“I built those ethical constraints first, and then I started exploring what AI could do.”

Part 1 of an interview with Kirsten Davis

About the Interview series

As part of my work in this space, I seek to highlight the folks I’ve been in conversation with or learning from over the last few years as we navigate teaching and learning in the age of AI.

If you have experiences around AI and education in higher education classrooms that you would like to share, consider being interviewed for this Substack.

---

Kirsten Davis is a Professor of Law at Stetson University College of Law in Gulfport, Florida. Davis has studied legal communication, education, and ethics for two-and-a-half decades. She recently served as the Provost's Faculty Fellow for Generative Artificial Intelligence and Higher Education at Stetson. Her current research focuses on the ways in which generative artificial intelligence impacts legal writers and readers. She founded the Legal Writing and Generative AI Convo Group, an online monthly forum for law faculty interested in generative AI and legal communication. She is the 2026 recipient of the Association of American Law Schools Section on Legal Writing, Reasoning, and Research Section Award and the Legal Writing Institute’s Mary Lawrence Innovation Award.

---

Lance Eaton: What does your own use of generative AI look like? Are you using agentic AI yet?

Kirsten Davis: Agentic in the simplest sense of the word!My two main ways of using AI are instrumental and exploratory. I use it personally in all sorts of ways: two or three different frontier models, based on their particular usefulness, and some law-domain ones, trying to figure out how the legal-specific tools can help lawyers. I tend to use Copilot frequently, which is connected to my Microsoft environment, to find things in my own work or files. I probably use that every day.

I use AI for researching, finding leads into information, and building small agents, usually around writing, like, “Can you help me figure out how to give useful feedback on this?” Thinking about feedback assistance is my most recent exploration.

My exploration is concerned with how might a law student use this tool, and what’s good or bad about it? Or how might a lawyer use this? It’s a kind of informal research. So I’m using AI every day in some way to help with my work.

Lance: One thing I keep thinking about is that it’s a tool that intersects with thinking. It doesn’t think, but how we engage with it reflects how we think. What does exploring look like for you? What might a chat log look like “let me see if it does this”?

Kirsten: I’ll give you an example. My guiding phrase for teaching right now is “sustained critical engagement.” I’m teaching the First Amendment this semester, and so I’m asking, ”How do I get students who want to explore with AI, or who might already be using it regularly, to use the tool in a way that promotes sustained critical engagement with the course content?”

Because I also teach writing, r, I think about writing in two ways: learning to write, and writing to learn. In the First Amendment class, it’s really about writing to learn. So I’ve been trying to develop a small assignments, at least regularly if not weekly, that encourage writing to learn as a way to promote sustained critical engagement with the course material. I use AI to help me think about a First Amendment topic and how I might encourage sustained critical engagement.Here’s a specific example of what I’m exploring, starting with a First Amendment case. New York Times v. Sullivan is a Supreme Court opinion about the standard for proving defamation in connection with public officials and public figures. The case established the “actual malice” standard, which requires that a a public official or public figure prove that a defamatory statement was made with knowledge of its falsity or with reckless disregard for the truth. This is a much higher standard than what is normally required in defamation suits, which is commonly negligence.

I asked AI for a way to help sensitize students to the difference Sullivan makes to the defamation standard of proof, and to experience what“real world” impacts it might make on speech. And I wanted to use AI as the way students would engage in the exercise. Using AI to help me, I developed a prompt that would ask the student to come up with a criticism about a public official. The AI would then ask k more and more pressing questions about the truth of the student’s criticism, pushing the student to experience what it might be like if a less-forgiving negligence standard applied instead of the Sullivan standard. aStudents reported that engaging in that written given-and-take process was effective in helping them ”feel” the difference between the standards and emphasized for them what would how confident you’d have to be in your criticism if the traditional standard applied.

So that was exploratory on my end: thinking about what I could design that would let the AI be involved, in a writing process and then seeing how well it worked as a way to broaden students understanding of something in the law.

