Week 29: Is a ravioli a dumpling?
revisiting representation and organization. anti-chihuahua sentiments are expressed
I was going to write a bit more about Rare Book Librarianship to introduce you all to that class, but my grandma emailed me after issue 28 and said that it was too theoretical. I think I can do a better job explaining the purpose of representation and organization, so I’m going to revisit that here, and the next issue will be about Rare Book Librarianship and probably a bunch of other things, because that’s how this works.
Representation and Organization = Examples and Relationships
Consider representation as the examples of categories. Chihuahuas are often considered awful examples of dogs because they’re so far removed from the characteristics and attitudes of what we think of when we think of a typical dog. A Golden Retriever or a German Shepard, on the other hand, might be considered excellent representations of dogs: they have the shapely ears, the energetic and loving dispositions, the fur, the body shape, and so on.
Consider organization as the relationships between and within categories. How might we organize dog breeds in relation to one another, such as in an infographic? By their body features? Their attributes? Their disposition? How and when and why they were bred?
This question-asking (we’ll never escape it!) is the primary undertaking of Representation and Organization: How do we define a category? And once we do, what goes in it?
Let’s continue with the animal example. We understand vertebrates as a subphylum, or a category within Chordates (vertebrates + sea squirts + lancelets), which is itself a category within the Animal Kingdom: mammals are mammals because they give live birth and breastfeed…except the platypus, which lays eggs. Birds are birds because they lay eggs, are feathered, have beaks, and fly…except penguins, which can’t fly.
When we make categories, we have examples that fit into those categories that are most essential: “essential” meaning here that they embody the essence of that category, and that those examples are necessary to justify the category’s existence. One might say that penguins are not very good examples of birds because they lack the characteristics, or essential criteria, we use to judge whether an animal is a bird. There’s a lot of cognitive and information research done on the firmness of these categories, and researchers ask people to rate the goodness (not morally, but essentially) of different examples in different categories. The results are pretty funny, especially when people universally agree, for example, that chihuahuas are terrible examples of dogs.
Here’s another example:
What is a sandwich?
To generalize, I’ll define a sandwich as two edible outer layers filled with some sort of inner layer—spreads, meats, vegetables, other foodstuffs. The outer layers are a vehicle for the inner layers. Webster’s says the outer layers are bread. Google’s definition adds that a sandwich is eaten as a “light meal.” In these definitions, we encounter criteria of structure, content, type, and even funtion.
But the graph above contends different structure vs content combinations that comprise a sandwich. Could those outer layers be sandwich bread, tortilla, potato buns, a waffle cone shell? And could the inner layers stretch beyond meat and PB&J to ice cream or french fries? Can one side of the outer layer be closed, like a hot dog in a bun? Can both sides be closed, like a burrito or a calzone? We understand open-faced pastrami and corned beef sandwiches, which pile the meat on top of the bread and leave it uncovered, as a sandwich. Is a pizza, then, a sandwich?
The lower right-hand box sends me for a spin: a poptart as a sandwich? Perhaps not. But a poptart as…a ravioli? A ravioli is an edible closed container around some sort of filling: meat, cheese, sauce, and so on. A poptart, too, is an edible closed container around some sort of filling: what I imagine is extracted nuclear waste coated in sugar and jelly flavoring and then pumped into a crumbly pastry cobbled together from sawdust. To that end, is a ravioli actually a dumpling, which shares the same structure and fundamental parts?
You may be reeling from reading this and be thinking to yourself, “Of course a ravioli is not a dumpling.” But why? Both are edible closed containers with some sort of filling that are then boiled or steamed to cook the shells. You could argue that they’re different because they’re made differently: ravioli are cut from one large sheet of dough, whereas dumplings are individually hand-rolled. Or you could argue that ravioli and dumplings taste and feel different, but then what happens when a chef flouts those criteria and stuffs a dumpling with cheese (is that a blintz?), or a ravioli with vegetables?
Or are ravioli and dumplings different in our heads because we associate them with different regions, languages, and cuisines? Different dining environments? How have our life experiences influenced our inclination to identify things as similar or separate them as different?
In writing this, it strikes me as somewhat strange that we primarily categorize dishes based on where they are from, rather than what they are made of. To paraphrase the comedian Sofia Niño de Rivera, who has a side-splittingly funny bit about how ridiculous it is to explain Mexican cuisine to a foreigner: every region has their rice and bean dish. So how are we distinguishing between them all?
