My Dearest Benson,

I have been perusing English Journal articles from 1964, and I must admit I have been feeling some ennui as most what of what I’ve been finding about new media in that year has been rather humdrum; I required a third cup of coffee just to keep at my reading! I was about to knock off for the day, when lo and behold I discovered an article that startled me like no other: “The Computer and Character Analysis” by Allan Ellis, Assistant Professor of Education at Harvard University! You can no doubt imagine my complete surprise and sheer delight at this unexpected archival find!

In this remarkable text, Ellis demonstrated a computer-aided textual analysis of the canonical high school novel, Huck Finn, in an attempt to convince English teachers that “that a computer, often thought of as an instrument to launch missiles or to solve equations, is capable of reading the text of a novel and of finding evidence that can help us to understand a theme or a character or an author” (p. 527). Ellis argued compellingly that English teachers should not dismiss the computer as simply a numeric device useful only for math instructions, but instead should work to explore how the computer’s superior counting abilities could be put to use for literary analysis.

In addition to touting the ability of computers to quickly calculate word frequencies in a novel, Ellis demonstrated an early form of digital sentiment analysis in which the computer program searched for recurring clusters of words that reflected particular emotions. This

sentiment analysis of Huck Finn’s dialogue enabled Ellis to offer empirical support for interpretive claims such as, “every time Huck is confronted with some societal convention or other- religion, school, superstition, family, manners, or even the rules of his friends' club-he is distressed or anxious” (p. 527). I was absolutely astounded that Ellis’ computer-based textual analysis was in some ways more complex in its calculations than our own work with Voyant more than fifty years later. Consider me humbled!

And yet--I also noted how Ellis at times overstated the computer’s power as a tool for literary analysis, even going so far as to say the the computer was “totally objective and unconcerned with proving anybody's point” (p. 527). Although I appreciate how computers can help us challenge our preconceptions about texts, I cannot help but notice how Ellis’ program was designed to look for particular emotions and themes while ignoring others. I am thinking, for example, of how Ellis’ computer-based analysis seemed not to notice instances of racism in the novel -- alas, it appears that the computer did not help Ellis analyze the text beyond the unmarked lens of whiteness.

Another thing that I found odd about this article was that there was very little mention of pedagogy--a true rarity in our beloved English Journal. I expect that part of the reason the classroom was not mentioned was that computers were scarce, large, and expensive in this period. Harvard professors may have had computer access, but computers were not present in everyday classrooms--though clearly Ellis and the editors of English Journal hoped they would be soon!

The process of preparing literary texts for computer-based analysis was laborious as well: Ellis’ experiment required that every instance of Huck Finn’s dialogue “was punched onto IBM cards” (p. 524). The use of the passive voice here fascinates me, making we wonder who exactly did this punching? (Likely underpaid and unrecognized women, I’m afraid).

Needless to say, I’ll be eager to hear what you think of our new friend, Ellis, and to find out if you find any other computer articles when you turn to reading 1965.

I hope you also are experiencing some respite from this dreadful summer heat this week, as we have been here in Cincinnati… Despite hailing from swamp-drenched Florida, I find the dastardly warmth of summer increasingly hard to bear -- my enfeebled constitution only grows more delicate as I enter middle age. When the mercury rises, a proper Campari Spritz truly is the only suitable cure. May we enjoy such refreshments and convivial conversation when we next meet!

Most Sincerely Yours,

Encl: Ellis, Allan B. (1964). The computer and character analysis. The English Journal. 53(7), 522 - 527.

Dearest Benson, you dashing rogue--

I was incredibly pleased to to receive this exciting missive! It appears we have new computer friends, Stowe and Maggio!

