Last November, OpenAI released ChatGPT, an artificial intelligence (AI) language model that generates human-like text in a way that feels conversational. When prompted to describe its capabilities, ChatGPT responded: “ChatGPT is a large language model that can generate human-like text based on the input it receives. It can be used for a variety of tasks such as language translation, text summarization, question answering, and text generation. It can also be fine-tuned to perform specific tasks such as conversation generation, sentiment analysis and more.”
The previous April, OpenAI released DALL-E 2, an artificial intelligence system that creates original images from a text description. It can also modify existing images to add and remove elements, or create stylistic variations.
So, what does this mean for teaching and learning?
(Scroll down for information about a talk on this topic on the NYC campus.)
I’ve been part of a dynamic email exchange among some of our faculty about the ChatGPT model. Here, shared with permission, are some of their thoughts:
John Misak points out that “you need to ask ChatGPT the right questions in order to get the desired response.” He shares an example in which he asked the ChatGPT model “to analyze Hamlet in regard to his having an Oedipus Complex and to quote Freud on the matter. It supplied in text citations for both the play and Freud. When I didn’t specify quotations for Freud, there was no reference to him at all.”
Navin Pokala mentions that the responses generated by ChatGPT are adequate for many students and assignments. He provided the same prompt three times. While one response was better than the other two, he stated that he would have to give full credit for all three responses. Jonathan Goldman
concurs, stating that ChatGPT provides “tepid-but-plausible” responses that are lacking in character or personality.
The conversation then moved to assignments that would minimize the temptation to use artificial intelligence.
Amanda Golden asks students to record themselves reading a page of a literary text. She then creates an audiobook and has students reflect upon the differences in reading and listening to the recording of the text. In a similar example, Lissi Athanasiou-Krikelis has her students watch a video interview of the author of a text they are reading for class and share a summary with main ideas and impressions. Jonathan Goldman asks his students to take a story or poem and re-write it set in their neighborhood and in current time. He writes, “one of my past students did ”Araby“ and had the protagonist try to impress his potential girlfriend by taking the G train and going and buying her vinyl at a pop-up record store in Williamsburg.” Lissi Athanasiou-Krikelis does something similar, asking students to connect readings to news reports or current events. “Students have connected Wizard of Oz to bitcoin, and Persepolis to the recent civic unrest in Iran.” Michael Schiavi asks students to create a conversation between two characters from different works. His example: “What would Mrs. Hale from Trifles and Nora from A Doll’s House say to each other about men’s treatment of them? Using quotation, how do we know that they might have this conversation? What conclusions might they reach?”
|
That said, you probably want to acknowledge – in your syllabus and in conversation with your students – the existence of these artificial intelligence models. They present an opportunity to discuss the ethics of using AI systems, how to do so appropriately, and how to acknowledge their use. After all, artificial intelligence is a tool, and we want our students to learn how to leverage it well.
Some examples:
Provide two versions of an assignment: the version you would normally assign, and a second version for students who want to use ChatGPT. Students who do not use ChatGPT complete the assignment as written, and submit it along with a signed statement that ChatGPT was not used. Students who do use ChatGPT will submit the assignment prompt to the ChatGPT model and copy the output. Then, using track changes, have the students correct inaccuracies, add additional detail, and offer alternative perspectives. All assignments can be graded on depth of knowledge – either of their original writing, or their changes to ChatGPT’s output. (Source: Ryan Watkins, Professor of Educational Technology Leadership, and Human-Technology Collaboration, George Washington University) Have students ask ChatGPT to critique a first draft of their response to a question, and provide suggestions for improvement. Then integrate those suggestions, and submit both drafts, along with the output from ChatGPT. (Source: Lucas Wright, senior education consultant, learning technology, University of British Columbia) My personal favorite is to incorporate ChatGPT into class discussions. You’ve often heard me talk about Think-Pair-Share, an in-class technique in which students reflect individually on a problem, then discuss it with a neighbor before the whole class talks about it. I like this technique because it involves everyone actively in the conversation, giving them time to rehearse and consolidate what they’ve learned. As suggested by Sarah Dillard, founder of Kaleidoscope Education, you can change this sequence to Think-Pair-Ask ChatGPT-Pair-Share. It gives students the opportunity to refine their ideas by using artificial intelligence, and to practice the skill of using it to enrich their own work.
|
So, is ChatGPT the future of learning or the death of critical thinking? It all depends on how you use it. What are you doing to incorporate this tool in your students’ toolboxes? Let’s keep this conversation going!