Kevin Frazier

Context

What do you teach and how have you been thinking about artificial intelligence in the context of those courses?

I used Midjourney in my THE-252 Lighting Design class in the Fall of 2023.  We used it to explore the ways collaborative artists can communicate in a visual medium, and how students can engage with a generative AI model in their work in an ethically responsible way, if such a thing exists in this space.  They incrementally explored new tools throughout the semester, culminating in an AI salon that they shared with other faculty in the Department of Theatre and Dance.  

I also use ChatGPT in the context of THE-111 Introduction to Theatre.  We’ve started exploring the boundaries of AI as a tool for creativity – what kinds of “plays” does it write?  How does its aggregate of information distill theory into practice?  Where does it fall short?  One of the things I say every time I use AI in learning spaces is that “AI is a springboard for creativity, not its replacement.”

One tool we will be diving into in class is the Diffusion Bias Explorer. It basically unpacks how the generative art models view certain professions and personality types in terms of gender and ethnic identity. 

The biggest worry in my experience is how it reinforces systemic biases and power structures.  The Diffusion Bias Explorer is a tremendous tool to explore those ideas, but it manifests in little ways too.  In class, we had a student enter the prompt “running late for class” and it rendered a woman.  I wondered if Midjourney was making assumptions about the gender identity of the user, so I entered the same prompt and got the same result.  In the absence of clarification, it will lean on its own collection of input to render a result.

Furthermore, when I put a picture of Mary (our chair), and added the prompt “chair of the department of theatre and dance” it turned her into a man. [This shows how] it’s a messy aggregate of so many systemic biases.

The implicit worry is that it can engender corner cutting, not richer work.  How do we find the way through that doesn’t vilify the technology, but also acknowledges its plentiful limitations, its boundaries, and its limits on the ways that it can streamline or augment academic pursuits?

Strategies

Has the development of generative AI changed how you teach?

Several times, I’ve used an exercise from my friend Will Lowry, who teaches Theatre and AI at Lehigh University, where we explore the ways that aggregating data can sometimes miss the point. It uses two made up image conceits, shows examples, then asks the class to render their own versions of those images based on the limited references they have, in the same way that Midjourney or DALI might. It’s revelatory, and introduces the concept of AI limitation and boundary to the students in a compelling way.

Before I began incorporating AI into my classes, I started using it in my creative work as a communicative tool, and as a means to explore the visual landscapes of the plays I was working on. This exploration also gave me an opportunity to test its limits, and to at least develop my own moral compass on its use. It also led to my creating and implementing a Midjourney component to my THE-252 course, using eight low-stakes generations as part of the exploration of the way light and image intersect.

This semester, in THE-111, I’m exploring ChatGPT more thoroughly, culminating later in the term in an exercise where we’ll have the AI platform write a play in collaboration with the students, again as an opportunity to test the boundaries of the program, and to explore its implicit biases. I’m also exploring the role of AI and creativity with my courses so I can develop and refine my own perspective, filtered through their experiences, not unlike the ways a generative AI works!

Impact

How have students responded to these changes? What have you learned and how will it shape your teaching moving forward?

Some students have responded very positively to the introduction of AI to the courses, and most use the tools already, so it feels like an extension of their lived experience. I’m eager to teach a full course on AI and creativity, so that we can spend the time it felt like we wanted to spend, without distracting from other course outcomes (this tension revealed itself in course evaluations, with some reacting positively and others feeling it distracted from the matter at hand).

I think that responding to industry trends is essential to my pedagogy, so I will continue to explore these tools in all of my courses, but finding ways to more elegantly connect them to the courses is my new priority!

Lessons Learned

What advice do you have for others who are asking similar questions and considering similar changes?

It does feel like a runaway train sometimes in the classroom, because it’s fun to play with and works quickly, but it also doesn’t understand anything, it only aggregates details. Basically it tells you what it sees, and then gives you the option to generate based on a set of choices made up of details it clocked, which also provides a window into how it engages with details. With image generators, it’s important to consider the plentiful limitations when using them, as well as the ethics of what you’re asking it to do. But it also helps to notice that, for example, it doesn’t really know what a hand is, it more understands that humans have something in that spot…which is why hands in v2, v3, and even v4 (of Midjourney) were straight up horror movies. The same thing happens when you ask it to “design a set” for a play. It might not know the play, but heck it’ll try anyway. It clearly knows what a theatre set is, but has no sense of storytelling or composition.

Disciplinary Insights

How might your discipline help students prepare for a world in which AI is ubiquitous?

Some corners of the theatre world are at the vanguard of these tools, and I was introduced to AI in the context of collaborative storytelling. Storytelling and art are excellent introductions to AI in my opinion, because it’s where you can see the cracks most clearly. Art requires subjective engagement, and soul, and these are things that AI models haven’t figured out just yet. They produce things that “feel” complete, and can even replicate the work of famous artists, but they exist only to appease the writer of the prompt; though the rabbit hole of what does art become when it enters the world of audience is another matter entirely. I focus on the connection between artist, medium, and meaning in the classroom, and hope to deepen these connections in future courses.

Responsible use is my primary concern with classroom integration. Students are going to use these tools no matter what, so it is important to use them well in my opinion. By using them alongside our students we can understand what parts of the classroom workflow they might attempt to circumvent, and find new pathways to learning and understanding!