How The Kinkaid School’s long standing interest in AI has led them to partner with Flint

What trends you have seen in ways that people at Kinkaid are using Flint, specific departments that have embraced it, etc.?

We're seeing it in a variety of different ways. I really like Flint’s primary developmental focus using learning via tutors (now called activities). To me, it is very similar to the development of custom GPTs, but it's much more accessible and easy to build as compared to a GPT or Anthropic project. We're seeing people begin to dip their toes into the water and explore. The 2022-23 school year was a pre-alpha year, last year was our alpha year, and this is our beta year.

We just closed a faculty and staff survey, which is the third survey that we've done with the same questions. One of the questions we added to this year's survey was, “Are you planning on integrating Flint within your classes?” This was targeted for the Middle School and Upper School. The responses were 25% yes, 25% no, a little less than 50% maybe. They're not sure yet. They've got to see what their colleagues do and are beginning to share amongst themselves.

Some of our science teachers have begun to build tutors around lab report writing. They're generating questions to replace some of the worksheets they had done previously. We're seeing 7th graders who are interviewing historical characters. We're seeing a lot of help with writing across the board, although we do have our skeptics and people who just aren't ready at this particular point in time. 

We've got an entrepreneurship course that is developing a tutor to help with ideating a passion project. I just got a request yesterday to spin up a group for our 9th-grade learning specialist who wants to build a tutor around the concepts of planning and time management. 

A sixth grade class used the image generation capabilities of Flint to develop visualizations of the character within the novel, Esperanza Rising. This project focused on the student’s developing the skills of close reading character analysis, and writing from a character's point of view. The use of image generation in Flint elevated the experience in some fresh, relevant ways. The students not only enjoyed creating the images of their characters, but they learned a lot about effective prompt writing and using AI as a collaborator. 

We've got a lot of ideas, and we're going to be supporting those ideas. 

Again, because of the way the interface is targeted, we're not really seeing work being done operationally at this point. We’d like to explore how the non-academic subjects in our Advancement, Admission, and Strategic Communication offices can use AI. For example, within HR, how are we utilizing AI to build efficiencies? I do know that the business office, our controller, as well as our CFO have begun to dabble a little bit more, but we don't have widespread adoption. I want to review those numbers out of the survey. But we haven't spent the time with AI on the operations side, because they don't fit into a regular departmental meeting structure.  Therefore, they don't get the same attention. That's an area I want to develop.

Math and science support have been behind a bit when it comes to LLM development. Flint has some additional features in place to support these subjects, so have you seen any interesting math and science usage come about?

Oh, for sure. Absolutely. That’s where some of the interesting growth is. Our physics teachers did a lot of work over the summer and have some really great ideas. We have a marine science teacher who mid-year last year was doing phenomenal work in terms of building rubrics and doing all this classwork with GPT3.5. We asked if she wanted to try Flint and she's just really taken off with it. She was one of our early adopters with students last year. Teachers are looking for ways to integrate this technology. 

Also, I want to share our Generative AI Menu with you. What we've done for this upcoming year is we have articulated how AI can be used on various different assignments. And again, this is version 2.2.

Earlier this fall, we conducted a Lunch and Learn with our Upper School teachers and shared the original version of this menu. We have rolled this out to our Middle School faculty as well. During this Lunch and Learn, some of our math and science teachers came up to us and said, “This is really good and really helpful, and this doesn't address all the ways that we might want to use this within math and science.” They started to make suggestions, which we told them we were open to. We decided to let them as departments first begin to try to define what usage looks like and then share it with us. We recently got their articulation, which we will soon thoroughly go through and vet to have conversations. Then, we will add the various different sets of tools to the menu. That’s the kind of collaborative ethos we're building.

I like to use this analogy to sum it all up. You learn when you drive that if you hit ice you're supposed to turn in the direction that you're skidding. It’s kind of the opposite of what you would think. We gave that example—which is really odd here in Houston because we don't deal with ice—but you don't slam on your brakes because then you lose control. You don't turn to try to correct because then you oftentimes overcorrect and cause a greater spin. You turn into the way that you're going to gain control. So, one of the terms we're using here is “we're turning into the skid.”

Usage is a lot stronger this year than the initial roll-out last year, what do you think is bringing teachers back to Flint this year? What besides the initial excitement are they drawn back to?

Usage is a lot stronger. Again, last year was our alpha year of development and we are considering this year our beta year. We had Flint come in during our end-of-year meetings to meet with a number of our faculty who were beginning to explore. Before they headed off to the summer, we wanted them to know we had made the decision to partner with Flint for the upcoming year and give them a chance to think about their experience as well as some of the work that we had done in the fall. So that's one of the reasons you're seeing an increase in usage.

What have you all been doing on your side to get teachers involved?

You know, we're letting it grow organically. We're not requiring it. It's really teachers sharing with other teachers. We do have regular meetings within the divisions to talk about and share what we've learned. We also provide times for purposeful play so that people work within it. One of the great things from our survey this year is that it showed, in the last six months, we started from a third of the people saying they have used generative AI personally or professionally to it now being 90% of our faculty and staff. And we had 205 people reply, which was our highest response rate yet.

We've provided this awareness through very thoughtful times. In fact, this afternoon I’m presenting the high level of the survey results to our administrative leadership team. Since they’re not all in one of the academic departments, they've not seen what we've done. So I'm going to share some of the results from the survey and then give them some time for some purposeful play. We're going to talk a little bit about prompting and what types of things they need to think about within their prompting and then give them a small activity exercise to work on and explore. 

Because again, if they're not asked to do that, they're not always thinking about ways to bring AI into their workflows. They will continue to do what they're doing. So we take the use of the time that we can. We've got Lunch and Learns scheduled every three to four weeks in the Upper School when we just allow people to come in and work because it's just about setting aside time.

Course development takes time. Developing a good Flint activity takes time. You can hack one together, and Flint helps automate some of the process, but to really get it to do what you want it to do takes some rule tweaking. That’s why we’re focused on providing the time and space for people to get together, figure out what is working and what isn’t, suggest rules to add to the AI’s behavior, and discuss ideas and learning. It's that organic sharing.