How do you teach someone to write in a world of robot authors?
Writing, since the Mesopotamians carved cuneiform onto clay, has seen authors use the tools available to them to communicate and express themselves. While the image of a “writer” sometimes seems stuck in the amber of whisky and typewriters, those tools have progressed from quill pens to spell check to Google Docs.
But what does it mean when AI gets more involved in the actual creation of the words themselves? And how do we produce and highlight quality writing when we face an incoming tsunami wave of robot-generated text?
It’s a pet subject of mine and of Deepnews (which cares about quality content), which we’ve explored through the lens of misinformation, marketing and more. One important part of writing we haven’t covered, however, is learning to write, which is the focus of Dr. Lucinda McKnight, a lecturer in Pedagogy and Curriculum at Deakin University in Australia.
McKnight has written a couple pieces over at The Conversation about how we should teach students to write when in the future (and to some extent the present) there are machines capable of producing passable text. The rise of tools such as GPT-3 has led her from a focus on the effects of testing, which can push people towards bad, adjective-laden writing, to considering what students’ interactions with AI may be.
“It was interesting to see how our high-stakes testing environment had narrowed the teaching of writing down to something very formulaic, that teachers were just trying to really tick boxes to get the students to comply with things like formulas for paragraphs, formulas for sentences, formulas for whole essays. They were doing very little writing that actually connected with the real world, that was anything like what a journalist might do or even someone maintaining their own blog,” she told me this week.
“One of the key things that I was thinking was that if school writing is becoming so formulaic, this formulaic kind of writing is exactly what machines can do so well and are already doing.”
So what kind of writing is the “human” kind? The elaborate and literary? Will the focus of teaching writing in the future be in creating purple prose so swollen with complexity that the sentences stick out like a sore, undeniably human, thumb?
McKnight doesn’t think so. She says that in a world of AI text, the skills that students need “ will be much more aligned with what we would describe as a ‘process writing’ approach, drafting, editing, collaborating, sharing and developing those kinds of soft skills that are going to be the skills that are useful in the future.”
“The emphasis, I think, is going to shift from the sort of humanist individual as writer to a more dispersed, networked notion of writing.”
Beyond the tools for generating text, in the context of education there is also the assessment of text, and whether machines can be used to help those on the teacher side. There are already tools for plagiarism detection that go beyond just measuring a students’ input against text online, and measure a students’ text against work that they have already submitted to see if there are abrupt changes in the patterns of writing that they have used previously.
Deepnews is of course interested in the evaluation of text, which we do through our quality score to judge its level of depth. Right now even the most advanced systems like GPT-3 have difficult generating high quality text in the Deepnews sense, though it is easy to imagine scores like ours needing to evolve to make sure that we are ahead of the curve of highlighting the added value that a human puts into a article (including the soft skills that McKnight talks about) as text generation advances.
Part of that work, in the world of technology as well as the world of education, is helping determine exactly where that human work occurs and how to make the most of it.
“Hopefully there will be a whole resurgence of creativity as something that humans can do that machines can’t do, or they can’t do yet. But I don’t think it will be a pure thing where there’s a binary division between humans and machines. I think that it will be something that is collaborative in the sense of working together,” McKnight said.
