My quick
experiment with Pangram in July led to me becoming the bookends in
a
story about Pangram and higher education, including my plan to
retire in 2027.
I’ve been writing about retiring for a few years on my personal blog,
said as much to colleagues when asked when I’d apply for Full Professor,
noted it in my annual reviews, and geeked out about it with anyone
discussing personal finance (i.e., Financial
Independence, Retire Early). But I suppose this was sort of a
“stealth announcement,” as one colleague noted.
Because of stuff—many reasons—Dear
Reddit, Am I the Asshole was to be my last scholarly book. I
thought I might then switch to teaching full-time for a few years, as
long as I enjoyed it (i.e., Coast
FIRE) and Nora wanted to remain at her job. A few years ago I
realized AI would upend education; now teaching is much less
fun, so I am changing things up, as noted in the story.
If students in Joseph Reagle’s Communication in the Digital Age class
last spring planned to have Claude or ChatGPT do their assignments, they
found out early in the semester it might not go well. Reagle told
everyone that, among other checks, he planned to run their writing
through an artificial intelligence detector called Pangram.
Reagle, a former Massachusetts Institute of Technology research
engineer who’d been a professor of communication studies at Northeastern
University for 15 years, had recently discovered AI misuse by students
almost by mistake. He’d assigned another class to contribute to
Wikipedia, telling them the work had to be substantively their own and
not a chatbot’s. But Wiki Education, a nonprofit helping with the
assignment, used Pangram to review the submissions and detected multiple
apparent violations of Wikipedia’s ban on AI content. Some students
seemed to have just copied and pasted AI output.
…
At Northeastern last spring, Reagle did end up referring a few
students for possible discipline. After the semester ended, he ran an
informal experiment, giving Pangram hundreds of samples turned in for
his digital communications class before AI chatbots existed. It got them
all correct — fully human. Then he did the same with 50 samples each
from several semesters after ChatGPT’s release. For the earliest batch,
it detected AI in 6% of assignments; the figure peaked at 42% the
semester before he started running homework through Pangram. Once
monitoring began, the share fell back to 6%.
As a deterrent, his approach seemed to have worked. And yet the
policing left him exhausted and demoralized. “It’s a very
confrontational, oppositional stance,” he says. At the end of the
semester, he received some of the worst teaching evaluations of his
career.
Reagle recognizes that students will be graduating into a world
saturated with AI content, and that employers might well encourage them
to create it themselves. He’s grown so unsure about what this all means
for communication — and how to teach it — that instead of faking it,
he’s decided to retire early, at 55, after the coming school year.
There are comments.