How to Learn to Write Like a Machine

I spent the past few weeks on two projects to help improve students’ academic writing. The first was an intensive course I already mentioned in an earlier post (a post, incidentally, that led to a data set I’m now crunching and will report back on later this month). In it, a colleague and I presented lessons in topics like clarity, concision, and coherence, and helped the students workshop early drafts of their master’s theses.

I’d like to think that the students’ writing will be better for the time we spent together, and the change I was able to easily measure — comparing homework from the start of the course with that from the end — showed incremental progress. More dispiriting was seeing that even when our lessons had been absorbed and applied, topics we hadn’t tackled still hampered the texts as a whole.

My second project provided a quicker pedagogical hit, which is slightly surprising because it wasn’t designed to teach. A Ph.D. student had asked me to edit one of the crucial chapters in his dissertation, which I did. Two weeks later, he asked for help with the introduction and conclusion. I started in and was struck by how much stronger this writing was, despite only small portions of the material overlapping with the text I’d edited earlier. When I remarked on this, he explained that he had read my earlier edit in tracked-changes mode and learned a lot from it.

This reminded me of how I learned to write — or specifically, how I learned to write for work, on the job as a young journalist. Sometimes I was lucky enough to have my boss pull me over to sit with her while she went through my stories. This made life easier for her, as I could answer questions then and there; but it also let me watch over her shoulder as she improved my work. Later, I “watched” by following pieces — mine and others’ — through the newspaper’s publishing system, observing the changes that were made by section editors, page editors, and copy editors. I got no explanation of why these choices were made — there’s no time at a daily news operation for typing out your reasoning — but I could guess, and recognize patterns emerging that I would later imitate.

I don’t need to tell you that pattern recognition is hot right now, in academe and elsewhere. No longer do computer programmers code their algorithms with instructions for how to do something. Instead (or in addition), they encourage them to learn by observing reams of data — the amped-up equivalent of my 20-something self watching a bunch of newspaper articles get edited every day.

As far as I can tell, this enthusiasm has yet to hit English departments, where editing is not exactly a celebrated pedagogical tool. Composition teachers I know write long, thoughtful responses to their students’ work but rarely tackle line-for-line improvements. Writing centers get deeper into the nitty-gritty, but still do this by giving advice rather than letting students sit back and watch their sentences get rewritten. I’ve certainly never seen a university writing center advertise in-house editing services as such.

Partly, of course, this is down to workload, but I do wonder if we’re giving machines credit for their ability to learn by observing while at the same time ignoring humans’ ability to do so. Our original ability to do so, I might add: We were here first! It may be time to persuade my department head to add “Interactive Editing 101” to our course offerings. Or link up with a newspaper and give students a virtual window on copy editors’ work—assuming copy editors aren’t phased out entirely in the meantime.

My Ph.D. student, incidentally, has landed himself a professorship and asked whether we could continue to collaborate. I’m not counting on much work; if his pattern-recognition skills up to now are anything to go by, he soon won’t need much more than a spell check.