I must confess I had been worried that the computer would never make it beyond the walls of Harvard in the 60s, but in the article you so thoughtfully enclosed we see a computer put to some good use in a high school setting. Of course, it is fascinating that the computer in this scenario was really only used for the rote task of calculating scores. All the multimedia tools and even the “Edex” student response machine -- what today we might so indecorously call a “clicker” -- were analog. Still, it was so exciting to see these teachers combining digital and analog media in creative ways--looking forward to a future when they hoped the computer could do so much more than just crunch the numbers. And yet--it certainly was revealing that most of the creative work here was done by the teacherxs not the students, alas.

Enough about our dear friends Stowe and Maggio -- I have found an exciting new chum to join our banter: Arthur Daigion, Professor of Education at University of Connecticut. In “Computer Grading and English Composition” (1966), Daigon reported on a study in which a computer was trained to evaluate high school writing in ways that correlated with the ratings of experienced English teachers. Although Daigon acknowledged that human graders could offer valuable feedback about writer’s ideas in ways that computers could not, he also noted that human English teachers had difficulty finding the time to give this kind of intensive feedback; “bearing a 130 - 150 typical student load for most teachers, leaves the instructor a choice between less than minimal

writing thoroughly graded and a reasonable amount of writing perfunctorily graded” (p. 47). If the computer could be taught to efficiently handle some aspects of writing evaluation and feedback, then teachers would be freed to do more “individualized teaching” (p. 52) about more complex writing concerns.

Daigon offered copious detail about how the computer could be programmed to grade. In addition to being trained to look for common misspellings, grammatical “errors,” and “solecisms,” the computer also could evaluate the frequency of uncommon words and complex sentence structures--both of which were seen as signs of superior style by experienced English teachers.  As an example of the kind of feedback a computer could be trained to give, Daigon shared this response: “John, [we are told that using first names softens criticism] please correct the following misspellings: believe, recieve. Note the ie, ei problem. You overuse the words interesting, good, nice; then was repeated six times. Check trite expressions. All of your sentences are of the subject-verb variety and all are declarative. Reconstruct” (p. 50). We can see here that Daigon’s computer was indeed quite precise and efficient at giving feedback about spelling and style, though I noted that this feedback gave us no sense of the topic or argument of John’s paper.

As our friend Ellis showed us, computers are devilishly good at counting words! It makes sense that a computer would be better and faster than a human at pointing out when writers are overusing certain common words -- one of my pet peeves as you know. And, computers love binaries: one versus zero, right versus wrong. So, it makes sense that the computer grading programs enacted a current-traditional writing pedagogy that

focused on the correction of “error” first and foremost. If Daigon’s computer ratings correlated with those of experienced English teachers, I take that as a sign that those teachers were emphasizing grammar and style above all else in their evaluations--enacting a writing pedagogy that was current then but has rightly fallen out of favor today.

Of course, Daigon admitted that it was a problem that the computer could not respond to the content of paper as a human would -- noting that the computer “cannot honestly say ‘I disagree with you.’ ‘I was moved.’ ‘Your topic was sensitively handled.’ ‘I am reminded of a personal experience’” (p. 50). I appreciated Daigon’s recognition that humans still have powers that machines do not, but then I was taken aback when he argued that “one could [also] say that the failing in a human contending with student composition is that he is not enough like a machine, which can accurately and consistently respond to discernible elements of style without tiring perceptibly” (p. 50). Oh brave new world! The computers are coming for our jobs!

For Daigon, the computer was not only faster than human readers, but it was also more objective. Daigon pointed out that teachers often let biases toward particular students influence their grading, but a computer would not mark down a student “for appearing in class needing a haircut and a bath during the greater part of the school year, looking at the teacher in a surly manner, or not being one of the ‘nice’ conventional students in the class.” (p. 52). When I read this, I found myself shaken to the core -- perhaps computers could be the solution to persistent problems of classist, racist, and sexist bias in education?