Lance: Interesting! Ok, you had mentioned in your use, the f-word (feedback), can I ask where you’ve been exploring with that?

Kirsten: It’s limited so far, because I’m reluctant. But I’m experimenting with low-stakes assessment. Not for grading; I’m responsible for the grading. The question I’m exploring is whether the quality of my feedback, or the turnaround time, could improve with AI assistance.

That is, could I build a bot where I uploaded some materials and student writing, using enterprise-protected tools provided by the university, and get a head start?on constructing feedback? I decided that if I was going to try this, I needed a “code of ethics.” So AI and I worked on developing some ethical constraints: AI would never generate feedback without my review; I’d ensure the feedback was accurate; I’d look at the assignment first and conduct my own review, gathering my own thoughts; I’d be familiar with the problem and experienced in analyzing it. I built those ethical constraints first, and then started exploring what AI could do. So far, my experiments with using AI for feedback on legal writing requires a lot of intervention on my part. AI will still fabricate content, even though it’s been given relevant materials, and even when I refine the bot’s instructions. AI doesn’t “see” deeply when it comes to legal analysis. AI is good at identifying and giving draft feedback for some of the common writing problems, and because I don’t have to rewrite that, the AI’s help gives me room to dig deeper.

In this process, I’ve found that something interesting happens in the feedback process. Anyone whose job it is to give written feedback will tell you that process is tedious and cognitively intense. With AI, however, I’ve found that the three-way dynamic–between the original text, my thinking, and the AI’s generation of content–changes my engagement. I’m interacting with that in a three-way relationship that’s a more cognitively interesting. I’m curious about my own interaction with the text.

So far, using AI is not a huge time savings, for the process but, as I said, the assistance increases my ability to go deeper with the comments I want to make and lets me give the student more robust feedback. I’m not experiencing anything revolutionary, just seeing how AI might improve productivity and quality. So far, I’m finding that AI can improve some of the quality of my feedback, in part because of its ability to generate words that would take my fingers more time to produce. But it’s certainly no replacement for my engagement or effort.

Lance: It makes me think about a workflow where I say something verbally, in shorthand, and it translates that. And it makes me wonder what that feedback would look like across different models. I wonder what it would look like with NotebookLM, which is a very different, interactive output model. So that’s less a question and more, now I’m off to experiment.

Kirsten: I think there are all kinds of ways we might reimagine feedback with AI. In first-year legal writing, the writing is highly structured, and has expectations that should be present in all assignments. In that context, I’m thinking about AI to facilitate my own feedback, keeping me significantly in the loop, rather than as an interactive tool a student would use. But I think there’s definitely room to create tools for students to use themselves. For example, I’ve always used comparative analysis, where a student takes their own writing and, through a structured set of steps, compares it to a model or sample. I wonder if AI could make that more enjoyable and effective. But again, the variability in the outcomes AI produces is what limits translating a comparative analysis into something more independently interactive through a bot. That’s the part I’m not doing yet: reconceptualizing what self-guided feedback might look like with an AI tool I provide to students. But that’s definitely on the table for exploration!

Lance: How else is generative AI showing up in your courses? You’ve shared a few; I’m curious what else.

Kirsten: It shows up not only in relation to pedagogical process but also in the law substantively. For example, we’re at the end of the semester in the First Amendment course, and the question comes up should AI speech be protected by the First Amendment, and in what context? These kinds of questions–ones that involve how the law applies to AI or how AI impacts the law–are showing up in all topics.

But I’ll set that aside, because what’s more interesting, I think, is the question that is coming up in legal writing and legal communication classes, where does learning AI skills belong? In a foundational, first-year setting? Or, in upper-level courses, after you’ve learned some foundational ln skills?

The timing is a subject of controversy in the discipline. Some think AI doesn’t belong at all in first-year courses; some think it might replace content we’ve always taught. I’m more middle-of-the-road: I think the foundational content traditionally learned in the first year now has to be integrated into a new way of thinking that allows you to make sense of AI as you apply that foundational content later.

