Sapir-Whorf for Data Visualization: The Tool Shapes my Reality

Ask any interpreter if language shapes a person’s reality, and you’ll have dropped a nice grenade in your day. Short answer: it’s complicated. Sapir-Whorf argues that language shapes our view of reality. The Inuit have a number of words for snow. Does that affect how they perceive snow? Depends on what side of the fence you want to take with your local linguist (or anthropologist if you’re up for some fun). Others say no – we all experience the same thing, but describe it differently.

I’ve determined the data visualization equivalent is how users of particular software packages relate to the greater “data viz” community, or, better put:

As someone from one (widely siloed) practice profession – interpreting – this speaks to me. You see, there are a number of realities for ASL interpreters that differ from spoken language. In short (United States focused lens):

1. ASL interpreters use 2 different modes of communication – 1 spoken, 1 signed. This means it’s easier in some ways to interpret simultaneously, but harder in others (GRAMMAR). Side note: sign languages, like spoken languages, have different grammar, histories, and practices.
2. ASL interpretation is governed by the ADA (Americans with Disabilities Act) in the US. Spoken interpreters go to Title VI (non-discrimination) and fall under the Office of Civil Rights. Terminology in the laws creates a very different reality for both the people receiving services and those providing said services. (Long lectures on specifics available over coffee and lunch. You’re buying.)
3. Training and certification for ASL interpreters differs from spoken language interpreters. Requirements in work places are also different.
4. Stigmas around clients are different. I’m not elaborating.

So, what do interpreters share?

• The responsibly of converting messages. This sounds as technical as it is. Interpreters do things with short term/working memory few normal human beings do. If you’re working with an interpreter, they’re doing deep dive analysis in seconds. Give them a medal, or at least some kudos.
• Cultural mediation. Sometimes (okay, often), things we do in 1 culture is REALLY offensive in another. And, we don’t mean it to be. Or, what we say sends warning bells in 1 culture that don’t exist in another. In English, when we get angry, we may threaten to choke someone – we don’t mean it (usually). Interpreters help smooth all this.
• Misery in joke telling. Jokes depend on 2 things: shared values and shared terms. Neither translates. The truth: more interpreters may tell an alternate joke with similar timing or cue the person to laugh. Sorry – bubble bursted – you’re not funny in other languages. But, I am. 🙂 (Bad interpreter joke)

So how does this relate to “data viz” as a profession?

Emerging Practice Profession

There, I said it. It’s still emerging from other professions, such as statistician, actuary, report writing from other specialized professions, and a long list of other things. What we do comes from a bit of everywhere and it shows when you ask an analyst/data vizzer/dashboard maker/Tableau consultant/__(fill in the blank)__ what they did before. The fact we can’t pick a name shows it.

I ran into this with interpreting. I did a paper on it even: interpreter vs translator. Even today, people say I did translating, but the term (in my world “professional bubble”) ties strictly to written conversion. I didn’t do it, kids. All my work happened live. But, most people felt a translator was more accurate (they’re not – it’s all method of delivery) while an interpreter took more liberties with the message (we don’t).

Emerging professions are messy. They come to similar patches of grass from all over. There’s no one shared history, so agreeing on “rules” means pulling at the blanket. Pull too hard, and the whole thing rips and we all end up cold. So, we all have to figure out how to settle under the blanket. Anyone who has ever tried this literally knows it’s not easy.

Practice professions are also hard to nail down as each person’s experience IS a factor to what comes out. Look at some of the other practice professions: doctors, lawyers, dentists, and so forth. More than just training, what these people have experienced IN and OUT of work comes into play. A doctor who suffers from the same skin condition may offer more practical advice on how to manage it than a doctor who has only ever treated it. As someone bilingual in a visual language with a background in interpreting, I analyze data differently than my counterparts with a computer science or mathematical background.

Names matter, but so do histories, values, and shared practices.

The Methods of Work

Sketching data by hand versus using software creates a different process. On one hand, freehand sketching on paper lets me draw as I like. But then, that’s it. There is no deeper interaction. Software lets me do what that application was designed to do. It was made for a reason – to show data – but within certain limits. Excel lets me make charts from wizards. I need to gather it a certain way, select the chart, and spend some time formatting it. SPSS lets me do certain types of statistically focused charts and tests. Microstrategy lets me hook to certain types of data and use a canvas to display charts together. Tableau uses a particular logic around small multiples so I can combine charts to filter. D3 lets me code all types of stuff.

All of these came from different needs. My limited focus is on Tableau. It came from a need to see the data in a way that allowed patterns to be seen easily. Small multiples are great for this. Presto, Tableau (or Polaris, if you wanna go old school).

Here’s where I quote Sapir-Whorf: The limits of the tool shapes my reality.

Tableau does great with transactional data. It lets me slice and dice to my heart’s content. I don’t have to code, I don’t have to plan, I don’t have to assume or know anything about my data. I can use Tableau to figure it out.

Tableau gives me a number of charts through Show Me and I can make a lot of others through knowing where to put things, writing some calcs, or – if I want to get fancy – shaping the data a certain way. I have certain types of interactions and “animating” I can do. But, there’s limits. Sankeys can be done, but they’re hard to maintain on live, changing data. I can do certain types of interactions, but achieving some of the “draw your own” New York Times style graphic means using something else.

The limits of the tool shapes my reality.

The Output

The output of these tools all differs. In Tableau, I have charts (worksheets), dashboards, and storypoints if I want to navigate someone through it. Excel gives me worksheets where I can put data, calculations, charts, pivot tables, and whatever else I add in with plugins. Logic tells me to organize both of these things that a particular way, but frankly, it’s my choice what I do. D3 lets me make all kinds of animations, but it’s rare to see a number of charts forcing interactions with other charts. Usually, these seem to be 1 chart with a lot of interaction and maybe some tabs or filters.

This output decides our best practices. While we may agree on some things (reducing ink and ways to communicate certain types of data), we also fight about:

• Exact chart choices
• Animation
• Chart/information density
• Data format
• User experience
• User navigation

These things rely on the methods of work and output. I can do certain types of charts in Tableau and others are a right pain. Maybe those aren’t best practice on real time data in Tableau. I can do a certain amount of animation, but I run into limits. I could go into a rant about browsers, Flash/JavaScript plugins, and life span, but I won’t. I’m sure someone already has.

Interpreters to have these fights too. Spanish interpreters tend to follow an interpretation more literally, while Arabic interpreters are having to reassemble the whole message in order to get it to fit. How do you define accurate? Spanish and Arabic interpreters face a lot more regional variations as these languages are spoken so many places. Is it sanos for sinuses? For some, yes, others will tell you that you’re way off the mark. Attaque de cerebral or embolia? (Warning – my Spanish is terrible and I’m pretty limited to medical terms – don’t judge.) Intervening caused the same types of discussions – ultimately, how much was too much?

Our tool decides a lot about our output.

So how do we come together?

If I use interpreting as a model, I’d propose:

1. Highlighting the shared threads in the narrative. What are things we all experience? What can we learn from?
2. Finding common ground. Our terms, rules, limits, and outputs will all be different. Can we borrow inspiration from other tools? Can we play safely even when it’s outside the norm?
3. Moving towards the grey. Realizing absolutes will be more limited as more opinions and experiences come into the fold. The realities faced by one group may not be wholly shared in another.
4. Continuing to ask questions and discuss it. Many of us may not relate to “data viz”. I’ve spent years trying to figure out what it is I do (dashboard maker, analyst, Tableau consultant???) and figuring out where I fit in. I’m thinking by the food – anyone know where to find the tea?