Lecture: On “An Algorithmic Criticism”

Stephen Ramsay’s Reading Machines: Toward an Algorithmic Criticism is the hardest text we’ll engage this term, and the first chapter, “An Algorithmic Criticism,” is likely to be the most difficult thing we read. That said, there’s much to learn here, and if your eyes glazed over at the first half of the chapter, it’s worth returning to because and pages 11-17 are not only far easier to understand, they demonstrate what Ramsay means by “an algorithmic criticism.”

Analysis vs. Criticism

To better understand the reading for this week (as well as the book as a whole), keep in mind that Ramsay’s book Reading Machines is about the possibility of using computers to help us engage in interpretive (hermeneutic) acts of written texts. Interpretive is the key word here, and Ramsay equates it with the act of criticism. As Ramsay notes in both the introductory “Preconditions” as well as the first chapter, “An Algorithmic Criticism,” computers are great at textual analysis which involves “counting, measuring, and (in a limited sense) verifying data” (18), that is, the production and/or comparison of “objective” data that can be quantified, such as the number of times a particular word is used or the stylistic features of a particular author or the rhyme scheme of a poem. What computers are not good at is the act of literary criticism which involves the interpretation of texts that is predicated on our “subjective” understanding of how texts create meaning and what that meaning suggests.

Recognizing the distinctions Ramsay is making between textual analysis and literary criticism is crucial to understanding the point of “An Algorithmic Criticism” and Reading Machines as a whole. The distinction Ramsay is making here is this:

Textual Analysis involves quantifiable data related to the features of a text: frequency of  words used, repetition of stock phrases, rhyme scheme, syntactical structure, words per sentence and sentences per paragraph, and stylistic features such as frequency of descriptive elements such as adjectives and adverbs or adjectival and adverbial clauses. So, for instance, textual analysis could ask such a question as “based on the stylistic features of the Federalist Papers‘ articles known to be written by James Madison, John Jay, and Alexander Hamilton, which of the three is the most likely author of Federalist Paper article X?” Textual analysis involves what we might call a “scientific” or “objective” understanding of the features of a text, and it produces quantifiable data that can be easily replicated and from which one may factual statements about that text.

Literary criticism, on the other hand, involves issues of interpretation that are more open-ended or based on the subjective experience of readers. Questions of criticism may take the form of “why is Odysseus’s argument to Achilles in the Iliad so rhetorically effective?” or “in what ways does Mark Danielewski’s novel House of Leaves conform to the conventions of postmodern fiction?” or “how does Shakespeare demonstrate Lady McBeth’s feelings of guilt over the murder of King Duncan?” or “in what ways is Terry Pratchett’s sword and sorcery novel The Last Hero indebted to Victorian Old Norse medievalism, and how much of that indebtedness might we trace through Robert E. Howard’s Conan stories compared to a direct knowledge and understanding of Old Norse literature?” Literary criticism produces interpretive readings of a text based upon subjective understandings of what the text says and does.

While evidence in analysis might involve the frequency a character uses a word as compared to other characters in that text, evidence in criticism involves the creation of a persuasive reading of that text based upon elements in the text, While the verifiability of analysis simply involves recounting and/or measuring the features counted and measured, verifiability of criticism (whether other deem it valid evidence) is based upon how persuasive the interpretive argument is.1

Computers, Ramsay tells us, are good at dealing with counting, measuring, and verifying data, that is, with analysis of texts, and that is how they have most often been used by literary scholars. In other words, computers are good at textual analysis.

The purpose of Reading Machines, and of “An Algorithmic Criticism” in particular, is to explore how we might use computers for interpretive purposes, that is, to aid us in literacy criticism. And that is what Ramsay does explore in the second half of the chapter, roughly pages 11-17.

In pages 11-15, Ramsay demonstrates how we might use a computer to generate quantifiable data – in this instance, word frequency by character and then by gender in Virginia Woolf’s novel The Waves – not to create data about features of the text which we then report as a form of analysis (a large proportion of shared terms of the women in the novel focus on traditionally womanly or domestic issues, while a large proportion of the shared terms between the men focus on technological, political, and professional issues), but as a set of quantifiable data that we use to focus our interpretive attention for the purposes of criticism (e.g., as many critics have noted, Lewis does seem concerned about his status as an Australian outsider while the other characters seem to either take no notice of it or simply don’t care, which in turn helps us understand why Lewis does X, Y, and Z).

In short, Ramsay is making the argument that while literary computing has focused on using computers to do what computers do best, that is engage in textual analysis (issues related to understanding the features of a text), we can ask computers to analyze texts in ways that produce data that we can then use to identify fruitful lines of interpretive inquiry in order to engage in literary criticism. In other words, while we can’t yet ask computers to interpret literary texts for us, we can use them for more than straight-forward analysis to describe the features of a text. While that description (analysis) can be meaningful and important for some kinds of inquiry, we can use that analytical data as a starting point for our own acts of interpretation. This computer-assisted interpretation is what Ramsay names an “algorithmic criticism.”

If you read “An Algorithmic Criticism” and it didn’t make much sense, I strongly suggest going back and rereading it, at least carefully skimming the first half the chapter and reading closely the second half.

Why We’re Reading “An Algorithmic Criticism”

While textual analysis does serve an important function in our understanding of texts, literary and otherwise, textual analysis is not the whole of literary or rhetorical scholarship. While I don’t want to imply hierarchies of scholarly practice here – I hope that last week’s examination of the history and practice of textual studies made clear just how vital textual studies is to literary scholarship even as it fell out of favor with the rise of literary criticism in the middle half of the 20th Century (starting with the rise of New Criticism) – only using computers for one kind of textual engagement (analysis) when literary scholars have two primary forms of textual engagement (analysis and criticism) is limiting. We’re reading Ramsay’s book because it explores the neglected side of computer-assisted literary scholarship.

We’re also reading “An Algorithmic Criticism” because it points to the use of computers as an inventive tool (inventio, 7, 16), as a tool to create (criticism) as well as a tool to decode (analysis). Just as we can use computers to engage texts by using computers to create new texts through glitching or through sampling and remixing texts, we can use computers to engage texts by creating new interpretive readings through acts of algorithmic criticism.


  1. One of my professors put it this way: literary criticism is a game, and the person who creates the most coherent narrative of a text while taking into account the largest amount of that text wins the game. That is, the more of a text you can use to support your audience-accepted reading of a text, the more comprehensive, and therefore valid, your act of interpretive criticism. In other words, while it is accurate to say that “Romeo and Juliet is about a romance gone bad,” such a statement doesn’t come close to addressing the why and how of the romance gone bad. On the other hand, the argument that “Romeo and Juliet is about the threats to civil society that are caused by the feuds between the powers of that society and the effects those threats can have on the social and interpersonal fabric of that society as evidenced by the ill-fated romance between the children of two feuding clans” is a far more comprehensive interpretation of the play because it alludes to the tragedies that befall not just Juliet and Romeo themselves but their parents, Juliet’s nurse, Friar Lawrence, Mecrutio, Tybalt, and Count Paris.

Leave a Reply

Your email address will not be published. Required fields are marked *