から安堵しましたし、初の女性・黒人・アジア系の副大統領が誕生したことを、本当に誇りに思います。

ただ、選挙の結果を見て、国の分断がまだこんなにもひどいことに驚いたのも事実です。スイングステート(激戦州)では、最終的に民主党が勝ったところが多いものの、一時期は数千票、0.1%以下の僅差だったりもして、結果が分かるまでの数日間は本当にやきもきしました。地図を見ても、驚くくらい、都市部と田舎とで政党が真っ二つに分かれています。これからこの分断をどう乗り越えていくかが大きな課題となります。

ステレオタイプを強化する恐れのあるイラスト

そんな時にたまたま、上記の「身近な品々に見る『民主らしさ』『共和っぽさ』」というイラストを友人が送ってくれました。もとは10年前に朝日新聞に掲載されたのが、最近になって「今でも傾向は変わらない」と言う解説とともに浮上したようです。

これを最初に見た時、ツッコミどころが多くて笑ってしまいました。興味深い視点ですし、分かりやすく可視化している姿勢が素晴らしいと思います。でも、可視化するからこその危険性も多分にあると思います。

まず、情報源は「広告会社と調査会社のデータおよび米有権者の声」とありますが、そうすると、少なくとも3つの情報源から得たデータとなり、それぞれ異なるサンプル、日程、調査方法であると想定されます。それをすべて一つの絵にまとめていること自体少し不思議だと思いますが、そうであれば、少なくとも広告会社や調査会社の名前を掲載した方がよいかと思います。

細かい点を見ますと、たとえば「共和党支持層」は、いろいろなタイプの人たちが混じっているように見えます。それこそステレオタイプに基づいて例を挙げると、軍人(「整えた短髪」)、富裕層(「ポルシェ」)、田舎に住む人(「フライドチキン」「カントリー」)がすべて一緒になっています。民主党の方でも、スターバックス(のおそらくコーヒー)を寿司と飲む、という不思議な構図になっています。これも、いろいろな情報源から集めたデータが同じ絵にあるから違和感があるのかと思います。

共和党・民主党支持者が視聴するメディア(「TV局」)はとても重要であり、FOXとMSNBCが両極端にあることは他でも立証されていますが、映画や音楽など、それ以外の点に関しては、それほど重要だとも思えません。むしろ、教育や移民政策、宗教等に関する考えを取り上げた方が興味深い気がします。さらには、少なくとも食べものに関しては、政党との関連性が低いことが分かっています。最近ニューヨーク・タイムズ紙は、冷蔵庫の中身の写真を見て、トランプ支持者かバイデン支持者のものかを読者が当ててみるというクイズを発行しました。現時点で読者は2500万回(!)推測してきましたが、正解率は52%。それだけ巨大なサンプルでも、まだ五分五分なのです。

人物のステレオタイプ

私が最も残念だと思うのは、この絵に描かれている人が4人とも白人だということです。2019年の時点で、登録している民主党支持者のうち、40%が非白人でした(このイラストが描かれた2010年でも、36%(2008年)以上だったと思われます)。また、民主党は「多様性は強みである」と明言しており、移民やマイノリティを支持していることも明らかです。したがって、民主党支持者の100%が白人として描かれているのは残念なことだと思います。

同時に、「正しい」形で人を可視化するのは大変難しいことです。6月にBLM運動を動画にし、物議を醸したNHKのビデオがよい例だと思います。黒人や他のマイノリティをこの朝日新聞の民主党・共和党のイラストに入れたところで、状況は改善しなかったかもしれません。あまりなじみがない人たちを可視化しようとすると、ステレオタイプに基づいた滑稽な絵になってしまいがちです。

多数派の人とそうでない人を描くことには、本質的な違いがあります。前者は既にいろいろな場で、様々な形で描かれており、一枚の新しいイラストは、読者の頭にあるイメージのコレクションに足されるだけであり、これまでの印象を大きく変えるわけではありません。一方、めったに描かれない人は、その人が所属するグループ全員を代表するような形になってしまいます。これは、映画の登場人物にも言えることだと思います。古いアメリカ映画のアジア系の登場人物は、ひどいステレオタイプに基づいていましたが(『ティファニーで朝食を』のユニオシ氏、『すてきな片思いの』ロンなど)、最近はアジアを中心とした映画や登場人物が増えてきているおかげで、多様な背景や個性が描かれています(『クレイジー・リッチ』や『フェアウェル』など)。

より良い形での可視化

朝日新聞のイラストには、興味深いデータが満載です。もし改善するとしたら、1)情報源の会社名を明記し、2)人物は描かず、3)上位1~2位だけを同じ絵の一部として描くのではなく、たとえば各カテゴリーのトップ5などを別々の表にして出す、といったことができると思います。イラストは、物事を理解し記憶する上で素晴らしいツールとなりますが、全体像が見えないと、誤解を招くことにもなりかねないと思います。

メディアは締め切りに向けて急いで作業を行わなければなりませんし、私がこうして後から批判するのは簡単なことです。でも、BLMのビデオに関して他の人も指摘したように、コンサルタントなど、何かしら事情に詳しい人に話を聞いて確かめることはできたのではないかと思います。より長期的な話で言えば、こういったステレオタイプを行わないように、私たちそれぞれがお互いへの理解を深める努力をすべきなのかと思います。

そのためには、対話を続け、教育を改善し、より多くの本を読み、映画・テレビ・本等における登場人物をより多様にしたりと、様々な課題があります。しかし、私たちは少なくとも、政治的な分断が続いているという事実、黒人の人たちが今も苦しんでいるという事実などを学びました。今回の選挙を受け、私たち皆で一緒に前に進み、相互理解を深めていけることを願っています。

Leave a Reply

Your email address will not be published. Required fields are marked *