Marjorie R. Hershey, Indiana University – Bloomington
This essay will be added to the second edition of Dr. Hershey’s ebooklet, How to Teach American Politics (and Other Subjects) Effectively. While this work is in progress, APSA Educate is pleased to feature below.
ChatGPT is a specific program of artificial intelligence using a dialogue format (which makes it a “chatbot”). It draws on patterns in the existing information on which it has been trained. It’s a bit like a tremendously more sophisticated version of the autofill in many email programs (in which, when you write “thank,” it adds on the basis of known probabilities, “you so much”).
You can use it to provide information and explanations, talk back-and-forth with it, ask it to respond like a particular individual would, improve students’ writing, and create many other written products that can serve as exam answers or paper assignments. Introduced in late 2022, ChatGPT learns from its interactions with users over time, so that it becomes increasingly more effective and sounds more and more like the language used by a real person. There are many other such AI tools and lots more in development.
Although programs are now being written to identify material produced by ChatGPT,[1] there’s currently no definitive, widely-accepted way to be sure whether a student’s paper or exam answer was written by the student or by ChatGPT. Even when these become available, ChatGPT and other forms of AI will keep learning. We can’t make them go away. So as teachers, we need to find productive ways to deal with them. Warning students not to use ChatGPT, or banning its use, is probably about as effective as telling teenagers not to do any of the many other things we’ve attempted to ban over time.
Although programs are now being written to identify material produced by ChatGPT, there’s currently no definitive, widely-accepted way to be sure whether a student’s paper or exam answer was written by the student or by ChatGPT. Even when these become available, ChatGPT and other forms of AI will keep learning. We can’t make them go away. So as teachers, we need to find productive ways to deal with them. Warning students not to use ChatGPT, or banning its use, is probably about as effective as telling teenagers not to do any of the many other things we’ve attempted to ban over time.
Here are three approaches to dealing with the widespread availability of ChatGPT:
First, I’d suggest that it works much better to structure a situation to increase the chance that students will do what you want, rather than relying on punishing those who don’t. One way is to assign types of exams and papers that ChatGPT isn’t good at. At least at the moment, it is very good at locating information (including “what did x author say about y?”) and putting it together in essay form. It is less good at doing analysis, such as: How can we apply the principles of the course to the following specific case? Consider structuring your paper assignments very specifically (“the paper must include the following…”). If you require successive drafts of a paper, you can offer comments on draft #1 that students must respond to in draft #2.
Given that ChatGPT’s coherence drops after two to three paragraphs, you might assign longer essays – though we know that real students’ coherence can also drop dramatically after two to three paragraphs. If you’re suspicious about a student’s exam or paper, a brief oral exam will usually reveal whether the student knows the material, just as it would in the case of any other type of exam or paper.
Another is to avoid take-home exams – or, if you really prefer take-homes, don’t include short essays that simply require students to define terms or summarize readings. Alternatively, you can ask students to collaborate on a question in class, with or without access to their electronic devices.
A third approach is to use ChatGPT as a tool.
A third approach is to use ChatGPT as a tool. If you assign a paper on a specific topic, you might ask students to use ChatGPT to identify x number of topic headings to research for the paper. If you teach about the impact of context or structure, you could assign students to ask ChatGPT about a topic in two different ways and then compare the results. You can use it to offer examples of argumentative writing or descriptive writing.
You could ask students to create an essay on a topic using the app and then critique the results they’ve received, by locating other research that supports or disconfirms the contentions of the essay. Remind your students that these AI programs are perfectly capable of spreading misinformation. An NPR reporter noted, “”There are still many cases where you ask it a question and it’ll give you a very impressive-sounding answer that’s just dead wrong.”[2] ChatGPT says about itself, “You should not take everything I say to be true and accurate. It’s always important to use your own judgment and common sense, and to verify information from multiple sources before making any important decisions or taking any actions.”[3] At the moment, ChatGPT isn’t able to rank its information sources on their accuracy or reliability, though that may well be possible before long.
Currently, ChatGPT doesn’t seem to include in its huge database a universe of scholarly articles, and when it does, they may be accompanied by material from the online equivalent of the National Enquirer. Students who use it will still need to check the results independently. Identifying and correcting misinformation is an essential skill. They also need to know that the program doesn’t (at least currently) cite sources consistently and that it’s most likely to be accurate on topics that are widely written about.
As these and other forms of AI develop, it’s tempting to either rail at technology or keep using the same teaching techniques in the hope that students don’t know about the advances. They do. In teaching as in the rest of our lives, adaptation is survival.
(By the way, if you want to sound with-it, GPT stands for Generative Pre-trained Transformer. No, I can’t define that, but you can ask ChatGPT…)
Endnotes:
[1] For example, see Greg Rosalsky and Emma Peaslee. “This 22-year-old is trying to save us from ChatGPT before it changes writing forever.” ”Planet Money,” NPR. January 17, 2023.
[2] Quoted in Emma Bowman. “A new AI chatbot might do your homework for you. But it’s still not an A+ student.” “Morning Edition,” NPR, December 19, 2022.
[3] Quoted in Billy Perrigo. “AI Chatbots Are Getting Better. But an Interview With ChatGPT Reveals Their Limits.” Time, December 5, 2022, https://time.com/6238781/chatbot-chatgpt-ai-interview/
Marjorie R. Hershey is a guest contributor to APSA Educate. The views expressed in the resources APSA Educate features are the authors and contributors alone and do not represent APSA’s views.
