What is it like to teach at the university level in the age of AI, and what does it tell us about the future of higher education?
At the beginning of the year, when ChatGPT hit the headlines as a threat to college education, I spoke about GenAI’s impact on education on WGBH in Boston. During the discussion, I emphasized that this technology represents an opportunity to change how and what we teach in universities.
I also document some reflections on my initial encounters with ChatGPT, and how it spurred me to rethink parts of my spring semester course.
After a semester of experimenting with this new technology in my undergraduate course, “Populism and the Rise of Anti-Democracy,” do I still hold the same opinion? In short, yes.
Acknowledging the presence of AI forced me to make changes to my teaching that I should have made a long time ago. It also pushed me to do something that I have never been good at as a teacher – to trust my students.
The question of generative AI in higher education has been rooted in anxiety over students using it as a tool for cheating. A secondary mainstream conversation has been an existential panic about the meaning of a university education.
Yet, my own experience has been different. I find that students are crying out for guidance and leadership on how to use AI for their lives and careers. Moreover, it presents an opportunity to reimagine and reaffirm what a university education means in this new world of generative AI.
I actively engaged with ChatGPT in a process similar to the one we outline in Reimagining Your Assessments in Light of AI, working in an iterative fashion with this technology to make my assessment stronger.
1. The first thing I did was to input exam questions into ChatGPT.
Its answers were good. Not A-grade material, but it got the main ideas and the basic facts right.
2. Next, I critically analyzed the output
I sat down and thought about what was ‘good’ about the AI answers and what was missing from them. By zeroing in what was missing from the responses I was able to clarify my expectations for students in their work. Specifically, this meant, contemporary examples from around the world, references to key academic texts, and in particular, critical analysis drawn from the writer’s own opinions.
This forced me to really articulate for students the specifics of what I was looking for in their work beyond the platitudinous phrases of ‘critical analysis’, ‘argument’, and ‘writing style’.
3. I openly discussed my experience with ChatGPT with my students
After this experimentation, I went back to my students and we had a discussion about what I had found. Together, we spent an entire class period discussing the strengths and limitations of ChatGPT for the context of my course. I said to them, “If you are just using ChatGPTwithout building on it, then congratulations, you have been replaced by it.”
Students responded with an avalanche of questions about referencing, contemporary examples from around the world, the use of political science concepts, and what an original argument actually looks like – all of the elements that generative AI was missing.
Over the course of one class session, we collaboratively moved expectations for student work beyond knowledge regurgitation and into higher-order creative and critical thinking.
Students in my course were allowed to use AI for any of their learning activities, as long as they referenced its use. All they were required to do was reference its use. After ‘stress-testing’ it, I felt confident that I knew its limitations and that my new assessment design would require them to focus their energies on more complex cognitive tasks that cannot be offloaded to GenAI.
4. I revised my final assessment
Since I started teaching, I’ve used final, essay-based assignments. However, after seeing the capabilities of ChatGPT, I was inspired to redesign my final assessment, giving students different options to demonstrate their mastery of course content– students could either take a more traditional final assessment or opt to complete a project-based option.
For the project-based options, students needed to create a visual story of populsim in one of the countries we studied as part of the course. They could use ArcGIS StoryMaps to combine videos, images, maps, social media, and anything else they found relevant to telling the story of populism in their chosen geographical area.
For the more “traditional” option, students were tasked with work that they might encounter if they decide to pursue a career in politics. First, they drafted a briefing paper for President Biden’s Forum on Democracy based on their understanding of the ‘threat’ of populism. Next, they analyzed the populism of Mexican President Obrador and drew conclusions for a general audience.
The changes I implemented meant that instead of grading a pile of repetitive essays at the end of the semester, I spent my time reviewing creative, original student work responding to authentic challenges in my discipline.
More Authentic Assessments and Better Faculty-Student Relationships
My experience points to the many options faculty have to adapt their courses to address the presence of ChatGPT. More fundamentally, my experimentation with ChatGPT in my classroom points to the opportunity we have at this particular moment to positively reshape the relationship between faculty and students.
The changes I made in my class were successful because I started from a place of trust with my students. Instead of framing the conversation about ChatGPT in my classroom from a place of punishment and admonishment, I had an open conversation with my students about technology, my expectations, and learning both in and beyond my course.
GenAI is here to stay. Corporations across many industries are investing heavily in AI tools for everyday work. We do our students an incredible disservice if we do not give them guidance on how to think about and use GenAI.
My story is one of many examples of educators who are working with their students to better understand the role of technology in higher education and the impact this intersection has on the rest of their lives.
Chat GPT was not used in the writing of this article. But autocorrect and autofill from Google Docs were.