After I picked myself up of floor, however, I realized that Daigon was overstating the case for the computer’s objectivity. After all, Daigon’s computer was programmed to penalize students for using supposedly non-standard grammatical constructions and words--the computer was in fact a more exacting enforcer of linguistic classism and racism than the human teacher. While a subjective human reader might overlook the use of a so-called “solecism” in a paper that otherwise made a strong argument, Daigon’s computer had no choice but to enforce the classist and racist rules it had been programmed to apply. After all, the computer was programmed by humans and so it was not a surprise that it reflected the biases of those teachers who built it! (I dare say that this remains true of automated essay grading programs today.)

I have enclosed Daigon’s article and I am eager to hear your thoughts. I worry that perhaps I am being too hard on dear old Daigon? (As critical as I may be of Daigon’s biased attempt to eradicate solecisms, I must admit that if I could write a computer program that would eradicate use of the dreadful word, “impactful,”  from student writing, I most certainly would do so post-haste!)

It would be jolly good of you to come visit me here in the Queen City. I am certain my feline companion, General Tso, would be delighted to see you -- as would yours truly. Although I find it unseamly for a gentleman to drink an old-fashioned in the heat of summer, I have all the requisite materials on hand should you desire. I might even join you in a mint julep, should I be forced to uncork my bourbon in these dreadful summer months.

Please do write again soon -- your letters are always such a joy to read. Although I miss the speed of email to be sure, there is nothing like the feeling of receiving an honest-to-goodness letter in my postal box.

Your obliged and affectionate friend,

Encl: Daigon, Arthur. (1966). Computer grading of English composition. The English Journal. 55(1), 46-52.

Dearest Ben (if I may be permitted such informalities),

Your epistle has delighted me as always. It hadn’t occurred to me to consider how strong a connection there was between the 1960s instructional technology craze and the recent excitement around MOOCs. Although I remain a MOOC skeptic, I grant that Ohio State’s approach to computer-assisted feedback is much more critical and rhetorical than what we saw with dear old Daigon. I’m tickled with the idea of the computer as not an objective authority, but rather as just one more peer reviewer with a special set of skills (and limitations).

After reading the Auciello article you so thoughtfully enclosed, I found myself wanting to understand more about the context of computer-assisted instruction in the 1960s beyond our beloved EJ. An extensive search of the “card catalog” turned up an exciting 1967 monograph by Edmund J. Farrell, English, Education, and the Electronic Revolution, that focused primarily (though not exclusively) on projecting how the emerging computer revolution would transform the teaching of English.

Surprisingly, Farrell dated the rise of computer-assisted instruction all the way back to 1962, when both IBM and System Development Corporation came out with systems in which a central computer was able to deliver instruction to a wide number of students at their own terminals (p. 20 - 21). The work of System Development Corporation was reminiscent of our friends Stowe and Maggio, as it entailed “a digital computer to control and select material presented and to analyze [student] responses], a magnetic tape storage unit, a typewriter for

printing out data analysis, a slide projector and screen for presenting education materials, and individual desks with keyboards for the students’ responses” (Halacy, 1964, p. 207 qtd in Farrell, 1967, p. 24). Amazingly, the computer in this case was serving not only as grader, but also a kind of teacher--both providing instruction and evaluating student work.

Before you get too scared that the computers are coming for our jobs, though, it is important to note that Farrell still saw a role for human instructors in computer-based classrooms:  “Freed by the computer from dull drill, the teacher of English could spend more time in tutorials and small group discussions, engaging students in a dialogue about human experiences  and values as they are transmitted through literature” (p. 68). I was relieved to see that Farrell’s classroom of the future still made room for human dialogue about complex texts, yet I found it curious that Farrell focused almost wholly on the human power of reading and discussion while remaining silent about how writing would fit into the computer classroom. I found myself worrying that perhaps Farrell assumed that student writing could be taught and assessed solely through the “dull drill” of the computer -- a frightening prospect indeed.

