By Joseph Wills
If you think back to most any class you’ve had that discussed data, you probably consumed it through your eyes. Unsurprisingly, the overwhelming majority of work done in data science is communicated in a visual manner: numerical and categorical data represented in two or three-dimensional space with shapes or colors. While this is obviously a time-tested and valid method of communicating data, the untrained eye often loses a sense of what it actually means; lines on a graph can be hard for a “layperson” to understand.
A novel approach to data communication was recently tested by researchers at Hong Kong University. They held a “Data Tasting Workshop” where they assessed the validity of “data edibilization” as a means to convey data. They used two methodologies to communicate data: first, they encoded data into food that retains its original appearance, changing its composition to reflect data. To illustrate the employment in agriculture of four different East Asian countries, the researchers placed a corresponding amount of sauce (native to each country) on a cracker. The amount of Kimchi dip, wasabi, curry spread, and broad-bean sauce on any single cracker corresponded to the amount of employment in agriculture in Korea, Japan, India, and China, respectively.
Ideally, through tasting the obvious differences in the number of sauce participants could easily discern the relative amount of employment in each country. Another example of this encoding method is a small salad encoded with Bureau of Labor Statistics data on available STEM positions and STEM degree earners. The number of breadcrumbs corresponded to the number of open STEM positions annually, while ingredients like ham, corn, and tomatoes represent the demographics of each set of degree earners, such as associates, bachelors, or masters. The participants would hopefully take away an understanding of the ratio of annual degree earners at each level, as well as their potential job prospects.
The second method of data edibilization was by representing data as a graph constructed from food: a cheese platter was arranged in a manner reminiscent of a pie chart and described data from a poll of Asian Americans’ attitudes towards their racial identity. White cheese cubes were those who identify as typical Americans, orange cubes represented Asian Americans who identified as coming from their country of origin, and a mix of the cubes represented a bicultural identity.
The immediate question that’s raised from this study is its effect on a person’s ability to internalize complicated data. The research suggests that “consuming data” benefits from a person’s natural attraction to food; many of the study’s participants noted that the food was particularly attention-grabbing because it was close to lunchtime. Participants rated edibilized data as more visually appealing and more fun. However, there are implicit flaws in gestating data: the 80% of participants who considered food easier to remember also considered it harder to understand.
The inherent flaw in edibilizing data is that the finer details are lost. A person may leave a “Data Tasting Workshop” with a better broad strokes interpretation of the how Asian Americans identify with Americans, but it can also cause misunderstandings -- one participant thought the mixed cheese cubes suggested Asian Americans who identified as purely American and those who were bicultural could integrate better than those who identified solely as from their country of origin. A misinterpretation While this may be true, it also suggests the fatal flaw of edibilization: for a person to represent data with food, they must interpret it to some extent.
Data edibilization is a novel method of communicating data, and while it could benefit from further research, it will necessarily fall victim to its creator’s interpretation. For a person to represent data with food, they must interpret it to some extent. It’s too easy for the person making the food to imbue within it their own biases -- on some level, what they think of the data will always be represented in the food. Negative experience as a matter of course affects people’s perceptions. Data edibilization could therefore be manipulated to manipulate what people believe. For example, a cookie with an amount of salt corresponding to population density could imprint on its eater’s mind an aversion to cities.
The first chapter of practically every statistics textbook explains how raw data -- numbers and facts -- can be manipulated to say almost anything. So much of the detail of data is lost that whatever discretion its audience had in their own interpretation is lost. Edibilized data is almost axiomatically pre-interpreted data; not encoded with the scientific method and an ability to judge the facts for oneself, but with the inherent biases that arise when something as subjective as food is involved. Unless it’s approached in a completely new way, this earnest -- if misguided -- method will rightfully be relegated to statistics’ most extreme periphery.
References:
Wang, Y., Ma, X., Luo, Q., & Qu, H. (2016). Data Edibilization: Representing Data with Food. Proceedings of the 2016 CHI Conference Extended Abstracts on Human Factors in Computing Systems - CHI EA 16. doi:10.1145/2851581.2892570