Running a Faculty Panel on Generative AI
Changing the format; improving the experience...
Over the years, I’ve gotten to organize, moderate, and be on lots of panels. They can be really interesting and exciting experiences where distinct and rich ideas get to interact with a flow that just doesn’t happen when you have a singular speaker.
This past week, I was asked to lead a panel of faculty and instructional designers about their usage of generative. Part of the goal is that this was my third visit to this university over the past year and I really wanted to flip the script a bit more from me being the outside expert to highlighting the wisdom, insights, and practices of the faculty at the institution.
In fact, it’s one of my primary messages when I talk and do workshops at different institutions: I’m a facilitator of the insights and abilities already of the room—I’m not the fountain of knowledge. I want folks to walk away less with thinking “wow, Lance really knows his stuff” and more of “wow, I’ve got amazing colleagues doing really cool things and I need to talk to them!”.
Thus, I was excited to have this third session where I moved from center of attention to MC to uplift the brilliance of their faculty. Now, doing a panel is great and also, I didn’t know if that was going to be as effective for what I was hoping folks would get out of it. In starting to question what the panel would look like and how it would impact the outcomes, I realized that I needed some objectives.
What follows is my process for developing the panel. It appeared to be a resounding success so I figured sharing it, how generative AI helped me (prompts & chatlogs included), and additional insights from the process.
Structuring the Panel Agenda
The first thing I needed was to get a sense of who my panelists were, so I came up with a short Google form for them to fill out. This gave me a sense of who they are and what their contributions might be. It helped me see if there would be any interesting ways of contrasting ideas with certain questions or order of questions.
As I gathered this information, I also wanted to brainstorm what the format could be. I started with ChatGPT to get some ideas for different forms—though first, I used my Prompt Enhancer GPT to refine my question. That gave me the following prompt after running it through twice:
Imagine you are a skilled panelist moderator, about to lead a 90-minute interview with a group of six diverse faculty members. Your mission is to craft an atmosphere that is not only engaging and dynamic but also ripe for deep, intellectual exploration. With this goal in mind, devise 10 innovative interview formats. These formats should be distinct from one another, each offering a fresh and unique approach to stimulate enjoyable, yet profound conversations. For each format, include the following details:
Interaction Style: Describe how the panelists will interact with each other and with you, the moderator.
Thematic Focus: Detail the central theme or topic for discussion in this format.
Engagement Techniques: Explain how this format will maintain audience interest and ensure active participation from all panelists.
Creative Elements: Suggest any unique or unconventional elements that could be incorporated to enhance the format's appeal and effectiveness.
Your aim is to enable a rich exchange of ideas, ensuring that each format varies in its structure, interaction style, and thematic focus. This variety is crucial to keep both the audience and the panelists continuously engaged and intellectually stimulated throughout the session. Embrace creativity and deep thinking in your response, showcasing your ability to think outside the conventional boundaries.
That felt a reasonable upgrade from my own prompt: “you are an expert panelist moderator. You are interviewing a panel of 6 faculty members over the span of 90 minutes. Provide 10 different formats that could be fun, engaging, and keep the conversation moving into meaningful and thoughtful spaces.”
The results (see here) were a range of different ideas about how to set up the panel. Most I was familar with and yet none of them jumped out to me as the full experience that I was hoping for. However, one of the options (World Cafe) was calling to me. I tend to like the World Cafe model, where the group stays the same and they go through a series of speakers. When you come back together there’s an opportunity to learn the similar but different conversation arcs that occur.
Thinking through the formats also helped me solidify the objectives of the panel. A conversation is always nice but what did I want out of it and how I would get it? I found myself settling on these objectives:
Learn from panelists about how they are thinking about and using generative AI in their work.
Make generative AI usage more tangible for consideration and use in teaching and learning.
Reflect upon how individuals will use generative AI for this upcoming semester.
The listing of different panels also had me thinking about one of the challenges with a panel—especially one where there were potentially going to be 6 people—there’s hardly any time to go deep with any one of them and if you’re really interested in what that person brings to the conversation, it might feel like a lost opportunity.
The structure came to me that we do a light World Cafe approach followed by a full panel. So we would start with the panelists each doing a 30-second pitch about what they plan to talk about in their breakout room and then let folks get into their breakout rooms for a rich conversation. After 20 minutes, people would be permitted to switch. Then, we would bring it back to the main room for a larger conversation.
