The Aporetic

What works best in teaching digital media? In my experience, it’s the dumb, promiscuous brute search that crosses many different kinds of texts. Like the “boundary transgressing animal” of anthropology, the dumb search shows new kinds of meanings and lays out basic structures of thought. We want less reliance on metadata, which is mostly “what other people think this is,” and more on encountering the unfiltered text.

I’ve written a book on the history of debates about money in the US. It argues that when people talked about money, they used the same kind of language, the same assumptions, terms and metaphors, that they used to talk about race. I found evidence for this not by looking at texts organized by subject–“this is a text about money, this is a text about race”–but by searching for specific words and terms across a wide range of texts. This turns up things I might never have found otherwise

For example, searching for the term “colored AND “greenback” in the Library of Congress’ “American Memory” site turns up thirty two instances of the words being used in the same document: many of them turn out to be either politically charged minstrel show or campaign songs that connected the issue of greenback paper dollars to the enlistment of American American soldiers. I would have been very unlikely to have found this connection my searches were bounded by metadata, or filtered by a lot of framing: it would have taken longer and the surprising juxtapositions would not have been so starkly revealed.

Of course, I was already pairing the words–“it’s not digital technology that’s at issue here,” you might argue; “you already had the question in your head.”

This is true: as a product of the “cultural” or “linguistic” turn in history, typical of my generation of historians, I was looking for metaphors and word pairings. I’d already come up with the question before the LOC had digitized anything.

That’s not an accident though. I’d argue that the same epistemology that led academics to think of all documents as texts produced the idea of digital searching, in the same way that most scientific advancements appear first in science fiction. It’s not that academics led computer scientists, or that Jules Verne inspired scientists: it’s that in two disciplines/realms of thought the same kinds of thinking were taking place at the same time, perhaps in the way cubism, movies, military camouflage and the theory of relativity can all be described as artifacts of early twentieth century modernity.

This kind of juxtapositional “shock” works extremely well in teaching history. I often ask freshmen to enter the search term “runway” in the online Pennsylvania Gazette. The Gazette was Ben Franklin’s newspaper, the leading paper in colonial America. The search turns up hundreds of ads for runaway indentured servants, apprentices and slaves. The ads are short and cryptic and weird: they don’t say what we want them to say.

Philadelphia, December 6, 1774.

TWENTY SHILLINGS REWARD.

FOR any person or persons, who shall apprehend a certain Francis Villaneuse, a French runaway servant, by trade a hair dresser, about 5 feet 5 or 6 inches high, swarthy complexion, black hair, tied behind, and speaks but bad English; had on, which he went away, a beaver hat, a Wilton suit of clothes, a good deal powdered, the breeches almost new, light blue worsted stockings, new shoes, brass knee and shoe buckles; he took his dressing utensils with him, and may endeavour to pass for a freeman; he having runaway about 4 or 5 months since, when he had procured a forged pass. Whoever secures said servant, so that his master may have him again, shall receive the above reward, and reasonable charges, by applying to HUGH HENRY, in Chestnut street.

Students quickly notice that the ads don’t give modern descriptors: that they often don’t seem to have a modern understanding of or concern with “race,” or with the difference between a servant and a slave; that they spend way more time on the clothes than on the man. The exercise has the advantage of ignoring the conventional categories we use to organize the colonial era, and allowing them to see freedom and unfreedom, property and value, in entirely new ways.

A similar exercise asks them to go to the LOC’s searchable database of WPA “life histories,” the accounts of ordinary Americans collected in the 1930s by journalists and writers working for the New Deal. I ask them to search for “things they like.” The result, broad and messy set of mixed documents, dramatizes that they didn’t think this way in the past. A student entered the phrase “working out,” because she liked to work out: they did not use that term in 1935, and she had to refine her search, and try different terms. In the end she wrote an excellent short paper on the different understanding of physical exercise in the 1930s.

Both these assignments work by ignoring textbooks and textbook narrative and by immersing students in primary sources, but beyond that, they work by the dumbness of the search, the fact that the search engine does not reproduce existing categories of thought or recognize intellectual boundaries: it’s promiscuous.

American Memory is an extremely valuable database, but it has possibly the worst web page and search interface in the digital domain. That’s partly because it’s trying to preserve the metadata of the library–the subject classifications and categories of the print library and the media collection–while also delivering “dumb,” promiscuous digital results. It’s a boundary transgressing animal itself, and there’s something fascinatingly awful about it. But a happy accident of its bad design is its slatternly promiscuity.

The best example is Google books. Searching there for “colored” AND ‘greenback” produces over 4,800 results. Refining it by date to before 1890 produces about 1000 results, adding “the N-word” to the search gets the results down to about 99, and gives a broad range of kinds of texts, devoted to a broad range of subjects, but which reveal how the two ideas–race and money–were intertwined.

It’s probably fair to say that while I want the search to remain promiscuous, I’m not sure I want it to be dumb. I’d like to have better filters than Google books now provides–a proximity search, for example. And I’d want to see better ways to “mine” the data, as described recently in the NYT and here. Maybe what I want from digital searching is not “dumb” but compliant. (Although I have to admit the linguistic pairing of “promiscuous” and “compliant” gives me pause).

But at all costs, it seems to me, we want digital searching to avoid reproducing the necessary sins of its parents.