ChatGPT
In these SoTL blogs, I attempt to share timely, contemporary research that may support your approach to teaching and learning in higher ed. Over the past several weeks, there has been a discussion in the higher ed assessment community on OpenAIs ChatGPT (Generative Pre-trained Transformer - Wikipedia November 2022). In an extreme summary [from Webb (2022)], “how GPT-3 works is quite complex but there are two points that we can all understand:
It’s training on large chunks of the internet, plus some books; and
It works by predicting the next word given a sequence of words.”
This week (without going into technical details), I would like to share preliminary ideas through several articles of how educators might be broadening their research on how ChatGPT could impact teaching (including assessment, measurement, evaluation) and learning.
The first [non-research] article is entitled, “The End of High School English” by Herman (Dec 2022). The author summarizes their analysis of ChatGPT as a “program that generates sophisticated text in response to any prompt you can imagine, which may signal the end of writing assignments altogether—and maybe even the end of writing as a gatekeeper, a metric for intelligence, a teachable skill.” The author concludes that “What GPT can produce right now is better than the large majority of writing seen by your average professor.” As a reminder, I try not to “tell” colleagues what to believe, but I hope that I can provide ideas from teaching and learning arenas, which may not normally come across your virtual desk.
Another recent article entitled, “The College Essay is Dead” by Marche (Dec 2022) states that “You can no longer give take-home exams/homework … Even on specific questions that involve combining knowledge across domains, the OpenAI chat is frankly better than the average.” Hutchins (2022) has shared ideas to this article that was compiled through discussions with faculty which include:
“AI (so far) is not very good at scaffolding work from one assignment to another, so any time you can build writing assignments that build on prior work, it's more difficult to rely on AI to produce meaningful content.
AI (so far) has no concept of "indexicality," it can't refer to external objects in a meaningful way. If your writing prompt is "based on the annotated bibliography your peers shared last week, write about the trends or themes..." AI has no idea what to do with a prompt like that.
AI (so far) doesn't know how to synthesize knowledge or make inferences. If you ask it to write a book report or summarize an article, it can do that surprisingly well. If you ask it to compare and contrast the themes of two different books, it struggles to do that coherently and tends to stack two book reports on top of each other.”
There have been conversations about ChatGPT in the tech arena, although it has quickly arrived on the Inside Higher Ed website through the article “Deconstructing ChatGPT on the Future of Continuing Education” by Schroeder (Dec 2022). The author shares that “heutagogy, or self-determined learning, a student-centered instructional strategy which emphasizes the development of autonomy, capacity, and capability” will become the driver of GPT. The article indicates that “early response by academics who have tested ChatGPT is generally amazement at the speed, quality and depth of insight.”
As with most SoTL, the major asset is to provide instructors with research-based data which can empower each of us to create dynamic, engaging learning opportunities and subsequently support our students in timely, focused ways. Although most of these articles are not heavily data-driven [yet], it seems plausible that many of these concepts will be added to SoTL research agendas soon.
References
Zotero library with ChatGPT resources related to higher ed teaching and learning.
Webb, M. (2022). GPT-3 and Plausible Untruths. National Center for AI.