Another question we have to ask is what is the role of AI in student work when students are trying to learn those foundational skills? For me, this is a question of assessment, as opposed to learning outcomes, though I think there are some new AI learning outcomes for the first-year, first-semester student. So AI is forcing me to ask, how am I going to assess students in my course? Do I stack AI skills on top of the foundational ones, or integrate them together? And what can I anticipate students doing with AI when they are under pressure? I call this the “big button problem.” Are they going to use that big AI button and not think about what they’re doing? Or are they able to stay engaged in the learning process while using AI? That’s back to the sustained-critical-engagement problem.

Ultimately, I’ve decided not to I ban AI in my classes. Instead, I’m working to redesign assessment to capture what students capable of doing with and without AI and to encourage them to think about the ethics of AI use.

AI is also showing up in my courses because I hear students talk about AI in a conflicted way. They recognize AI is available and will probably be required for legal practice in various ways. But there’s a consternation, even resentment, in this generation of students who rightly recognize they’ve invested a lot in learning how to do things and develop expertise without AI, but now they face the possibility that AI will make some of that work and expertise less important. I think they’re right to be concerned, but I’m also still skeptical about the degree to which AI can accomplish what a thinking human can, and I try to help my students prepare for that future of using AI tools while still bringing human expertise to the work

So AI is showing up everywhere, and it’s biggest impact is in in my legal writing classroom, where I’ve essentially had to redesign the courses to make sure they get what they need as graduates right now.

Lance: And these are all first-year law students?

Kirsten: For the most part, yes, all first-year law students. When I say first-year, I mean the 1L year.

Lance: As we’re navigating this, what are you finding most curious or intriguing about AI?

Kirsten: I think about AI in the context of both legal education and legal practice. And what’s intriguing to me in both contexts is the degree to which it can “do” or mimic legal analysis. This result requires rethinking, what it means to produce legal communication “on your own.”

The idea of “on your own” is a big issue, , particularly in legal education, where we’re trying to assess a human’s thinking,: analytical, evaluative, and research skills. What “on your own” means is intriguing now. What does it mean to possess knowledge, or outsource its production to a machine or, alternatively, possess it in a hybrid way? Are those different categories? That’s really interesting.

This plays out not only for students but also in context of tenure and tenure-track faculty work; that is,, what does it mean to produce scholarship, and what are the ethical boundaries for doing that work? I’m sure you’re hearing that issue arise in other fields, but in the humanities, in which I’d include law, the questions are sometimes different. The value in humanities scholarship often seems based on the author’s ability to generate ideas; so, that leads me to ask, should generating ideas be done “on your own,” without AI, in that context? Or it is okay to use AI for brainstorming scholarship?

On the other hand, scholars always used other sources;, building on other ideas, getting ideas from the work others have done. So, I think legal scholars have to ask what it means to produce scholarship “on one’s own” and what the ethical boundaries of that work are.

When you turn to practice, for lawyers, things look different. The goal of the lawyer is to get functional document. Do we care how that document is created as long as it serves its purpose?? In many ways, no, except perhaps that we care whether the lawyer has treated confidential information appropriately. What we care about is that the document does the work expected. So, for the purpose of practice, are we training new lawyers to be able to do the work necessary to get the document, regardless of how it’s generated?

This question is not entirely new. Lawyers have always had forms and brief banks from which to borrow ideas and text. This is true for judges, too, but judges also have to be worried about distinguishing between the adjudicative and administrative functions when using AI.

Here’s what that means. Litigants rightly have expectations about how a judge decides a case; that’s the adjudicative function, to decide cases. When a judge uses AI for adjudication, numerous questions arise. Should they use AI for brainstorming? To write part of the opinion? To edit? Where does judge’s independent judgment, which is required by judicial ethics, begin and end when using AI? And if a judge has clerks writing draft opinions for them and the clerks are using AI, what are the limits?

