A Great Way to Think About Data Science: The Bowtie

In any data project about people, we’re starting from a broad, language-based set of conceptions, definitions and questions, figuring out how to turn those ideas into quantifiable numbers that we can engage with statistically, then turning those numbers back into answers or information related to complicated, language-based meaning. We might use numbers to make decisions, but ultimately decisions and knowledge are held in a linguistic realm.

Our perspectives, biases, and even prejudices enter this process at the moments when we attempt to transform an idea into a number and when we attempt to turn it back out into an idea.

Scope and Complexity:

One of the unavoidable ways that subjective perspectives are embedded into the data process is in selecting what kind of number you want to use to represent your concept.

For example, if I want to know about the prevalence of hunger in my country, what number should I collect? Self-reported “hungriness” on a scale from 1 to 10? Rates of malnutrition hospitalizations? A BMI average? How many times terms related to “hunger” appear on social media? The amount of imported and domestically produced food per capita?

Any of these might represent only a single dimension of a complex idea like hunger…