Fuzzy Theory & Final Thoughts
We could run this exercise with any category of anything: food, animals, plants, office furniture, kitchen supplies. Anything that gets classified undergoes this question process of why it is this and not that. Ultimately, these categories are fuzzy, and information and cognitive scientists know that, which is why they named it “fuzzy theory.” The boundaries of different categories become blurred depending on the criteria. When we look at established criteria, we can see why a robin is an essential example of birds, whereas a penguin is lacking. But if we were to shift the criteria away from
beaks,
lays eggs,
feathers, and
flies
to criteria that favors the penguin, suddenly the robin is not such an essential example after all. Notably, most things can be or are actually categorized as many things, because most things have more than one relationship. Penguins may be poor examples of birds, but they are excellent examples of Antarctic1 creatures, swimmers, and animals that mate for life.
At this point, you may be asking why we even bother organizing and representing information at all, if even the best attempts at creating such categories will always be biased. Put plainly, we begin classifying things basically from birth, and we never stop, so we might as well examine what comprises those categories, and why, and what we consider to be excellent—or essential—representations of those categories. Doing so is very useful to understand how we make meaning in the world. It also avails us of how someone else might do so differently.
Information professionals have the monstrous task of expressing those internal classifications in intuitive user interfaces. If cataloguers and data scientists fail to represent information the way they think most people would, they will obscure that information to those who attempt to seek it.2
A final illustration: In class, we took a bunch of our professor’s junk (her words, not mine), and attempted to relate the things to one another with hierarchical categories. My group had, among other things, a penny, a bottle of perfume, a comics coloring book, a pencil, a paintbrush, a guitar pick, a DVD of Texas Chainsaw Massacre 2 (extra gore edition), a rose made out of resin, a plastic Simba from The Lion King, a picture of five stoic women taken at the turn of the 20th century (henceforth referred to as “the Baddies”), a mountain on a postcard.
Our collection of stuff was quite similar to those of other groups, who came up with deeply creative categories, including “objects, the purpose of which is revealed through creation.” It was a clever way to organize disparate art-making materials.
My group came up with “Lions.”
Both categories are useful. The former, high-level philosophical designation is a “descriptive” category, one that identifies the function of objects. That kind of category is crucial for the back-end of information organizing, where developers teach computers the meaning of different kinds of information.
The latter, low-level, “what you see is what you get” designation is a “presentational” category, one that identifies what objects look like. This kind of category is crucial for the front-end of information organizing, where users are far more likely to input simple key-words like “lions” to find what they’re looking for.
Representation and organization involves figuring out how people imagine and relate things [books, stories, animals, history, sports, etc.] to one another and trying to reflect that in a system so that people can then find things. And it only gets more complex the more variables you add, which is where choice comes into play. That’s another issue. For now, when you encounter categories, ask yourself, who is deciding what makes a sandwich a sandwich, and what forces shape the criteria they pick?
Housekeeping and birdseeking
house
I mentioned in Issue 27 how we often do not own e-books, and now The New Yorker has published an article about it.
bird
Not a bird, but this imbued me with the energy of one thousand suns:
More later.
My grandma emailed me after reading this and pointed out my error in referring to penguins as “Arctic” creatures. They are Antarctic. The accurate category would have been “polar creatures.”
Determining how to represent and organize information only becomes more complicated the more variables you have. Even with dogs, sandwiches, and ravioli, we stumbled into a number of ways our categories are flawed. What happens when you have an item that appeals to four different fields of study for four different reasons: the content, the format, the composition, the production process. How might an English professor versus a professional bookbinder search for the same book? Which features do we have to highlight in order to ensure that both people can find what they’re looking for? And at which point do we exhaust those lists, because we have limited time and manpower, and accept that we cannot sort things into every category all of the time?
Would a burger be equivalent to a chip butty using this diagram? Also, very easily accessible explanations!
Brilliant and funny take on the subject. Much more enjoyable and thoughtful than my recent reading about how to evaluate the "accuracy" (and thus the legal admissibility as evidence ) of "judgments" made by "machine learning" trained AI programs. The epistemological question is the same - what makes a sandwich a sandwich? (Or an arrestee a flight risk who should be denied bond pending trial?) Are machines as good as human "experts" at such categorization ( and prognostication) decisions? Or are we training them to mimic human categorization and organizational biases? Interesting times in libraries and law courts.