The questions of adjudication aren’t just about keeping information confidential–although judges have those worries ,too. It’s about the rule of law and the integrity of a judicial system.

Lance: It’s adjacent but it reminds me of Congress passing thousand-page bills they put together the night before. The scale of that is now possible with AI.

Kirsten: Right. And the legal system really cares about the independent judgment of the officers of the court, including judges. State bars have made clear preserving the independent judgment of the lawyer matters when writing with AI,, and that’s also true of judges when they adjudicate claims. The duty itself is not a point of controversy. The legal system, at least so far, expects the independent judgment of humans. But how does that play out “on the ground,” so to speak how are those expectations enforced, and what are the boundaries for these kind of uses? We don’t have great answers yet

Lance: What you need is some AI. No, I’m kidding. You need some bots, right?

Kirsten: To help me figure this out? You might have a point.

Join us for part 2 in the next post!

---

The Update Space

Upcoming Sightings & Shenanigans

• Keynote speaker at the Reimagining the Liberal Arts in the Age of AI Conference, July 21-23 at the University of Mary Washington.

• Panelist on a general sessions panel (AI Leadership Choices: Building the Ecosystem from Foundation to Scale) at UPCEA’s SOLAR conference in Boston on July 30 at 2:30pm.

• EDUCAUSE Online Program: Teaching with AI. Virtual. Facilitating sessions: ongoing

Recent Recordings, Resources, & Writings:

• Davis, L., & Eaton, L. (May 2026). Expanding OER with GenAI. EDUCAUSE Review.

• AI x Higher Ed Podcast with Anand Rao & Stefan Bauschard. Episode: Universities Must Adapt to AI—Here’s How They’re Doing It (May, 2026)

• Damm, C., & Eaton, L. (2026, March). From prompt to practice: A framework for transparent GenAI use in higher education. EDUCAUSE Review.

• Eaton, L., Nemeroff, A., & Sun, X. (2026). AI-assisted course design and development. In K. S. Ives, M. Cini, & R. Schroeder (Eds.), AI applications in online higher education administration: Strategies for maximizing returns and improving outcomes. Routledge.

• Margin of Thought with Priten: Season 1, Episode 5: How Can We Center Pedagogy During the AI Tech Wave? (February 2026)

• Online Learning in the Second Half with John Nash and Jason Johnston: EP 39 - The Higher Ed AI Solution: Good Pedagogy (January 2026)

• The Peer Review Podcast with Sarah Bunin Benor and Mira Sucharov: Authentic Assessment: Co-Creating AI Policies with Students (December 2025)

• David Bachman interviewed me on his Substack, Entropy Bonus (November 2025)

• The AI Diatribe Podcast with Jason Low (November): Episode 17: Can Universities Keep Pace With AI?

• The Opposite of Cheating Podcast with Dr. Tricia Bertram Gallant (October 2025): Season 2, Episode 31.

• The Learning Stack Podcast with Thomas Thompson (August 2025). “(i)nnovations, AI, Pirates, and Access”.

• Intentional Teaching Podcast with Derek Bruff (August 2025). Episode 73: Study Hall with Lance Eaton, Michelle D. Miller, and David Nelson.

• Dissertation (2025): Elbow Patches To Eye Patches: A Phenomenographic Study Of Scholarly Practices, Research Literature Access, And Academic Piracy

• AI Syllabi Policy Repository: 200+ policies (always looking for more- submit your AI syllabus policy here)

Finally, if you are doing interesting things with AI in your higher ed classrooms, consider being interviewed for this Substack or even contributing. Complete this form, and I’ll get back to you soon!

We periodically host small-group workshops and leadership sessions for higher ed teams. You can learn more about our current offerings here.

---

AI+Edu=Simplified by Lance Eaton is licensed under Attribution-ShareAlike 4.0 International