As I kept reading Farrell, though, I started to see that he ultimately outlined a critical humanistic vision of computer instruction that remains relevant today. Resisting the tendency to see computers solely as the domain of math and science experts, Farrell insisted that
“the electronic revolution must have the guidance of humanists if it is

to be that which it can be, the instrumentality for releasing the creative potential of each individual, rather than an ingenious means for degrading human life” (p. 68). Here, Farrell helps me realize that although humanities and the computer may often seem to be in tension, it is precisely this tension that makes their combination so necessary. We need humanists to stay involved in instructional technology conversations so that we can help ensure that computers are used in more humane, creative, and equitable ways.

On a personal note, I must say that your reference to Colonel Pepper brought back some fond  memories -- it truly ‘tis such a pity that he and the General never met. Then again, if they had joined forces (and gained opposable thumbs), we would have been done for I am afraid!

I have enclosed some photocopied passages from Mr. Farrell’s monograph for you to peruse. I hope I have included enough tantalizing bits to inspire you to procure you own copy.

As I write, I sit here in the sweltering heat, fanning myself with your last letter. It has become frayed, I am sorry to report. I apologize for treating your treasured missive so rudely. I must buy a proper fan...

Yours ever, truly,

Encl: Excerpts, Farrell, Edmund J. (1967). English, education, and the electronic revolution. Champaign, IL: National Council of Teachers of English.

Ben, you sly dog --

Who would have thought that we would see the return of good old Daigon?! I read the enclosed article with much interest -- despite its tarnished state. And, I must say I like Daigon’s arguments much more this time around. I was pleased see him growing more critical of the ways in which computer programs often reduce learning to simple yes / no and true / false questions -- I love his futurist vision of a multimedia computer program that would help students engage open-ended, “thought-provoking” questions about the texts and contexts of poems.

And yet… does it strike you as a little disconcerting that we as computers and writing scholars find ourselves most cheering when Farrell, Daigon and the like make powerful arguments for how the humanistic value of literary study can transform computers? Have we crossed sides in the literature vs. composition divide or maybe just returned temporarily to our roots as idealistic literature majors? Perhaps, perhaps … (it would not be the first time I’ve been accused of such) ... but I expect the reason we find ourselves connecting with these literary arguments so much is that we have yet to see an article that actually imagines the computer as tool for student writing. In this period, it was at the nexus of literature and computing where the most radical imagineering was happening!

When it comes to writing, however, it appears that English teachers imagined computers solely as grading tools, as Daigon did in his first article. And, I’m afraid I must report that this trend continued into the

1970s (see enclosed). The first article that rung in the illustrious decade of our collective births was “Essay Grading by Computer: A Laboratory Phenomenon?” (1971) by Henry Slotnick of the National Assessment of Educational Progress (Ann Arbor) and and John Knapp of the  University of Illinois. I must admit that I wanted to dislike this article, and yet in the end I found it to be a remarkably thoughtful and balanced review of the potentials and pitfalls of computer-based grading at the time.

Reporting on their own research on computer grading and reviewing other studies (including Daigon’s), Slotnick and Knapp argued that computer grading was not yet ready for widespread use, and then they went on to elucidate a variety of technological and curricular changes that might make it possible for computer grading to be used in a limited way in college writing classes in the near future. Although Slotnick and Knapp found in their own studies (and in the literature they reviewed) that it was possible to program a computer that would arrive at similar evaluative judgements as human readers, they importantly pointed out that the computer had not actually learned to read papers like a human would. Whereas humans tend to evaluate elements of student writing in holistic, intuitive ways, “computers look for approximate qualities of writing--qualities which are logically related to the intrinsic qualities sought by human judges but are stated with the finite clarity required to make them programmable” (p. 76). One example of this is that computer could be trained to recognize linguistics features of “subordination of thought” (p. 75) and that these might correlate with

the complexity of the argument. But, we must recognize that that human readers bring a great deal more contextual knowledge to evaluating any student argument than merely looking for patterns of subordination in sentences.

