Dr. Scott Graham’s “AI Has Entered the Chat: Promise, Peril, and the Future of Writing in Higher Ed” Talk (Recording Linked) 

Instructors and students alike are increasingly aware of the ways generative AI technologies are changing writing and higher education. This term, WIC invited Dr. Scott Graham (University of Texas-Austin) to discuss how instructors can navigate uses and misuses of artificial intelligence (AI) in the classroom. In “AI Has Entered the Chat: Promise, Peril, and the Future of Writing in Higher Ed,” Dr. Graham, who uses AI to study health and bioscience communication, reviews the many ways students and instructors alike are using generative AI technologies (including Chat-GPT) in and beyond the classroom. 

In the talk, he highlights potential benefits and drawbacks of AI use, and shares ways that instructors might employ large language models (LLMs) as an electronic pedagogical resource in the writing classroom, as well as tips for designing assessments that facilitate students’ learning about the writing process. He also discusses the many limits and drawbacks of this tech, including the ways that LLMs produce false information and sources through “hallucinations,” as well as the dangerous data collection and storage methods used by major AI tech companies. 

You can watch the recording of Dr. Graham’s talk here. 

Dr. Scott Graham’s workshop “ChatGPT in the Classroom: Practicalities & Pedagogies” (No Recording Available) 

Following his talk on using generative AI responsibly in the classroom, Dr. Scott Graham hosted a workshop titled “ChatGPT in the Classroom: Practicalities & Pedagogies.” In the workshop, he explained how generative AI technologies like Chat-GPT gather data and create outputs based on that data; he then led attendees through various prompt generation exercises using Chat-GPT, discussing the tool’s limitations and capacities along the way. 

In the first exercise, attendees prompted Chat-GPT to summarize a book or reading they had recently assigned to students, and then shared their observations on the AI’s outputs with one another. Dr. Graham then reviewed the basics of prompt engineering, the practice of developing inputs for the strongest possible outputs from generative AI tools. In the following exercise, attendees tested the limits of Chat-GPT’s ability to localize its outputs by prompting the tool to craft a letter related to issues in Corvallis and the Oregon State community.