No, not some guy named Alan; that’s a capital i, not a lowercase L.
A while back I wrote a couple of posts about experimenting with ChatGPT to see whether I had a reasonable shot at spotting student work that was using the tool.
With school starting up this week, I’ve been thinking about what sort of policy to have about student use of AI. My university gives us teachers a lot of freedom as to our course policies; the official institution-wide policy is that student use of AI for assignments is prohibited “without the express permission of the professor”—which means we can give permission for anything we think is appropriate and academically justifiable.
So I did some more playing around with ChatGPT, and also with Claude.ai.
I began with ChatGPT, specifying, “Write a 700-word essay in the style of www.danolinger.com on the topic of sanctification.” I wanted to see whether it could write a blog post that sounded like me. (I know what you’re thinking; hold off on any judgment for a bit.) What it wrote—immediately—was pretty good. Although the title didn’t reflect my style here on the blog—you may have noticed that I like titles that start with “On”—it was generally pretty good as to content and basic style. I did notice differences in mechanics; it spelled out the names of the biblical books and used colons rather than periods to separate chapter numbers from verse numbers in references. But it did use the Oxford comma, though it used more commas than I would have in other constructions.
Sidebar: do you know why I use periods instead of colons in Bible references? Because a colon requires the Shift key, and the period doesn’t. Efficiency.
I thought I’d see how it handled a secular topic, so I asked, “Make the topic the migration of the monarch butterfly.” I’m no expert on the seasonal peregrinations of lepidoptera, particularly danaus plexippus, but what it churned out seemed very good to me.
Now, I had asked it to use the style of my blog’s website, where there are, as of this writing, 691 posts, and the speed with which it had responded to both of these prompts made me wonder whether it had even glanced casually in the direction of the blog. So my next prompt was “Delineate the parallels in style between your essay and those on danolinger.com.”
I was astounded by its reply—seven nicely formatted points identifying seven elements of my blog’s style, ranging from “conversational tone” to “use of relatable examples” to “integration of wonder and mystery.”
Yikes. ChatGPT is the most attentive reader I’ve got.
One more test—one that I think might cause it to stumble. “Add biblical references to the essay.”
Now this is an essay about a non-religious, and more specifically, extrabiblical topic, and I’m asking this non-brain to make biblical applications about something it doesn’t even mention. So what does it do?
It brings in Psalm 19.1, a key passage for the doctrine of natural revelation. Then it parallels God’s guidance of the monarch with his guidance of his people, citing Proverbs 3.5-6, and then the idea of finishing the race, from Hebrews 12.1-2. And it finishes with Ecclesiastes 3.11 and Psalm 104.24.
Bravo, ChatGPT, bravo. I’m impressed.
Later, at the recommendation of a friend, I popped over to Claude.ai and asked for the same thing, for the topic of the early date of the Exodus. Excellent essay.
Then “Write a 500-word short story set during the Exodus.”
Pretty good one. The kid’s even named Levi.
“Write a haiku about it.”
Waves part, path revealed
Israelites follow God’s call
Freedom’s light shines forth
“Write a sonnet about it.”
“Write a poem about it in Greek.”
It even included the line “Οἱ δὲ ἐν φόβῳ ἔφευγον πρὸς τὴν θάλασσαν Ἐρυθράν.” Pronominal use of the article. Nice touch.
Then I asked for a poem in Akkadian.
And it did it.
Transliterated—not in cuneiform—but recognizably Akkadian. (I’m not qualified to evaluate the literary quality or the accuracy of said poem.)
Yikes.
These tools are getting better all the time.
But no, I’ll never use them to write a blog entry.
So what’s my AI student policy?
For Greek and for Bible Doctrines, prohibited. I want the students to do the thinking—all of it.
For my Gospel of John class, I’ve decided to try an experiment.
“Any use of AI must be specifically credited with quotation marks and a footnote. AI output should not predominate in what you turn in.”
I have no idea whether or not this is a good idea. We’ll see how it goes, and I’ll ask the students for feedback at the end of the semester.
No chatbots were harmed in the creation of this blog post.
Photo by Andy Kelly on Unsplash