And, of course, today’s computer grading programs are not much better. As Les Perelman has shown, it is possible to submit an incomprehensible nonsense essay to a modern automated essay grader and have it receive a high score because it meets required word counts, uses high-level vocabulary, and is relatively free of nonstandard grammatical constructions (Anson & Perelman, 2017; Perelman, 2014; Kolowich, 2014). At a moment in which educational stakeholders are increasingly looking to computers to evaluate standardized tests of student writing (Smith, 2018), it is crucial that we stay alert to what is lost when we reduce writing evaluation to simply a matter of counting words and applying simple rules.

In addition to pointing out the perennial problem that human readers and computer readers rely on differing evaluative logics, Slotnick and Knapp also highlighted a somewhat simpler technological problem that needed to be solved--the problem of how to make the process of preparing papers for evaluation more efficient. In particular, Slotnick and Knapp looked forward to the development of an  “accurate character reader (a machine which can read typed or printed material and put it on magnetic tape or punch cards)”(p. 80), which would allow computers to quickly read student writing that had been composed with the more ubiquitous technology of the typewriter.

Once the character reader problem was solved, Slotnick and Knapp suggested that computer graders might be integrated into composition classes as an ancillary tool for providing feedback on first drafts. In this model, the student would submit the first draft of a paper to the computer, and the computer would provide feedback following programmed rules specifically aligned with the particular assignments. Both student and teacher would receive a printout that:

would suggest to each student ways in which he might improve the paper before he submitted it to you [the teacher]. It might also tell him how the paper compared to his earlier writings; it might indicate areas in which he was improving (according to standards you have set) and where more work needs to be done. It would definitely tell him that if he is confused, or thinks the comments are inappropriate, that he should see you, the final judge. You would also receive a copy of the printout. And while you may be impressed with the computer's functioning, you may feel that it is inadequate or otherwise overly responsive in certain areas. You would report your criticisms to the person responsible for the programming, and/or inform your students to ignore certain things the computer had told them. In short, your awareness of the computer's functioning would be used to improve its (the computer's) utility. (p. 87)

Finally! A vision of computer feedback that does not totally scare me -- indeed, I am left intrigued. I appreciate here how the computer was not positioned as a replacement for teacher - student dialogue, but rather as a possible instigator of further human conversations about writing and revision. I also think it crucial that Slotnick and Knapp realized

that one-size-fits-all response programs were untenable and (as a result) writing teachers needed to play a role in the ongoing development and refinement of the response programs they would use for particular assignments.

Reading Slotnick and Knapp has inspired me to consider making more use of machine-based response in my classes. I can see how asking students to have their work read by a range of computer protocols could help us open up some exciting conversations about the potential benefits and limitations of using computer feedback to spur revision. Even something as simple as making a wordle of essay (showing most frequent words) can be a way to help students focus and re-see their work.  I still am not comfortable with the idea of the computer as a sole evaluator, but I am willing to allow computers in as another possible audience that students must consider when they are writing for web -- no doubt, understanding the logic of search engine or the social media algorithm is a crucial part of audience analysis for digital writing (Gallagher, 2017). Of course, I am wandering pretty far from Slotnick and Knapp in my musings here, but I do so since I find their broader model of the computer as an additional reader in the classroom (alongside students and teachers) to be a compelling one.

I suppose it is time for me to step down off my soapbox -- you know well how conversations about automated essay scoring can get me hot under the collar!  

On another note, I must admit that I too would welcome the ascendance of our cat overlords -- they could not do any worse than our current leaders, that is certain! I, however, am not willing to submit essays to be graded by cats -- I am afraid they would too easily be bribed with tuna to award unmerited scores.

I am very curious to see what new computer articles we may find as we keep reading the 1970s EJ. The decade has been off to a great start with Slotnick and Knapp...

Please do write soon!

Your steadfast colleague and faithful friend,