I liked that and also knew that we were going to lose time and momentum in the transition period when folks were moving to other speakers. I decided to do only one speaking session with the trade-off being they would have more time in the one session they went to (30 minutes instead of 20).
Creating a Focus
If folks were going to be talking about generative AI in distinct ways based upon what each panelist brought into that room, there needed to be a slightly new focus to consider when we came back (and had another 40 minutes). I mean, sure, 10 minutes could be summarizing and providing highlights across the room, but there’s still more to be discussed.
That left me thinking about where the panel’s attention (now joined together) could be directed. It had to be relevant and grounded in the university and community. Again, I wanted folks to walk away from the session feeling more empowered and comfortable engaging with generative AI if that’s where they found themselves. We knew the panelists were going to share lots about what they were doing but what about the audience themselves?
That’s when I had the realization that I could actually bring the audience to the panelists in a different way. If there was time for them to do a short survey about their own thoughts, then that could be something the panelists engaged with. I took to ChatGPT again to help create a survey. First, I used the Prompt Enhancer and then asked ChatGPT to help create the survey (Chat log here) which I then turned into a Google form.
The goal would be to collect the information, process it a bit with ChatGPT while they were in breakout rooms, and present it back to folks after we shared out from the discussion.
That itself would provide good content for conversation as the panelists engage directly with questions and ith the collective thoughts of the group in front of them. I figured that would make rich conversation before turning to final questions and last comments in the last 15 minutes.
The Panel Schedule
Here was the final structure of the panel conversation:
9:30am: Introduction by Lance. I outline the session and goals, while giving a little bit of time for any folks to complete the questionnaire.
9:40am: Meet the panelists. Each panelist gave a 30-second pitch about their focus in their breakout room. (Pro-tip: I got these from panelists in advance and had them write them out to me which helped them stick to it).
9:50am: Breakouts with Panelists. Participants chose the room of interest and had a 30-minute conversation.
10:20am: Shareout. A participant from each room shares out the key parts of their conversation—1 minute max share.
10:30am: Review Survey Results. Lance shares the results of the survey and initial analysis by ChatGPT.
10:35am: Panelists Discuss Survey. Panelists share insights, reactions, and implications for the results of the survey along with occasional questions/comments from participants.
10:50am: Final questions. Any lingering questions are asked by the attendees.
10:55am: Final comments. Each panelist has a final sentence/thought to share about the conversation.
The Panel Flow
It went well and I definitely feel like there’s something here as a format to explore—with generative AI or any panel conversation, really.
I communicated with the folks leading the event and they made sure to share it with participants—we also reshared it at the start of the conversation. All in all, we had probably 95% folks attending fill it out, which was a solid completion rate. The results were quite the mix and were perfect for a good conversation.
During the breakout rooms, I was able to analyze the data. Again, I used Prompt Enhancer to improve my questions (Chatlog here) and then get this analysis from ChatGPT to share back with folks. These materials I put into a Google Doc that I had shared with them earlier.
I think the panel’s success derived from its different structure—one that required a bit more of attendees than just listen and ask questions. It required a bit more of them at times including the survey and deciding which room to go to. But it also reflected back at them who they were and what they were figuring out with generative AI. Sharing and discussing the results of the survey wasn’t an abstract survey but provided a sense of who they were and what they were figuring out.
Besides the energy of the conversation, it was clear from the chat that folks were really interested in the conversation, sharing their ideas and learning from one another. It was one of those sessions where it was hard to keep track of both the vocal and the textual conversation.
Moments like these in working with faculty is part of what gives me hope. It’s why I try to bring the optimism to these conversations. In many ways, it’s like our classrooms and how we work and think about students. If we can create the right conditions, make folks feel valued and recognized, and then create opportunities for them to explore, learn, and share, it can make a difference and better foment a community of learners.
It’s part of what gives me hope and why I try to bring some optimism to these conversations. Just like we believe with our students that if we take on something in the right way and make it inviting and accessible, we can
Question for readers
Have you been doing any faculty panels? Have you played with the format? In general, how are faculty finding faculty-based panels around generative AI?
AI+Edu=Simplified by Lance Eaton is licensed under Attribution-ShareAlike 4.0 International
Very nice idea, Lance. It's in the vein of generative AI as collaborator.
I've been on some faculty panels about AI, as well as seen others. Format... nothing unusual yet.
Lance,
This is an amazing walk-through of the process, how you gathered, used, and analyzed data, and how you worked as both a mirror and a lamp in this panel discussion! Thank you so much for taking the time to share this in such a specific and interesting way.