
6 minute read
WordGames
Joseph Thibault takes an optimistic look at process tracking and the business of evaluating effort and writing.
For those who’ve seen WarGames – the 1983 movie about nuclear brinkmanship between a 17-year-old and a military AI – the scene where “Shall we play a game?” gets typed out letter by letter might have popped into your head the first time you saw ChatGPT. That movie scene evoked a certain playful intelligence reflected on the bulky CRT monitor.
These days, every generative text and chat gives that WarGames vibe on speed, evoking an anthropomorphized writing process. It’s an illusion that gives you the sense ChatGPT is making effort and taking time to write, spitting out a stream of text that you can almost read, but well faster than you could ever write yourself. Actual compute times are obfuscated by its rule-based user experience, giving you a sense of exerting effort (artificially slowing down the artificial intelligence to be more “human-like”).
Despite the value in establishing trust and evaluating a student’s skill in real time, faculty rarely witness student writing as it happens. In online settings, it would be crazy to ask students to write via Zoom (unless being proctored). The good news is that you need not look over a student’s shoulder or require a screenshare to see a student’s process: there are an increasing number of tools that give you a front row seat. Get your popcorn.
These tools leverage a little-known fact: the Google Docs revision history shows only a tiny window to the expansive data Big G collects from writing. The first time I learned this was from an amazing tool called DraftBack, a browser extension for Chrome created by writer and developer James Somers (https:// jsomers.net/). His tool (introduced with a wonderful explanation in essay form) exposed what’s hidden in the default revision history of Google Docs: individual key presses. He’d cracked it open and exposed it through a beautiful time-series replay interface.
This tool, created in 2014, gave me and others an early view into what I believe will be one of the most important developments in 2025 for writing in class: a complete re-framing of writing as a cascade of data–an unfurling process–with traits like duration, edits, size, and space. Since 2022, “Process Tracking” tools have proliferated beyond Mr. Somer’s early exploration. They’ve made their way into
several academic integrity tools, spurred the addition of more robust revision history for popular text editors such as Moodle’s editor of choice, TinyMCE, and even been built into the Grammarly Authorship tool targeted at students (FWIW, my company Cursive did it first, and we don’t limit process tracking to Google Docs).
In a recent collaboration between the MLA-CCCC Joint Task Force on Writing and AI, academics explored the opportunities and risks that process tracking might afford students and faculty in the classroom. The verdict? A split decision.
On the one hand, those against the tools say process tracking is the next iteration of surveillance edtech. Focusing on the character-by-character creation of a document is not nearly as valuable as the finished product. It’s invasive, it’s onerous, it’s a slippery slope. As one commenter put it: what happens when the same process tracking is required for our (Faculty) own writing, publication drafts, or tenure documentation? In this view, the focus on the writing process through technology starts from a foundation of mistrust.
On the other, process tracking builds transparency through the writing process. It passively creates accountability without requiring additional effort. In an era that appears to be fraught with hard questions about AI-detection and -use, it makes visible what otherwise is opaque. You can see where the writer started, what the writer cut, and that huge block of text they pasted from Wikipedia. It can also differentiate contributions from other authors. And these are just the benefits to reviewers.
The most obvious benefit for students is that it can protect against false accusations from after-the-fact AI detection methods that ignore student writing time and effort. It also creates a robust revision history that can be mined as a reflexive tool. It gives teachers and students alike the necessary tools for spelunking the different phases of writing that otherwise would only be visible through frequent check-ins, conferences, or journaling. Want to see the ideation or outlining phase? Just rewind a little. Want to fast forward to the final stages of editing? Seek towards the end. No rewinding necessary.
Researchers have studied the process of writing through qualitative observation for decades. Quantitative data collected during the writing process has only had a place in the same research arc for mere years. If there’s some benefit to be gleaned from process tracking, for writers, readers, or both, I’m all for it.
Imagine reading a wonderful piece of prose with students and then playing back its writing process while the class watches. It unfurls character by character until it reaches the final state. Together, the class talks through the process, recognizes the effort, analyzes the word selection, revisions, and editing process. Your class recognizes the friction of creating the piece from the first keystroke to the last, which may have taken place over many minutes, hours, or days. Students suddenly see the cumulative effort of thinking and working as a process rather than just the end result. The context of hard work is both encouraging and sobering, but most of all, it helps to show the value of process, not just the product.
When he first launched Draftback, in 2015, James Somers said, “I worry that most people aren’t as good writers as they should be… they don’t realize it’s supposed to be hard; they think that good writers are talented, when the truth is that good writers get good the way good programmers get good, the way good anythings get good: by running into the spike. Maybe folks would understand that better if they had vivid evidence that a good writer actually spends most of his time fighting himself.”
Without process tracking for human writing, the example we’ll be left watching is the automated stream from the playful AI in a chat window: effortless, errorless, immediate, inhuman.
Want to play a game? It asks. No thanks, I think I’ll just write.